Mutagenesis and Carcinogenesis Risk Evaluation for AAV and Lentiviral Gene Therapies

Abstract

Fueled by the identification and invention of novel gene delivery vectors, gene therapy efforts now hold promise for treating a wide range of diseases and are seen as a crucial part of growth for the biopharmaceutical industry. Currently, recombinant adeno-associated virus vectors (rAAVs) and lentiviral vectors (LVs) are the main vectors used in gene therapies that are approved or tested in human clinical trials. Meanwhile, ongoing research continuously reveals unprecedented knowledge of viral vectors on the host genome, which may subsequently affect the mutagenic and carcinogenic potential of these therapies. This article summarizes the content and addresses the commentary from the scientific symposium entitled “Mutagenesis and Carcinogenesis Risk Evaluation for AAV and Lentiviral Gene Therapies,” conducted at the 43^rd Annual Meeting of the American College of Toxicology, November 2022 in Denver, CO. The objective is to summarize the current understanding of rAAV and LV related mutagenicity/carcinogenicity risk, describe the methods and interpretation of results to guide risk assessments, as well as the current regulatory landscape on the carcinogenicity and mutagenicity assessment of rAAV and LV gene therapy products.

Background of the Symposium

Over the last few years, several gene therapies have been approved by the US Food and Drug Administration (FDA) and European Medicines Agency (EMA). ¹ There were more than 1,000 active gene therapy (GT) trials listed by Clinicaltrials.gov at the time this symposium occurred, targeting a wide range of unmet medical needs from oncology to recessive genetic disorders.^2,3 The use of viral vector in GT has several safety concerns in clinical trials including the risk of insertional mutagenesis. ⁴ Regulatory agencies recommend characterizing the risk of insertional mutagenesis for viral vector-based GT products for various indications.^5-7 In response to this potential risk, the scientific symposium, held in November 2022, as a part of the 43^rd Annual Meeting of the American College of Toxicology, brought together industry and regulatory scientists, to discuss and contemplate ways to address the evolving challenges in the GT field.

The symposium served as an introduction for scientists who are interested in obtaining background information around GT, and as a comprehensive overview of the history and current scientific knowledge on recombinant adeno-associated virus vector (rAAV) and lentiviral vector (LV) related mutagenicity/carcinogenicity risk. In addition, case studies and scenarios, using either rAAV or LV as examples, demonstrated how to assess this risk in applicable drug development programs. Lastly, a regulatory agency representative presented his view on assessment of mutagenicity/carcinogenicity of gene therapies. The individual presentations are summarized below.

The opinions expressed here are those of the authors and do not necessarily reflect official policy, nor strategies of the authors’ affiliated organizations.

Introduction

Vector-mediated genotoxicity may occur when components of a viral vector are unintentionally integrated into the genome (Figure 1(A)). This event may potentially result in the activation of protooncogenes (Figure 1(B)) and/or insertional mutagenesis that, in turn, may cause the disruption or dysregulation of gene expression leading to clonal expansion and oncogenesis.

Figure 1.

Vector Mediated Genotoxicity. (A) Insertion of unwanted gene from viral vector (orange) into the host genome near oncogenes (green or blue), and (B) insertion of the unwanted gene may lead to activation of oncogenes through various mechanisms.

Integration of vectors into the genome involves both viral and non-viral factors. ⁴ The viral factors include vector design (presence of strong enhancer and/or promoter sequences, presence of splice-donor or acceptor sites, and polyadenylation signals favoring generation of alternative transcripts), and host integration site (IS) or insertion site profile. The nonviral factors include transgene function, target cell, disease state, age of recipient, and epigenetic determinant such as presence of histone modifiers, cell culture condition, and proximity to DNase hypersensitive sites (HSS). ⁴

There are four proposed mechanisms for vector-medicated genotoxicity as summarized in Figure 2. These are promoter insertion, promoter activation, gene transcript truncation, and epigenetic gene modification. ⁴

Figure 2.

Four mechanisms of vector mediated genotoxicity. Box/arrows in purple represent vector and blue represent host genome. (A) Promoter insertion upstream of cellular transcription (up), or into the promoter and 5’ of target gene (down); (B) promoter activation through insertion near the target gene (up), or into the 3’ untranslated region (UTR, down); (C) intragenic integration leading to loss of function mutation in tumor suppressor genes, and (D) epigenetic gene modification such as virus-induced methylation of histone and DNA. Figure is modified with permission from David, RM and Doherty, 2017. ⁴ Abbreviations: LTR = long terminal repeat; mRNA = messenger ribonucleic acid; Pro = promoter; UTR = untranslated region.

To mitigate and minimize the risk of integration, three types of vector modifications may be utilized; (1) deletion of U3 from 3’ LTR (long term repeat) prevents LTR mediated gene activation (Figure 3(A)); (2) using an enhancer-blocking insulator which prevents enhancer mediated transcriptional activation of transcriptionally silent genes (Figure 3(B)), and (3) use of barrier insulator to block silencing heterochromatin from inactivating genes (Figure 3(C)) ⁴ The standard panel of genotoxicity assessment as described in the ICH S2 (R1) guideline document are not relevant for detection of vector mediated genotoxicity because these techniques are focused on DNA damage or mutation and involve short exposure and expression period. Vector mediated mutagenesis may take weeks, months, or even years. Assays to detect vector mediated genotoxicity should be able to quantitate insertional mutagenesis reproducibly, measure gain/loss of gene function, and locate the insertion within chromosomes. Unfortunately, regulatory agencies have not developed guidance specific to assessing risk of vector-mediated genotoxicity and do not ask for a specific assay to be used. Depending on the viral vector, there are various assays that may be used for assessment of genotoxicity. For gene editing therapies, these assays include but not limited to genome wide analysis of off-target modifications such as insertion/deletion quantification and/or oligoduplex integrations, IS analysis, molecular translocation assays to assess translocation events between on-on and on-off target sites, karyotyping, fibroblast soft agar transformation assay or growth in low attachment (GILA) assay, and IL-2 cell proliferation assay.

Figure 3.

Mechanisms of minimizing risk of integration. Box/arrows in purple represent vector and blue represent host genome. (A) Deletion of U3 from 3′ LTR, (B) enhancer-blocking, and (C) barrier insulator. Figure is modified by T. Parman with permission from David, RM and Doherty, AT. ⁴ Abbreviations: LTR = long terminal repeat.

Overview of Genotoxicity, Integration, and Tumorigenicity of Recombinant Adeno-Associated Viral Vectors (rAAVs)

Adeno-associated viruses (AAVs) were discovered in 1965 and thought to be non-pathogenic.^8-10 They were found to replicate only in the presence of adenoviruses. The tropism of the naturally occurring AAVs depends on the AAV strain. To date, there are a total of thirteen identified AAVs, AAV1–AAV13. More than 80% of the world’s population is AAV positive and AAV genomes have been found to be abundant in human tumors and normal tissues but lack clonality. ¹¹ In the early 1990s, wild type AAVs were shown to have a high propensity to integrate in AAVS1 site within chromosome 19.^12,13 More recently, rAAVs have been used for gene delivery with nine products approved. The most recent FDA-approval AAV delivered gene therapy is Kebilidi (first for direct administration to brain) which was approved in 2024 and Glybera, which has been removed from the market due to cost of the therapy. Integration of rAAVs is rare mainly because they lack the Rep region of the viral vector that is required for replication. ¹⁴ However, in humans, when integration occurs, it is observed in chromosome 17 and 19 near hotspots such as CpG islands, or ribosomal DNA repeats. ¹⁴

When integration occurs in mice, rAAVs tend to insert in or near gene regulatory elements, hotspots, or oncogenes. ¹⁴ In a series of studies conducted with mouse models of MPSVII (Mucopolysaccharidosis Type VII), Ornithine Transcarbamylase (OTC), Sandhoff disease, or methylmalonic acidemia, treatment of neonates from these disease models led to a dose-dependent occurrence of hepatocellular carcinoma (HCC) within 13–24 months post-treatment due to their sensitivity to rAAV integration and tumorigenicity.^15,16 These studies also demonstrated that in tumor-prone mouse species, specific component of vector design such as use of chicken β-acting or thyroxine-binding globulin enhancer/promoter, treatment in the neonatal period, and vector DNA-related contaminants may increase vector integration and infer greater risk of tumorigenesis in this species. The rAAVs integration in these mice was mainly in the Mir341 within the Rian Locus.^15,16 Given that humans do not carry Rian locus, the human relevance of rAAV induced HCC caused by integration within Rian locus in mice is unclear.

Interestingly, occurrence of HCC has not been reported in large animals particularly Hemophilia B and Hemophilia A dogs that received a one-time GT treatment delivered by rAAVs (rAAV2, 8, or 9) through 8–10 years of observation, or in nonhuman primates receiving gene therapies delivered using rAAV1, 5, or 8 through 1–5 years of observations.^16,17 In Hemophilia dogs that received a one-time AAV-delivered gene therapy, over 1,700 integration sites were noted in the liver from six dogs and expanded cell clones were observed in five dogs with integrations near genes involved in cell growth; however, none developed liver nodules or transformation. ¹⁸

Currently there are numerous rAAV delivered gene therapies that are in various phases of clinical trials. Table 1 provides selected examples of these investigational therapies with majority being in Phase I/II. ¹⁹ The most prominent AAVs used for gene delivery are AAV2, followed by AAV8, AAV9, and AAV1. ¹⁹ To-date, there has been no measurable genotoxicity among clinical trials or the marketed drugs that could be associated with rAAV GT.

Table 1.

Selected Examples of Ongoing Clinical Trials Employing rAAV Vectors ^a .

Disease	AAV	Sponsor	Phase	Clinical Trial ID
Hemophilia A	AAV2/6	Pfizer	III	NCT04370054
Hemophilia B	AAV5	CSL Behring	I/II	NCT02396342
AADC Deficiency	AAV2	Krystof Bankiewicz, UCSF	I	NCT02852213
SMA	AAV9	AveXis	III	NCT03461289
LCA	AAV5	MeiraGTx	I/II	NCT02781480
Fabry	AAV2/6	Sangamo Therapeutics	I/II	NCT04046224
DMD	AAV9	Pfizer	I	NCT03362502
Pompe	AAV8	Actus Therapeutics	I/II	NCT03533673
HIV Infection	AAV1	International AIDS Vaccine Initiative	I	NCT01937455
Alzheimer’s	AAV2	University of San Diego, Case Western University	I	NCT05040217
Canavan	AAV9	Aspa Therapeutics	I/II	NCT04998396
Parkinson’s	AAV2	National Institute of Neurological Disorder and Stroke	I	NCT01621581

Abbreviations: AADC = aromatic L-amino acid decarboxylase deficiency; AIDS = acquired immunodeficiency syndrome; AAV = adeno-associated virus; DMD = Duchenne muscular dystrophy; HIV = human immunodeficiency virus; LCA = Leber’s congenital amaurosis; SMA = spinal muscular dystrophy.

^aInformation taken from clinicaltrials.gov.

To be clear, HCC has been observed in humans after AAV gene therapy, particularly hepatic directed therapies. The occurrence of HCC in these patients has been mainly attributed to three factors ²⁰ (1) prior history of viral hepatitis, nonalcoholic fatty liver disease (NAFLD), hepatitis B (HBV), Hepatitis C (HCV), and human immunodeficiency virus (HIV); (2) prior existence of advanced fibrosis and early cirrhosis; and (3) integration of wild -type AAV2 in oncogenes. It should be noted that it is unclear and remains to be determined whether rAAV gene therapy vectors that lack rep can integrate into oncogenes in human hepatocytes. Based on these findings, specific locations for integration of AAV into the human genome likely vary from the locations within the mouse genome. In general, AAV-associated increased risk of HCC in human is hotly disputed, with a number of rebuttals arguing that the presence of this ubiquitous virus is simply association and not causation.^21–23 More recently and after completion of this workshop, a new investigation into causes of HCC in one patient that received Etranacogene dezaparvovec was conducted. ²⁴ This gene therapy was used in phase 3 HOPE-B trial (NCT03569891) and is comprised of a liver-directed rAAV5 vector containing a codon-optimized Padua-variant human factor IX transgene and a liver-selective promoter approved in USA for treatment of hemophilia B. Molecular and vector integration analysis of the liver of this one patient established no relationship to rAAV administration. ²⁴

Overview of Genotoxicity, Integration, and Tumorigenicity of Lentiviral Vectors (LVs)

LVs are recombinant form of retroviral vectors (RVs). For over 40 years, RVs have been widely used in GT for more monogenic disorders, cancer, and infectious diseases.^25,26 They are able to provide stable (long-term) and efficient expression of the delivered transgene.^25,26 Unfortunately, first-generation RVs for GT (such as certain first-generation gamma-RVs) caused adverse events linked to insertional mutagenesis. From 2003 to 2011, treatment of RV delivered gene therapies were shown to cause cell proliferation, chromosomal aberration, and carcinogenic events including myelodysplasia and leukemia, due to insertional events in various genes and modification of their expression levels.^25,26 In 2007, common insertion sites (CIS) for RVs were determined to be primarily clustered in specific regions, mainly near MDS-EVI1, PRDM16, LMO2, and CCNE2 loci; however, not all insertion sites were considered to be associated with increased risk of tumorigenicity. It was also discovered that CIS are mainly present near transcriptionally active regions. ²⁵

Subsequently, different RVs with enhanced safety features, such as the use of modified LVs, were developed to overcome the safety concerns with the early RVs. Unlike other RVs that only transduce dividing cells, LVs can transduce dividing and nondividing cells while still providing a stable (long-term) and efficient expression of the transgene.

Currently, LVs are mainly used in ex vivo modification of cells to generate chimeric antigen receptor (CAR) T-cells for cell therapy. LVs are derived from human immunodeficiency virus (HIV) but are replication incompetent and have evolved over the years to 1^st, 2^nd, and 3^rd generation with the latter generation having less risk associated with genotoxicity. The 1^st generation of LVs was found to have high risk of integration due to gag, pol, and regulatory accessory protein vif, vpr, vpu, and nef under the control of a ubiquitous cytomegalovirus (CMV) promoter in the packaging construct.^2,27,28 The 2^nd and 3^rd generation LVs have lower risk of integration because the regulatory elements are no longer in the packaging construct. Instead, the vif, vpr, vpu and nef are removed from the 2^nd and 3^rd generations. For the 3^rd generation LVs also known as self-inactivating LVs (SIN LVs), a separate regulatory construct is created, and the transfer construct has the ability to self-inactivate. In fact, SIN LVs lack the enhance-promoter element in the LTR region responsible for higher risk of genotoxicity. None-the-less some level of risk associated with LVs remains due to their unintended generation of replication-competent provirus, and non-random genome integration.^2,27,28

The mechanism of integration and genotoxicity of LVs includes integration across transcribed genes and predominantly in the introns. ²⁹ SIN LVs can lead to genotoxicity through (1) upregulation of the expression of genes flanking the IS particularly when they carry strong enhancer-promoter sequences in internal positions; and (2) induction of aberrant splicing and/or premature termination of endogenous transcripts of the target gene which can cause loss-of-function, or gain-of-function mutations. ²⁹

To assess the risk of LV integration and insertional mutagenesis, the Cdkn2a^−/− mouse model lacking tumor suppressor proteins p16^Ink4a, and p19^ARF was generated. The neonates of this mouse model are highly tumor sensitive, allowing maximal tumorigenicity assessment. By injecting various SIN LVs to newborn Cdkn2a^−/− mice, ²⁹ investigators have demonstrated the SIN LVs with LTR configuration and transgene expression cassette in an internal position have better safety profiles than vectors with active LTRs with respect to risk of insertional mutagenesis. It was demonstrated that the mechanism of insertional mutagenesis depends on four main factors, as summarized in Figure 4, and can upregulate oncogenes or downregulate tumor suppressor genes. These mechanisms are ranked based on their relative risk of occurrence as follows: (1) presence of strong promoters primarily cause oncogene activation via promoter insertion (122 days post dose); (2) vectors with strong enhancer/promoters cause enhancer mediated activation of oncogene (183 days post dose); (3) SIN vectors with moderate or weak enhancer/promoter cause both enhancer-mediated oncogene activation and tumor suppressor gene inactivation (207 days post dose); and (4) presence of insulator cassette to diminish the enhancer-mediate activation of oncogenes can also lead to dominant inactivation of tumor suppressor gene (204 days post dose). One other mechanism that was proposed during the studies conducted by Cesana et al ²⁹ was the presence of splice doner sites downstream of the promoter sequences in the construct and the potential for occurrence of aberrant splicing.

Figure 4.

Potential mechanisms of genotoxicity associated with SIN LVs. LV.SF.LTR and SIN.LV.SF represent strong promoters primarily causing oncogene activation via promoter insertion (122 days post dose), SIN.LV.SF.GFP.PRE and SIN.LV.SF.PRE represent vectors with strong enhancer/promoters which cause enhancer mediated activation of oncogene (183 days post dose), SIN.LV.PGK.GFP.PRE vector with moderate or weak enhancer/promoter cause both enhancer-mediated oncogene activation and tumor suppressor gene inactivation (207 days post dose), and INS.SIN.LV.SF.GFP.PRE represent vector with insulator cassette to diminish the enhancer-mediate activation of oncogenes can also lead to dominant inactivation of tumor suppressor gene (204 days post dose). Oval shaped boxes show examples of oncogenes that may be up- (↑) or down- (↓) regulated. The oncogenes in bold/or larger fonts represent most frequently affected oncogenes. Figure modified by T.Parman with permission from Cesana, D.; Ranzani, M.; Bartholomae, C., et al. (2014) Molecular Therapy 22(4): 774–785. ²⁹ Abbreviations: GFP = Green Fluorescent protein, INS = Insulator, LTR = Long Terminal Repeats, LV = Lentiviral Vector, PGK = Phosphoglycerate Kinase Promoter, PRE = Posttranscriptional Regulatory Element from the woodchuck hepatitis virus, SIN = Self-Inactivating, SF = spleen focus–forming virus (SF) enhancer/promoter.

To overcome the potential risk of genotoxicity of new generation of LVs, non-integrating LVs (NILVs) are currently under development.^30,31 NILVs are intended to be used for vaccinations, cancer therapy, site directed gene insertions, gene disruption strategies, and cell programming. They are generated through mutations in the integrase region of pol either through Class I or Class II mutations. Class I mutations prevent integrase activity (mutations within integrated catalytic domain), genomic DNA binding and vector DNA binding (mutation within C-terminus), and integrase multimerization of LVs (mutations within N-terminus) as shown in Figure 5.

Figure 5.

The diagram shows the HIV integrase gene with its three domains and identifies the class I mutations in various regions of the integrase gene that will render it defective thereby lowering the risk of genotoxicity by NILVs. The green circles represent a nucleotide number in gene sequence. Figure created by T. Parman using information from Banasik and McCray Jr. 2010. ³¹

Class II mutations impair integration as well as many other viral life cycles. Class I mutations are more suitable for generations of NILVs. Although NILVs are being designed to be non-integrating, at least through integrase mechanism, they have been shown to have integrase independent integration called Illegitimate integrations, which usually occur at the sites of chromosomal breakage mediated by nonhomologous end joining (NHEJ). Potential solutions are being developed that involve efforts such as inhibiting cellular factors in the double-strand break repair pathway, or limiting the linear form of the vector DNA which integrates much more efficiently than supercoiled DNA. In NILVs, transgene expression is reduced and transient in dividing cells. A potential solution for this has been described through the use of stronger promoter/enhancer elements, but this may lead to insertional mutagenesis from NILVs. Other solutions include removing or reducing inhibitors to episomal transgene expression, or inhabitation of cellular restriction factors. ³¹

Table 2 presents selected examples of ongoing clinical trials employing LVs. Additional clinical trials using LVs can be found at clinicaltrials.gov.

Table 2.

Selected Examples of Ongoing Clinical Trials Employing Lentiviral Vectors.

Condition	Intervention	Sponsor/collaborators	Trial Stage	NCT No.
ALL	CART-19 T Cells	University of Pennsylvania	Phase II	NCT02030847
ADA-SCID	Autologous EFS-ADA LV CD34+	Orchard Therapeutics, NIAID, NHGRI, NHLBI, UCLA	Phase I/II	NCT01852071
		Orchard, UCLA	Phase III	NCT04140539
		Orchard Therapeutics, CIRM, UCLA	Phase I/II	NCT02999984
Age-related macular degeneration	Endostatin and Angiostatin LV	Oxford Medical	Phase I	NCT01301443
CALD	Autologous ABCD1 LV CD34+ Cells	Bluebird Bio	Phase II/III	NCT01896102
Β-Thalassemia	Autologous βA-T87Q-globin LV CD34+ cells	Bluebird Bio	Phase III	NCT03207009
				NCT02906202
Sickle cell disease	Autologous βA-T87Q-globin LV CD34+ cells	Bluebird Bio	Phase III	NCT04293185
HIV infection	Wt gag-TCR and α/6-gag-TCR-modified T cells	University of Pennsylvania, Adaptimmune	Phase I	NCT00991224
Wiskott-Aldrich syndrome	Autologous WAS LV CD34+ cells	Genethon, Hôpital Necker-Enfants Malades	Phase I/II	NCT01347346

Abbreviations: ADA = adenosine deaminase deficiency; ALL = acute lymphoblastic leukemia; CALD = cerebral adrenoleukodystrophy; NHLBI = National Heart, Lung, and Blood Institute; NHGRI = National Human Genome Research Institute; NIAD = National Institute of Allergy and Infectious Disease; SCID = sevre combined immunodeficiency; UCLA = University of California, Los Angeles.

To date, the FDA has approved 13 lentiviral vector cell therapies, with the most recent one being TECELRA®, which is the first FDA-approved cell therapy for treatment of solid tumors. As of Nov 2024, of all FDA-approved LV modified cell therapies, only SKYSONA, used for treatment of Cerebral Adrenoleukodystrophy (CALD) has been associated with insertional mutagenesis in humans. SKYSONA (Elivaldogene autotemcel or Eli-cel) is an autologous ex-vivo modified cell therapy that is generated using lenti-D LV, which adds cDNA carrying the gene ABCD1 into the patient’s own CD34+ hematopoietic stem cells. The modified cells are then administered back to the patients for treatment. In the clinical trials of SKYSONA, 3 out of 67 patients developed hematopoietic malignancies within 5 years after the treatment, and all three were Lenti-D LV-mediated insertional oncogenesis. In a 2022 FDA advisory committee meeting with Bluebird bio and various experts in the field of cell therapy, the mechanism of SLKYSONA’s oncogenicity was thought to be related to the presence of a very strong viral promoter, MNDU3 (based on the U3 region of the Moloney Murine Leukemia Virus), in an internal position upstream of the ABCD1 cDNA driving high level of gene expression to all hematopoietic lineages (non-specific). The integrations occurred mainly near MECOM gene, a known myeloid oncogene in 53 out of 54 patients treated. Other IS included PRDM16, MPL, and MIR100HG.

CALD is a progressive, irreversible, and fatal neurodegenerative disease that primarily affects young boys. SKYSONA was demonstrated in clinical trials to be very effective, maintains event free survival for up to 7 years in approximately 87% of the treated patients, reduces CALD-related events by 72% compared to untreated patients with early active CALD, and majority of the patients-maintained baseline neurologic function and normal IQ. SKYSONA is significantly better than the existing allogeneic hematopoietic stem cell transplantation for which patients need to either wait a long time to find a matched cell doner, or accepting unmatched cell doners with an extremely high risk of rejection due to graft versus host disease. With SKYSONA, patient’s own cells are used, which eliminates both the long wait period and the potential for graft vs host disease. Although the risk of tumorigenicity exists, it was considered by FDA that the benefit of SKYSONA for this aggressive and fatal disease outweighed the risk of tumorigenicity, and was eventually approved.

Table 3 provides a selected list of recent clinical holds and reason for being placed on hold. With the exception of SKYSONA none has been related to genotoxicity or oncogenesis in humans.

Table 3.

Table of Selected Clinical Holds and Reason for Holds ^a .

Disease	Phase	Product	Vector	Sponsor	Reason on Hold	Clinical Hold Lifted	Year Lifted	Clinical Trial ID
Hemophilia A	III	PF-07055480	AAV2/6	Pfizer/Sangamo	To update study protocols for clinical management of elevated Factor VIII levels	Yes	2022	NCT04370054
Pompe	I/II	AT845	AAV8	Astellas Gene Therapies	Mild Peripheral Sensory Neuropathy	No	–	NCT04174105
SMA	III	Onasemnogene abeparvovec-xioi	AAV9	Novartis	Dorsal Root Ganglia (DRG) Toxicity	Yes	2021	NCT03505099
XSCID-1	II	MB-107	LV	Mustang Bio	CMC Issues	Yes	2021	NCT01512888
XSCID-1	II	MB-207	LV	Mustang Bio	CMC Issues	No	–	NCT01306019
SCD	I/II	HGB-206 (Lovo-cel)	LV	Bluebird Bio	Hematologic Malignancies (exonerated)	Yes	2021	NCT02140554
SCD	III	HGB-210 (Lovo-cel)	LV	Bluebird Bio	Partial: An adolescent patient with persistent, non-transfusion dependent anemia	Yes	–	NCT04293185
TDT	III	HGB-207 (Beti-cel)	LV	Bluebird Bio	Hematologic Malignancies (exonerated)	Yes	2022	NCT02906202
TDT	III	HGV-212 (Beti-cel)	LV	Bluebird Bio	Hematologic Malignancies (exonerated)	Yes	2022	NCT03207009
CALD	III	ALD-104 Lenti-D (SKYSONA, or eli-cel)	LV	Bluebird Bio	Insertional mutagenesis leading to Hematologic Malignancies	Yes	2022	NCT03852498
PKU	I/II	HMI-102	Liver-Tropic AAVHSC15	Homology Medicine	The need to modify risk mitigation measures in response to observations of elevated liver function tests	No	–	NCT03952156
PKU	I/II	BMN 307	AAV 5	BioMarin	Tumors seen in transgenic mice in an interim study	No	–	NCT04480567
Methylmalonic Acidemia	II/III	SEL-302	AAV (undisclosed)	Selecta BioSciences	CMC issues	Yes	2022	?

Abbreviations: AAV = adeno-associated virus; AML = acute myeloid leukemia; CALD = cerebral adrenoleukodystrophy; LV = lentivirus; MDS = myelodysplastic syndrome; MMA = methylmalonic acidemia; PKU = phenylketonuria; SCD = sickle cell disease; SMA = spinal muscular dystrophy; TDT = transfusion-dependent β-thalassemia; XSCID-1 = X-linked severe combined immunodeficiency type 1.

^aup to Aug 2022; From clinicaltrials.gov.

Regulatory Considerations for Insertional Mutagenesis

At the time of the symposium, there were eight FDA-approved gene therapies, seven with no reported oncogenicity and one with insertional mutagenesis; the latter was approved based on significant benefit to patients.

Oncogenesis of rAAV vectors integration has been reported in newborn mice and associated with integration near the Rian locus, which only exists in mice, not humans. Oncogenesis or tumor formation has not been reported in large animals such as hemophilia dogs up to 10 years post treatment or nonhuman primates up to 6 years post treatment.

LVs integrate specifically across actively transcribed genes, and predominantly in introns. The main factors leading to integration and high risk of genotoxicity is presence of strong promoter-enhancer elements in the vector construct.

In general, the FDA may request evaluation of genomic integrity which includes large insertion or deletion, integration of exogenous DNA, chromosomal rearrangement, and potential oncogenicity/insertional mutagenesis through assessment of clonal expansion or unregulated proliferation (only for ex-vivo modified cells). At present, there is no major guidance on specific methods to use, or assay specification requirements (e.g., cut-off points).

Due to risk of integration and oncogenesis, albeit low, FDA and other regulatory agencies do ask for long term follow up (LTFU) of patients of 5 to 15 years based on weight of evidence (WoE). ³²

Preclinical Assessment of Mutagenicity and Tumorigenicity for an Ex Vivo Lentiviral Gene Therapy Product

LV vectors can integrate genetic information into the host genome and are desired for their capability for stable introduction of genes that will be passed down to daughter cells throughout cellular development.

In contrast to early RVs, various elements and production changes are introduced into the currently used 3^rd generation LV vectors that enhance the safety of their use for GT.^2,33,34 As part of the preclinical safety assessment of LV-based GT products, in vitro tools to evaluate vector integration and cellular transformation potential are used in tandem with integration site analysis (ISA) of in vivo tissues from GT-treated animals. The results of these assessments can be used to provide a WOE assessment of mutagenicity and tumorigenicity risk of LV GT products.

For LV vector-based gene therapies, in vitro assays such as the murine cell-based in vitro immortalization (IVIM) assay and the next generation surrogate assay for genotoxicity assessment (SAGA)^35,36 were developed to explore the mutagenic potential of certain GT vectors in hematopoietic stem cells and to qualitatively predict mutagenic risk to humans when compared to positive, transforming control vectors. While the use of these in vitro tools has been accepted by the regulators as informative on assessing mutagenic risk of GT, there are inherent limitations in these assays that reduce their ability to support a quantitative risk assessment (Table 4). Notably, these analytical tools are not currently validated or executed under Good Laboratory Practice (GLP) guidelines. Discussions of standardization are ongoing within the industry. As a part of the case-by-case evaluations, essential to the evolving and diverse GT space, the assays used to evaluate integration profiles and/or safety should be clearly explained and scientifically justified in their adequacy of overall safety evaluation. Lessons learned from exploratory clinical follow-up investigations of ISA from treated patients will also likely provide valuable information as the field continues to progress.

Table 4.

Advantages and Limitations to Commonly Used Tools for Preclinical Insertion Site Analysis With LV-Mediated GTs.

Endpoint	Assay	Advantages	Limitations
Genotoxicity/Mutagenicity	In vitro immortalization (IVIM)	• In vitro system - lower cost and turnaround time	• Murine stem cells
	Surrogate assays for genotoxicity assessment (SAGA)	• Evaluate multiple vectors at once	• Not yet ‘qualified’ or ‘validated’ & not GLP
		• Specialized personnel essential for conduct & results interpretation	• Specialized personnel and computational tools essential for conduct & results interpretation
		• Lack of positive controls in nonclinical models	• Limited understanding of in vitro-in vivo concordance
Tumorigenicity	Integration site analysis (ISA)	• Use in nonclinical and clinical settings	• Not yet ‘qualified’ or ‘validated’ & not GLP
		• Tissue from animals or humans administered the actual drug product (i.e., LV-transduced HSCs)	• Specialized personnel essential for conduct & results interpretation
		• Informs on both safety and efficacy (clonal dynamics, reconstitution, durability)	• Lack of positive controls in nonclinical models

Abbreviations: GLP = good laboratory practice; HSCs = hematopoietic stem cells; LV = lentivirus.

In addition to the in vitro and ex vivo tools (i.e., ISA of tissues from GT-treated animals), choice of animal models and their potential impact on endpoints like tumorigenicity should be considered in preclinical safety assessment of cell and gene therapies (C&GTs). For LV GT, both immunocompromised, humanized mouse models (e.g., NSG mice) and disease mouse models are often used as part of the preclinical assessment package. There is some evidence that oncogenesis may be impacted by disease state ³⁷ and incorporation of safety endpoints into pharmacology/proof of concept studies is an increasingly common approach for preclinical safety assessment of these therapies in development. A better understanding of the impact of disease biology, as well as the clinical translation of relevant biological systems (e.g., dynamics of hematopoiesis between species) will also advance our understanding of safety of these therapies as the field continues to evolve.

Regulatory Perspective on Assessing Mutagenicity/Carcinogenicity Risk for Viral-Based Gene Therapy Vectors

The Role of General Pharmacology and Toxicology Assessments in Assessing the Mutagenicity and Oncogenicity Risks for Viral-Based GT Products

The risk of vector integration of viral-based GT products is one of the nonclinical safety concerns that needs to be addressed in a systematic manner. The preliminary evaluation of potential integration risk for a viral-based GT vector begins with a comprehensive characterization of the GT vector including, but not limited to, vector designs, its critical quality attributes, major release specifications, ³⁸ and ultimately the inherent ability of the viral genomic material to translocate to the nucleus and availability of viral proteins and/or interactions with cellular proteins to mediate translocation and/or integration of the viral genetic material. Viruses with DNA integration as part of their life cycle, such as lentiviruses, that have been engineered as GT vectors, have higher risks of insertional mutagenesis and oncogenicity, when compared to other viral vectors like AAV, without a DNA integration step in its life cycle. ²

Information from product characteristics will help inform the conduct and design of the preliminary pharmacology and toxicology studies. In the early stage of product development, the cumulative nonclinical data from the early and definitive studies will help guide the succeeding steps whether additional vector integration and/or tumorigenicity studies will be warranted. Specifically, careful analyses of the pharmacology, biodistribution, pharmacokinetics, and toxicology studies will be central to the nonclinical strategy and decision-making as regards to the need for further evaluation for the potential for vector integration, mutagenicity, genotoxicity, and/or oncogenicity in on-target and off-target tissues/organs. This overall WoE approach follows a stepwise and integrative decision-making process that is science-based and data-driven. Note that the specific characteristics of a particular GT product will determine the direction of the nonclinical testing strategy that allows for flexibility in approaches, as long as these are scientifically justified.^7,39

In some instances, observations from the general toxicology and other nonclinical studies might indicate that the vector could have integrated in a genomic location that could affect the cellular morphology and physiology, including changes in gene expression, cell viability, proliferation, persistence, changes in organ weight/size, and/or presence of abnormal lesions or resulting in organ toxicity(ies). ⁴⁰ For any observed tumors or growth lesions, the origin and whether the vector contributed to tumor development should be investigated. Additionally, one should also consider potential toxicities with concomitant therapies, immunomodulatory state in the study animals and/or target subject population, and the specific disease indication/specific target population, may contribute to vector-related toxicities ⁴¹ or modulate the risk for insertional mutagenesis.

Nonclinical Considerations in the Assessment of Vector Integration and Oncogenic Risks for Viral-Based GT Products

The major factors to consider when assessing vector integration and risk of insertional mutagenesis and oncogenicity include whether the GT vector is a known integrating or non-integrating vector and its inherent tissue tropism, the presence of genome editing components, ⁴² the vector backbone design, the route of administration, the dose levels to be administered, potential for long-term persistence, susceptibility of the target cells/tissues/organs to insertional mutagenesis, and risk factors in the target subject population that make them susceptible to tumor formation. ⁴³

If and when additional vector integration and oncogenicity studies are scientifically justified, the following are some of the major nonclinical testing considerations: (1) comparability of the nonclinical product (e.g., early versions or surrogate vectors) to the intended clinical product; (2) the selected animal species/disease model and how representative these are to the target indication/patient population; (3) the study conduct and design elements including adequate study duration to allow for potential tumor formation, inclusion of concurrent appropriate control groups, and adequate number of animals per group to ensure statistical significance of any biological observations such as any background incidence of tumor formation; (4) study endpoints; and (5) the nonclinical studies should mirror the clinical protocol, as feasible. ⁷

The nonclinical vector to be tested should be as similar as possible to the intended clinical vector with regard to the vector identity, viral genome titer, manufacturing process, final formulation and storage, transduction and/or editing efficiency, and should reflect the major product release criteria. ⁷

With regard to the selection of animal species or models, including cell/tissue types for the evaluation of insertional mutagenesis, the important considerations include the comparative physiology/pathophysiology to humans, permissiveness to vector transduction and/or genome editing, pharmacological response to the product, and the feasibility of clinical delivery system or procedure. ⁷

The vector and transgene biodistribution assessment is an important aspect that could help inform for potential need for integration and oncogenicity studies. When designing nonclinical studies, one should use sensitive and quantitative methods to detect product DNA sequences in relevant animal species and look at both target and off-target tissues.^38,44 It is imperative to properly archive/store tissues/organs collected for tissue assessment for possible future analysis including vector integration. In cases of persistent vector presence, drug developers should take a risk-based approach to assess the potential for integration and oncogenicity. For additional guides when designing, conducting, and interpreting your vector biodistribution studies, please refer to the “Guidance for Industry: Long-Term Follow-up After Administration of Human Gene Therapy Products (2020)” ⁴⁵ and the “International Pharmaceutical Regulators Programme (PRP) Reflection Paper: Expectations for Biodistribution (BD) Assessments for Gene Therapy Products (June 2018).” ⁴⁶

Stepwise Approach to Assessment of Risks From Insertional Mutagenesis and Oncogenicity

An important aspect of assessing the risk from vector integration is leveraging the available nonclinical and/or clinical safety data from similar vector-based GT products. ⁴⁷

With regard to the method(s) that can be used to assess the risk of vector integration, this determination should be guided by the specific biologic properties of the vector GT product including the target transduced cells, the growth kinetic potential of these cells, and the kinetics of transgene expression (particularly important if the transgene has growth factor functions). In general, a test method should be considered if it can contribute useful information to better understand the risk of insertional mutagenesis and/or oncogenic potential of a vector-based GT product.

In vitro immortalization (IVIM)^36,48 and SAGA^35,49 are two in vitro assays that have been used to assess potential for mutagenesis for integrating vectors. Results of these studies can be correlated with safety endpoints in long-term nonclinical in vivo studies (e.g., gross and microscopic evidence of tumor formation, persistence of vector and transgene in target and off-target tissues, and any delayed toxicities that may be related to genomic instability or insertional mutagenesis).

Integration site analysis is conducted to identify the frequencies and location of vector integration and could also provide information on clonal expansion, depending on the methodology used.

With increasing numbers of viral-based GT products encoding for or in combination with genome editing components, there is a potential for greater risk for mutagenicity due to vector integration ⁴² and/or DNA cleavage events in on- and off-target sites, and inefficient DNA repair activity. These events may be characterized by oncogene activation, disruption of tumor suppressor genes, disruption of protein-coding sequences/regulatory elements, translocations, and/or genomic rearrangements. The disruption of protein-coding regions may result in the production of truncated/novel peptides/proteins that may be antigenic or overproduction of transgenes or generation of fusion proteins with adverse effects.

The characterization of vector integration and/or nuclease-mediated editing events can be tested using sequencing-based methods and orthogonal approaches including in silico prediction, deep sequencing or whole genome sequencing, and biochemical and cellular approaches, ³² as applicable.

There are several challenges when addressing safety concerns for vector-based GT products, including those with genome editing components, which include issues in establishing assays and thresholds for predicting and identifying adverse genomic modifications; limitations associated with various disease models/animal species for safety evaluation and subsequent identification of potential risks; and how to account for genomic variation across the target subject population.

From a clinical standpoint, LTFU may be needed for assessment of long-term risks and delayed adverse events based on vector persistence, integration, and/or genome modification.

Challenges and Future Considerations for Assessment of Insertional Mutagenesis and Oncogenic Risk

The challenges associated with evaluating the oncogenic risk of a diverse group of AAV or LV GT products include determining the mechanism of action/process of oncogenesis and the lack of scientific consensus regarding selection of the most relevant in vitro methods and animal models to evaluate tumorigenic/carcinogenic potential or ability to predict clinical outcome. There is a continued need for a better understanding of the risk to humans, potential mechanisms and contributing factors for oncogenicity, and potential avenues to reduce risk. One suggestion is to encourage the scientific community to continue to investigate these issues further and share data in the public domain.

It is imperative that developers of GT products to work closely with drug regulatory agencies to provide feedback early and consistently during the entire product life cycle. This may entail great familiarity with the regulatory agencies’ current thinking and regulatory framework, guidance documents, and taking advantage of opportunities to interact via meetings (formal or informal) with regulators regarding specific issues and concerns that are specific for the particular GT in development. The primary goal of these interactions and meetings is to provide useful feedback from regulatory agencies that will help drug developers to achieve a successful IND submission and develop effective and safe drug products.

To avoid the potential for clinical hold in the IND submission, it is advisable for drug developers to directly address any issues or concerns that arise from nonclinical studies, solicit input from regulatory agencies prior to investing in and conducting definitive nonclinical safety/toxicology studies, and ensuring that complete study reports are included in the briefing package/IND submission.

Nonclinical Issues Encountered in Viral-Based GT IND Submissions

Common pitfalls encountered by drug developers in their IND submissions will usually involve inadequate information to assess risks to the target subject population, including insufficient product characterization, limited or absent nonclinical safety data, incomplete safety study reports, inadequate nonclinical study design and/or poorly conducted nonclinical studies (e.g., major differences in the tested product and the intended clinical product, irrelevant route of administration, dose levels, study duration, inappropriate/irrelevant test systems, and endpoints that are not clinically meaningful).

In summary, the nonclinical assessment of vector-based GT products for risks of insertional mutagenesis and oncogenicity should follow a comprehensive WoE approach and keep in mind that nonclinical studies are designed to support the administration of a specific product for a specific clinical indication.