Adeno-Associated Virus Gene Therapy: Plasmid Contamination & Vector Genome Disruption in Liver
Adeno-associated virus (AAV) gene therapies show promise for treating a variety of serious genetic conditions, including hemophilia1,2,3, muscular dystrophies4 adn spinal muscular atrophy (SMA)5. As of 2025, there were seven AAV gene therapies approved by the US Food and Drug governance6, with many more in clinical trials. The most common adverse effect of intravenously administered AAV gene therapies is hepatotoxicity, routinely treated with high dose steroids. Occasionally, liver toxicity is severe, and some patients have experienced fulminant liver failure.
The mechanisms underlying hepatotoxicity are incompletely understood, and it has been postulated to be caused by innate, humoral and cellular immune responses to the vector capsid, genome or transgene product14,15,16, by impurities within the vector preparation17,18 or from a direct toxic effect1). Liver injury was managed using steroids and tacrolimus. Tacrolimus was successfully withdrawn 7 months after infusion, and steroid treatment continued for 19 months.
A needle core liver biopsy,taken 7 weeks after infusion,showed mild perivenular and portal fibrosis and a single focus of porto-central necrosis. There was mild portal tract expansion, including a portal ductular reaction and periductal and intraepithelial neutrophils.There was a moderate portal inflammatory infiltrate composed predominantly of CD4- and CD8-positive T lymphocytes and occasional plasma cells,with mild interface inflammation and moderate lobular inflammation with foci of hepatocellular cholestasis (Fig. 2). Few CD20-positive B lymphocytes were detected. These histological features are consistent with those previously reported in children with hepatitis associated with wild-type AAV2 infection29,30 and in ’indeterminate’ pediatric acute liver failure31. Adenovirus immunostaining was negative (Fig. 2h).
The incomplete genome coverage of AAV2 and HAdV suggested that the results did not derive from a wild-type infection (Fig. 3a). To investigate this further,we aligned the reads to the manufacturing plasmid sequences used in OA production. We found good coverage of the OA vector genome as expected,but also of the pSMN plasmid backbone and of pAAV2/9,and some reads mapping to pHelper (Fig. 3c). the reads originally classified as AAV2 or HAdV-C aligned onyl to sections of the viral genomes that are part of the OA manufacturing plasmids (AAV2 rep, HAdV E4, E2A, L4 and VA regions), suggesting the presence of plasmid sequences in the liver tissue rather than wild-type virus infection (Fig.3a). A specific PCR for HAdV, targeting a region of the genome that is not present in the pHelper plasmid, was negative.
The presence of the pAAV2/9 plasmid sequences also potentially explains why multiple AAV serotypes, other than AAV2, including AAV4 and A
Tissue showing a positive signal (control patients showed 0.4-1.5% positive cells; Fig. 4 and Extended Data Fig. 2).We observed both nuclear and cytoplasmic positive signals.To detect plasmid sequences, we designed probes complementary to regions of the manufacturing plasmids that are absent from both the therapeutic OA vector genome and the human genome. Analysis confirmed the presence of the bacterial origin of replication in pSMN, pHelper and pAAV2/9 plasmids in 5.1% of cells (probe vector-pHelper-C1, 0.2-1.1% positive in controls), as well as a sequence from the AAV9 cap gene present in pAAV2/9 in 5.8% of cells (probe AAV-HeB-T1-VP1-O1-C1, 0.2-1.1% positive in controls) (Fig. 4). The contaminant plasmid-specific sequences were found at lower levels than SMN, in agreement with the metagenomic sequencing.
Fig. 4: in situ hybridization.
a-c, In situ hybridization (ISH) for the detection of SMN1 in formalin-fixed paraffin-embedded (FFPE) liver tissue. a, A strong positive red signal was detected in the nucleus of ballooning hepatocytes separated by areas with severe immune cell infiltration (*). The box in a is magnified in b. b,Higher magnification of SMN1-positive hepatocytes next to the immune cell infiltrate (*). c, Dense nuclear signal for SMN1 and a mild-to-moderate, punctuated signal within the cytoplasm of hepatocytes and immune cells (*). d,e, ISH for the detection of ma
Alignment dot plots showing individual nanopore reads (x axis) aligning to representative sequences of the OA manufacturing plasmids (y axis).Red dots indicate alignment to the forward strand,and blue dots indicate alignment to the reverse strand. a,Explanation of dot plot format. b, Alignment against the vector region of the pSMN plasmid. c, Alignment against the entire pSMN plasmid.d, Alignment to the pAAV2/9 plasmid. e, 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 were selected; the number of reads in each category can be found in Extended Data Table 3, and diagrams for all reads are provided in the Supplementary Information. See the Supplementary Information and Methods for description of similar dot plots generated for human reads.
Rearranged sequences may derive from recombined plasmid contaminants outside vector particles, mispackaged recombined DNA from manufacture and/or recombination events after infusion. Many of the structures we observed were longer than the maximum packaging length of an AAV vector (up to 15 kb, while the packaging limit is approximately 5 kb (refs. 2930
The findings, published in Frontiers in Pediatrics in 2023, suggest a potential role for HHV-6 in pediatric liver conditions, though its exact contribution is still being investigated.43