AAV Vector Integration Sites in Mouse Hepatocellular Carcinoma

See allHide authors and affiliations

Science  27 Jul 2007:
Vol. 317, Issue 5837, pp. 477
DOI: 10.1126/science.1142658


Adeno-associated viruses (AAV) are promising gene therapy vectors that have little or no acute toxicity. We show that normal mice and mice with mucopolysaccharidosis VII (MPS VII) develop hepatocellular carcinoma (HCC) after neonatal injection of an AAV vector expressing b-glucuronidase. AAV proviruses were isolated from four tumors and were all located within a 6-kilobase region of chromosome 12. This locus encodes several imprinted transcripts, small nucleolar RNAs (snoRNAs), and microRNAs. Transcripts from adjacent genes encoding snoRNAs and microRNAs were overexpressed in tumors. Our findings implicate this locus in the development of HCC and raise concerns over the clinical use of AAV vectors.

Adeno-associated viruses (AAVs) are promising vectors for gene therapy. We previously reported a high incidence of hepatocellular carcinoma (HCC) in AAV-treated mice with the lysosomal storage disease mucopolysaccharidosis VII (MPS VII) (1). Similar malignant hepatocellular changes were also observed in other mouse models of AAV-mediated gene therapy (2, 3). In each case, the underlying molecular mechanism is unknown.

We examined HCC formation in mice after neonatal intravenous injections of 1.5 × 1011 genome-containing particles of an AAV vector expressing the human β-glucuronidase gene from a β-actin promoter and a cytomegalovirus (CMV) enhancer (AAV-GUSB) (fig. S1). Six of 18 (33%) AAV-treated MPS VII mice developed HCC, compared with 1 of 25 (4%) mice treated with bone marrow transplantation (table S1). Wild-type mice injected with the AAV-GUSB vector or a version lacking the β-actin promoter (fig. S1) also developed HCC at significantly higher rates (56% and 33%, respectively) compared with untreated normal mice (8.3%). No tumors were observed in transgenic mice overexpressing GUSB from the same expression cassette as that in AAV-GUSB.

We attempted to isolate vector-chromosome junctions from six tumors present in six different mice by inverse polymerase chain reaction (PCR). A single, unique amplification product was detected in four samples (Fig. 1A). Each junction was similar to those previously described (4, 5) and contained a portion of the 5′ vector inverted terminal repeat (ITR) with transgene transcription proceeding in a telomeric direction. We compared the copy numbers of junction-specific fragments in tumor tissue and in normal liver. In the three cases where adjacent tissue was available, the junctions were undetectable in normal-appearing liver (<1 copy per 100 genomes). Junction copy numbers ranged from 3 to 27 per 100 diploid genomes in tumor tissue (Fig. 1B). Because murine hepatocytes can be tetraploid or octaploid (6) and because tumor samples contain a variety of supporting and inflammatory cells that presumably lack vector proviruses, these values are consistent with a single AAV vector integration event resulting in clonal expansion of transformed cells. However, we cannot exclude the possibility that additional, undetected vector integrations were also present in the tumor samples.

Fig. 1.

(A) Vector-chromosome junctions are shown with colors indicating sequence origin: black, chromosome 12; blue, AAV; green, microhomologies between AAV and chromosome 12; and red, inserted nucleotides. Chromosomal and AAV ITR (flip or flop orientation) positions are indicated. (B) Quantitative PCR for junction copy numbers in tumors and in adjacent, normal tissue [primer sites shown in (A)]. Normal samples were below the limit of detection. NA indicates not available. Error bars represent 1 SD. (C) The genomic locus containing the integration sites is shown with relevant portions expanded. Imprinted transcripts expressed from maternal (pink) or paternal (blue) alleles, areas where transcription has not been confirmed (dashed lines), and microRNAs (triangles) are indicated. (D) Transcript levels in tumors divided by those from adjacent normal-appearing tissue. Fold differences for mice with junctions 1, 3, and 4 were averaged and plotted versus chromosome 12 positions. Blue dots are assayed transcripts (two overlap for Rian and Mirg). Red lines are sites of vector integrations.

All four junctions mapped to a 6-kilobase region of chromosome 12 (Fig. 1C). Two of the insertion sites were located just 12 base pairs apart within the mir-341 microRNA transcript. Microarray analysis showed that genes adjacent and telomeric to the AAV vector proviruses were dramatically overexpressed (Fig. 1D), and they were up-regulated in all three tumor samples studied, suggesting that the transcriptional changes were due to provirus insertions.

Thirty-four of the 382 known mouse microRNAs, with thousands of potential target genes, are located within this locus. The highly overexpressed Rian and Mirg genes contain multiple small nucleolar RNAs (snoRNAs) and microRNAs, respectively, that could have profound effects on host gene expression. The fact that every junction isolated was present at the same locus and resulted in similar changes in expression suggests that these events promote a critical step in tumorigenesis. A similar locus on chromosome 7 was not dysregulated [Supporting Online Material (SOM) text].

Our findings implicate insertional mutagenesis by AAV vectors in the development of hepatocellular carcinoma. Because humans have a syntenic locus on chromosome 14 that has been linked to several cancers (7, 8), these findings raise safety concerns over the clinical use of AAV vectors.

Supporting Online Material

Materials and Methods

SOM Text

Fig. S1



View Abstract

Stay Connected to Science

Navigate This Article