New Modalities and Carcinogenicity Assessment

Considerations for Adeno-Associated Virus Gene Therapy, Insertional Mutagenesis, and Carcinogenesis Potential

Adeno-associated virus (AAV) gene therapy has emerged as a promising strategy for the treatment of monogenic disorders. This approach has been successfully trialed for the treatment of the inherited bleeding disorders hemophilia A and B. These pivotal studies led to European Medicines Agency (EMA) and Food and Drug Administration (FDA) approvals of Etranacogene dezaparvovec for hemophilia B and Valoctocogene roxaparvovec for hemophilia A. Despite these advances, there are questions on the cellular mechanisms via which AAV vectors persist and whether this treatment could lead to long-term safety concerns.

Wild-type adeno-associated virus (wt-AAV) is a replication-deficient single-stranded DNA virus, with a genome consisting of Rep and Cap coding regions, flanked by inverted terminal repeats. Although wt-AAV primarily persists in extrachromosomal episomal forms, site-specific integration mediated by the Rep78/68 complex has been characterized. The wt-AAV is considered to be nonpathogenic based on a high seroprevalence of AAV antibodies in the population. These assumptions have been challenged in a recent study that reported wt-AAV integration in a small number of hepatocellular carcinoma biopsies. There are conflicting findings in this area, and further research is required to evaluate whether wt-AAV integration may play a minor role in this process.

Recombinant adeno-associated virus (rAAV) vectors are produced by replacing the viral coding regions with a promoter/enhancer and gene of interest. These fundamental differences limit the translatability of studies of wt-AAV. Early murine studies demonstrated that rAAV predominantly persists in an episomal form. In some of the studies, however, rAAV integration was seen following both local and systemic delivery. In some murine studies, there has been evidence of insertional mutagenesis associated with recurrent integration events into the murine Rian locus. Studies have been conducted in large animal models in order to gain a better mechanistic understanding of rAAV vector persistence and potential mutagenesis. These have demonstrated long-term therapeutic expression with no evidence of tumorigenesis.

One study in the hemophilia dog model demonstrated predominant episomal persistence, with low integration frequencies in areas of chromatin accessibility. Detailed analyses using long-read sequencing demonstrated only fragmented integrated forms, supporting the hypothesis that transgene expression is derived from episomal forms. In another similar study, hepatic integration with clonal expansion of hepatocytes containing rAAV vectors was identified, albeit without tumorigenesis.

A recent study has challenged the concept of transgene expression only from episomal rAAV. In this study, full-length integrated forms were identified in samples from nonhuman primates treated with different rAAV constructs. Data from a small number of human liver and muscle biopsies from individuals treated with rAAV revealed, similar to large animal models, the presence of both episomal and integrated rAAV. ¹ Integration events in these samples occurred at low frequencies with no evidence of recurrent integration sites, insertional mutagenesis, or clonal expansion. The demonstration of integration in pre-clinical and clinical samples has led to theoretical concerns that integration could result in insertional mutagenesis. Integration site analysis has been conducted on biopsy tissue as part of the investigation of seven malignancies reported in clinical gene therapy trials. ² These studies have identified no evidence of insertional mutagenesis.

In summary, data from pre-clinical and clinical studies of rAAV vectors have shown that although AAV vectors predominantly persist in episomal extrachromosomal forms, rAAV integrates into the host genome at low frequencies. Integrations site analysis in large animal models and clinical biopsies has shown no evidence of insertional mutagenesis to date. Although these results are reassuring, there is a need for further research into the relevance of rAAV integration to help inform long-term outcomes.

Oligonucleotides: Evolution of Carcinogenicity and Risk

The potential for oligonucleotides to cause cancer has historically been evaluated through 2-year bioassays in rats and mice. No human-relevant findings have been observed to date. More recent oligonucleotide programs have successfully leveraged data from approved products to streamline carcinogenicity assessment, including replacing the traditional 2-year mouse bioassay with a 6-month transgenic mouse study.

Recent draft guidance for nonclinical safety assessment of oligonucleotide therapeutics and platform technologies may provide further opportunities to streamline carcinogenicity assessment of oligonucleotides in the near term. This presentation provided an overview of carcinogenicity assessment approaches and findings in rodents, followed by a discussion of proposed opportunities to limit carcinogenicity assessment of future oligonucleotide-based therapeutics when scientific justification supports it.

Cell Proliferation in Rat Mammary Glands as a Method to Assess Potential Carcinogenicity

Following the evaluation of carcinogenicity risk evolution in the field of oligonucleotides, a case presentation was jointly presented in which the utility of cell proliferation (CP) in rat mammary glands in the evaluation of the carcinogenic potential of insulin was challenged.

In 1992, the publication of a study showing the carcinogenic effect of the insulin analogue, insulin X10, in rat mammary glands raised concerns regarding the increased cell proliferative effect of modified insulin analogues in vivo. ⁷ Consequently, EMA started to recommend assessment of CP, and assessment of CP in the mammary gland has often been included in nonclinical studies evaluating the carcinogenic effects of new insulin analogues, in some cases using X10 as a positive control.^1,4,5 The value of this CP assessment has been challenged in recent years, particularly because histopathologic examination of mammary gland tissue and identified macroscopic changes are now included in more recent assessments.

The authors compared results from four nonclinical rat studies evaluating the effects of repeated dosing with insulin detemir or degludec in mammary glands. ² Studies included human insulin and/or X10-dosed groups as comparators, CP, and histopathologic evaluation of mammary glands. In addition, data from an investigative rat study with repeated dosing of human insulin and X10 was included in the assessment. The results of the studies indicated that neither human insulin, insulin detemir, degludec, nor X10 induced mammary tumors or increased CP. In agreement with published data in rats, CP fluctuated during the estrous cycle, and thus, the applied methods were able to identify changes, if present.^4,5

Based on the four studies evaluated, the authors found that the mitogenic effect of insulin in rat mammary glands is weak and that in vivo CP evaluation in nonclinical studies was not predictive of the carcinogenic potential of insulin; thus, the value of including this endpoint is debatable, as is the inclusion of X10 as a positive control.

Carcinogenicity Risk Assessment of Protein Degraders

Targeted protein degraders (TPDs) have recently emerged as a novel drug class. Their mode of action utilizes E3 ubiquitin ligases to degrade therapeutic proteins of interest selectively. This modality offers excellent potential for targeting previously intractable drug targets. This presentation addressed the biology, unique characteristics, and aspects of potential carcinogenicity assessment of the two main TPD classes, the molecular glue degraders, which have drug-like properties similar to conventional small molecules, and proteolysis targeting chimeras (PROTACs; also termed heterobifunctional TPDs).

The cereblon E3 ligase complex is the most widely utilized complex in drug discovery, and several cereblon-based TPDs are in late-stage clinical development for oncology indications. A significant safety consideration for cereblon-based TPDs is the degradation of unintended neosubstrates, in light of the teratogenicity findings associated with thalidomide and other immunomodulatory imide drugs, which is thought to be mediated, at least in part, through a cereblon-neosubstrate mechanism.^3,6 Due to the inherent complexity of TPD-mediated cereblon-neosubstrate binding, unintended neosubstrates are challenging to predict.

As cereblon and other E3 ubiquitin ligase complexes are employed in drug discovery for non-life-threatening indications, carcinogenicity risk assessment will be required. There is no regulatory requirement to treat this new drug class differently from conventional small molecules in carcinogenesis assessment. However, several properties of TPDs could influence the weight-of-evidence assessments and carcinogenicity study design. Three hypothetical weight-of-evidence case studies were presented for consideration and to highlight the complex challenges of this emerging modality. Aspects to include in a weight-of-evidence approach for carcinogenicity risk assessment with TPD include that, compared to functional small molecule inhibitors, toxicodynamics may be different for TPDs due to their catalytic degradation properties, and drug metabolism may be more complex for PROTACs. These aspects can and should influence the design of toxicology studies and risk assessment. Furthermore, species differences in TPD interactions with cereblon and neosubstrates can complicate translational safety, including rodent carcinogenicity studies. In summary, TPDs offer an exciting prospect for new drug discovery, but their differing properties from conventional small molecules should be taken into account in carcinogenicity risk assessment.

Carcinogenic Risk of Chimeric Antigen Receptor-T Cell Therapy

Chimeric antigen receptor-T cells (CAR-Ts) are human gene therapy products in which T cells are genetically modified to recognize a desired (mostly oncology) target for therapeutic purposes. Presently, there are 3 platforms, autologous and allogeneic (also called ex vivo), which share relatively similar costs and safety challenges, and in vivo CAR-Ts, which share similar off-target transduction concerns as gene therapy products. To date, only autologous CAR-T cell products have been approved, and the requirement for a robust carcinogenicity assessment of these products has been heightened by the perceived potential risks of second primary malignancies (SPMs), thought to be linked to insertional mutagenesis.

Although there are very limited data to support this association in the post–CAR-T setting, evidence-based solutions and scientific strategies for risk mitigation are expected by regulators. Regulators recommend that a case-by-case nonclinical testing strategy based on in silico, in vitro, and in vivo pharmacology or toxicology models be applied, as appropriate, in conjunction with available nonclinical and clinical data from related products.

The speaker presented that a weight-of-evidence evaluation should include an assessment of individual components of the CAR-T cell products and process and engineering strategies that may be linked to a potential carcinogenic risk. These include the biology of the CAR construct (specificity, signaling, additional modification, or “armoring”), vector characteristics (integration, self-inactivating, cell-specific promoters), donor-related risks (high-fidelity genomic tests, eg, whole genome sequence analysis or karyotyping), and implementation of a clinical program that includes site integration analysis or baseline screening for clonal hematopoiesis.

Although there is no scientific consensus on which in vivo models to use, studies in murine xenografts or pigtail macaques (for in vivo CAR-T) with pathology and molecular pathology endpoints can also provide valuable evidence of transduction specificity and T-cell trafficking and proliferation profile, which can be used in the overall carcinogenic risk assessment.

The session organizers hope that the attendees and readers find the information presented in this format as helpful and impactful as we found it at the symposium, and our sincere gratitude to our speakers for their time, energy, and expertise, so freely shared with the Toxicologic Pathology community.