Contaminating manufacturing plasmids and disrupted vector genomes present in liver tissue following adeno-associated virus gene therapy

Main

AAV gene therapies show promise for treating a wide variety of serious genetic conditions, such as haemophilia^1–3, muscular dystrophies⁴ and spinal muscular atrophy⁵. As of 2024, there were seven AAV gene therapies approved by the FDA⁶, with many more in clinical trials. The most common adverse effect of intravenously-administered AAV gene therapies is hepatotoxicity, with most patients experiencing a rise in serum liver enzymes, which is routinely treated with high dose steroids. Occasionally, liver toxicity is severe and some patients have experienced fulminant liver failure^7–10. Hepatotoxicity tends to be more severe in older patients with a higher body weight, who receive higher vector doses, and who may require additional immunosuppression^11,12. Pre-existing liver disease appears to be a significant risk factor⁷.

The mechanisms underlying hepatotoxicity are complex and incompletely understood. Hepatitis has been postulated to be caused by innate, humoral and cellular immune responses to the vector capsid, genome, or transgene product^13–15, or by impurities within the vector preparation^16,17, or from a direct toxic effect^18,19. Acute sinusoidal endothelial injury resembling capillary leak syndrome has also been well documented in non-human primates using both empty capsids and therapeutic transgenes²⁰.

Onasemnogene abeparvovec (Zolgensma®, OA) is an AAV-vectored gene therapy for spinal muscular atrophy (SMA), a neurodegenerative disease caused by deleterious variants in the SMN1 gene which encodes the survival motor neuron protein (SMN)²¹. OA is manufactured using three plasmids (Figure 1): the vector plasmid (pSMN), which contains SMN and a promotor region between two AAV inverted terminal repeats (ITRs), pAAV2/9, which contains AAV2 rep and AAV9 cap genes, and pHelper, which contains the human adenovirus (HAdV) genes necessary for AAV replication^22,23. The resultant vector preparation contains therapeutic recombinant AAV (rAAV) particles which have an outer AAV9 capsid, containing a human SMN coding sequence between AAV ITRs, in a self-complementary structure. Apart from this desirable construct, there are also manufacturing process-related impurities including empty capsids, reverse packaged manufacturing plasmids, genome fragments, and recombined products^24,25. Transcriptionally-active sequences from a Rep-Cap manufacturing plasmid have been identified in mouse liver after systemic administration of a good manufacturing practice (GMP) produced and purified rAAV preparation²⁵. These manufacturing issues are complex to study and resolve, and the FDA has released guidance on reporting and validating the steps in the manufacturing process²⁶.

• Download figure
• Open in new tab

Figure 1: Schematic of plasmids used to manufacture onasemnogene abeparvovec and its mechanism of action

OA is produced by transfection of HEK293 cells with a vector plasmid (pSMN), containing SMN between AAV inverted terminal repeats (ITRs), an AAV plasmid containing AAV2 rep and AAV9 cap genes (pAAV2/9), and a helper plasmid containing HAdV genes such as E2A, E4 and VA RNA genes (pHelper)^22,23. SMN, survival motor neuron; AAV, adeno-associated virus; HAdV, human adenovirus; ITR, inverted terminal repeat; ssDNA, single-stranded DNA; dsDNA, double-stranded DNA. Produced using biorender.com.

We investigated a patient treated with OA for type 1 SMA who experienced significant symptomatic hepatitis after infusion. [Potentially identifying clinical details withheld in preprint]

Histology of the liver biopsy showed a marked periportal and lobular lymphocytic infiltrate with interface inflammation, patchy hepatocyte necrosis and ballooning degeneration, with relative sparing of bile ducts (Figure 2A, B). This histological pattern is similar to that seen in children with hepatitis associated with wild type AAV2 infection^27,28 and in “indeterminate” paediatric acute liver failure²⁹. In situ-hybridisation for SMN confirmed hepatocyte vector transduction in the patient liver (Figure 2C, D).

• Download figure
• Open in new tab

Figure 2: Liver biopsy findings

A) Liver biopsy of the patient shows marked periportal and lobular inflammation as well as interface inflammation (^*); numerous hepatocytes with ballooning degeneration are present (^**). Box in A is magnified in B. B) High magnification of ballooning hepatocytes highlighting the swollen cytoplasm. C) RNA in situ-hybridisation detecting the SMN1-gene shows a strong positive red signal in the nucleus of ballooning hepatocytes separated by areas with severe immune cell infiltration (^*). Box in C is magnified in D. D) High magnification of ballooning hepatocytes shows the positive signal in the nucleus and a mild to moderate, punctuated signal within the cytoplasm of hepatocytes and of the immune cells (^*). Inflammation in the liver is shown by immunohistochemistry detecting CD4 (E), CD8 (F) and CD20 (G). Bars, A and C, 400 micrometres, B, 60 micrometres, D, 100 micrometres, E-G, 300 micrometres.

We conducted untargeted short-read metagenomic sequencing of DNA and RNA from the residual patient liver sample. Analysis identified multiple serotypes of AAV, human mastadenovirus C (HAdV-C) and human betaherpesvirus 6 type B (HHV-6B) in the DNA-seq, while RNA-seq did not identify any pathogen transcripts, consistent with lack of viral replication (Table 1). For AAV2 and HAdV-C, the sequencing reads only aligned to the sections of the viral genomes that are part of the OA manufacturing plasmids, suggesting presence of plasmids in the liver tissue rather than wild-type virus infection (Figure 3A). In contrast, HHV-6B reads covered the breadth of the HHV-6B genome (Figure 3B), suggesting natural HHV-6B infection. Accordingly, a specific PCR for HAdV was negative (targeting a region of the genome which is not present in pHelper), and a specific PCR for HHV6 was positive (Ct value= 26.2). Aligning reads to the manufacturing plasmid sequences, we found good coverage of the vector genome, but also of the other regions of the pSMN manufacturing plasmid and of pAAV2/9, with some reads mapping to pHelper (Figure 3C).

• Download figure
• Open in new tab

Figure 3: Metagenomic sequence alignment

A) Genome coverage of wild-type (WT) AAV2 and HAdV-C from Illumina sequencing reads. Approximate locations of the genes present in the manufacturing plasmids are marked along the x-axis. AAV2 alignment uses more stringent mapping parameters to more clearly differentiate between any AAV2 and AAV9 derived sequences – see methods. B) Alignment of Illumina sequencing reads to the HHV-6B genome shows reads cover the breadth of the genome. C) Alignment of Illumina sequencing reads to approximate manufacturing plasmid sequences shows presence of plasmid sequences. NB From the negative control, 10 reads aligned to the pSMN sequence. No reads aligned to the pAAV2/9 or pHelper sequences.

To further confirm the presence of the manufacturing plasmids in the liver, we performed untargeted long read metagenomic sequencing of the liver sample (Oxford Nanopore Technologies). Again, we found reads aligning to AAVs, HHV-6B and HAdV-C (Table 1), along with the gene therapy vector sequence. Reads mapping to all three of the manufacturing plasmids were also found, primarily to pSMN (including the region of the plasmid that is not included in the therapeutic vector genome), and pAAV2/9 (Table 2). Sequence analysis of individual reads showed high levels of vector genome concatemerisation and complex genome structures with rearrangements (Figure 4A-D, Table 2). The concatemeric patterns observed show similarities to those seen in replicating AAVs using rolling hairpin and rolling circle amplification³⁰. Many of the structures observed were longer than the maximum packaging length of an AAV vector (approximately 4.7kb), suggesting that recombination may have occurred in vivo. Alternatively, these sequences may represent plasmid contaminants from manufacture. The majority of pAAV2/9 reads also contained regions of the other OA manufacturing plasmids, indicating recombination between plasmid sequences (Figure 4, Table 2). Most of the complex structures and recombination involved the region between the ITRs of pSMN, the rep/cap region of pAAV2/9, and the region of pHelper containing the HAdV-derived genes (Figure 4, Table 2).

• Download figure
• Open in new tab

Figure 4: Alignment dot plots

Alignment dot plots showing individual nanopore reads (x axis) aligning to representative sequences of the OA manufacturing plasmids (y axis). Red dots show alignment to forward strand and blue to the reverse. A) Alignment against vector region of pSMN plasmid. B) Alignment against entire pSMN plasmid. C) Alignment to pAAV2/9 plasmid. D) Alignment to regions of all three plasmids – the vector region of pSMN, AAV rep and cap within pAAV2/9 and the HAdV gene region within pHelper. Representative images selected – numbers of reads belonging to each category can be found in Table 2.

Complex AAV vector genome rearrangements with structural rearrangements and concatemers have previously been demonstrated in macaque liver after treatment with rAAVs^31,32 and in human hepatocytes in a humanised mouse model³³. Similar complex concatemeric structures have been noted in liver samples from children with hepatitis associated with wild-type AAV2 infection²⁷. To our knowledge, this is the first metagenomic analysis of liver tissue from an rAAV-treated patient to comprehensively elucidate the vector concatemer structures and complex genome rearrangements, which also included some manufacturing plasmid sequences.

Both short and long read metagenomic sequencing approaches in this study provide evidence that sequences from all three manufacturing plasmids were present in the liver of a patient with hepatitis after treatment with OA. Long read sequencing revealed large quantities of complexed plasmid DNA and many reads contained elements of multiple plasmids. It is possible that recombination events between plasmids took place during the manufacturing process, resulting in transduction of the AAV rep and cap genes and the HAdV helper genes as contaminants in the therapeutic vector preparation. Theoretically, this could allow replication competence of the vector genome in liver cells, thus explaining the replication-associated vector genome complexes we observed. The contribution of these genomic structures to the pathogenesis of rAAV hepatitis, and the putative role of contaminating plasmids and their potential immunogenicity, remain to be determined. It would be important to ascertain whether these genomic structures are also present in rAAV-treated patients without hepatitis.

Due to insufficient liver material, we were unable to assess whether the rearranged vector genomes were chromosomally integrated or episomal. Random, low frequency integration of various rAAV vectors in patient tissue is now well recognised^34–37. AAV integrants in complex concatemers containing mixtures of rearranged and truncated vector genomes have been demonstrated in liver tissue of non-human primates after intravenous administration of rAAV8 vectors³¹. The significance of HHV-6B in the liver is unclear from this single case description. HHV-6 can act as a “helper” virus in wild-type AAV2 replication, and its genome contains a homologue of the AAV rep gene³⁸. HHV-6 has also been found in liver tissue in a proportion of children with hepatitis associated with wild type AAV2 infection, although also sometimes in controls^27,28, and in children with acute liver failure of unknown cause^39,40.

Overall, our metagenomic analysis of liver tissue from a patient with hepatitis after OA treatment has revealed extensive disruption and concatemerisation of vector genomes, alongside numerous contaminating manufacturing plasmid sequences, with evidence of complex genomic structures and recombination events. We also identified HHV-6B in the liver. We postulate that presence of certain manufacturing plasmid sequences or helper viruses may allow replication of the vector genome within cells, giving rise to complex concatemeric structures. Future work is needed to determine the frequency and pathological significance of complex DNA structures in patient liver cells after rAAV gene therapy, whether they are episomal or integrated into the host genome, and how this may relate to the known hepatotoxicity of rAAV gene therapies.

Declaration of competing interests

GB is PI of clinical trials Sponsored by Roche, Novartis, Sarepta, Pfizer, NS Pharma, Reveragen, Percheron, Biomarin, Scholar Rock, and has received speaker and/or consulting fees from Sarepta, PTC Therapeutics, Entrada Therapeutics, Pfizer, Biogen, Novartis Gene Therapies, Inc. (AveXis), and Roche, and grants from Sarepta, Roche and Novartis Gene Therapies. UCL has received funding from Sarepta, Roche, Pfizer, Italfarmaco, Santhera.

FM is the PI of the Novartis sponsored trials in which OA was studied in the UK, and is also involved in clinical trials sponsored by Biogen, Roche, Sarepta Therapeutics, Genethon, PTC therapeutics and Solid Bioscience. He has received consulting fees from Pfizer, Sarepta, Roche, Biogen, Novartis, Solid, Dyne Therapeutics, Entrada, PTC and Edgewise.

MS is the sub-I of the Novartis sponsored trials in which OA was studied in the UK, and is also involved in clinical trials sponsored by Biogen, Roche, Dyne. She has received consulting fees from Roche, Biogen and Novartis.