AAV Vectors for the Nucleolus

During the replication of wild-type adeno-associated virus (AAV), viral capsids are assembled in the nucleolus,¹ and the capsid proteins interact with the nucleolar proteins nucleolin (NCL) and nucleophosmin (NPM1), possibly promoting their transport from the cytoplasm into the nucleolus.^2,3 In the case of replication-incompetent AAV vectors, the nucleolus is also important because vector virions can be found inside the nucleolus soon after entering the cell, and both NCL and NPM1 are involved in this trafficking.⁴ The nucleolar localization of vector virions is probably responsible for the unusual preference of AAV vectors for integration within the ribosomal DNA (rDNA) repeats, which accounts for 3–8% of all nonhomologous integration events,^5,6 or 10- to 30-fold more than expected based on the amount of rDNA per cell. In this issue of Molecular Therapy, two reports from the labs of Markus Grompe and Mark Kay exploit this tendency of AAV virions to localize to the nucleolus by incorporating rDNA sequences into the vector genome and transducing mouse hepatocytes with these “rDNA vectors.”^7,8 This simple trick produced dramatic effects, resulting in a 30-fold increase in chromosomal integration, about one third of which occurred within the rDNA repeats.⁸

The nucleolus is a multifunctional nuclear organelle best known for its role in the synthesis of ribosomes. It assembles around the tandem rDNA repeats present on multiple chromosomes that encode the rRNAs, is organized into morphologically distinct regions, contains its own cohort of proteins, small RNAs, and ribonucleoproteins, and plays an important role in the life cycles of many viruses,⁹ including AAV, as noted above. Figure 1 diagrams the numerous potential transduction pathways of both rDNA and non-rDNA AAV vectors and highlights possible interactions with the nucleolus. Existing data suggest that a large proportion of input genomes are initially delivered to the nucleolus regardless of their sequence, perhaps by capsid interactions with NCL and NPM1 (ref. 4). Although the exact timing of capsid disassembly and uncoating of the vector genome is not clear, some of the non-rDNA vector genomes integrate in rDNA, but most exit the nucleolus and are converted to double-stranded episomal forms through second-strand synthesis, annealing, and/or circularization.^10,11,12,13 Some of these nuclear vector genomes may integrate into host chromosomes at random locations, or into homologous sites at about a 10-fold lower frequency if designed for gene targeting.¹⁴ By contrast, most of the rDNA vector genomes remain in the nucleolus, presumably by binding the nucleolar factors that recognize rDNA. This results in a much higher frequency of both nonhomologous and homologous integration into the rDNA repeats. Although distinguishing these two types of integrants was technically difficult, the two studies estimate that homologous integration at the rDNA repeats was 5- to 50-fold lower than nonhomologous integration.^7,8

Figure 1.

Adeno-associated virus vector transduction pathways. Possible transduction mechanisms of non-rDNA (blue genomes) and rDNA (red genomes) adeno-associated virus vectors. NCL, nucleolin; NPM1, nucleophosmin.

The rDNA vectors were able to produce therapeutic levels of factor IX and fumaryl­acetoacetate hydrolase (Fah) in murine models of hemophilia and hereditary tyrosinemia, respectively. Although this had previously been achieved with conventional AAV vectors, the rDNA vectors have potential advantages. Because the integration frequencies are so much higher, one would expect stable, long-term transgene expression. This could allow a single rDNA vector treatment to last for a patient's entire lifetime, produce more consistent levels of the transgene product, and ultimately translate into lower vector doses, making gene therapy less expensive and more practical. However, Lisowski et al.⁷ noted a potential disadvantage of rDNA vectors, which is that transgene expression was lower than expected during the phase when episomal vector genomes predominated. Although the basis for this observation was not determined, it could be due to nucleolar specialization in transcription by RNA polymerase I, whereas the transgene promoters required RNA polymerase II.

Is rDNA a safe place for a vector to integrate? These papers argue that it is, because disrupting a single copy of the many rDNA repeats should not produce a change in rRNA expression, and there are examples of other transposons that integrate in rDNA without untoward effects. The rDNA repeats are also segregated away from the cellular proto-oncogenes that might be activated by vector-encoded enhancer–promoters. In addition, the process does not require a double-strand break, so potentially genotoxic, site-specific endonucleases are not delivered to the cells. Thus, rDNA vectors should be safer than non-rDNA AAV vectors, which have been shown to promote hepatocellular carcinoma after integrating into specific oncogenic loci.^15,16 However, capitalizing on the improved safety potential of rDNA vectors will first require that random integration into other loci be greatly reduced, which still occurs with their present design.

Given that these studies describe the first generation of rDNA vectors, there are many areas where we might see advances in the future. For example, can the AAV capsid be engineered to enhance nucleolar transport and uncoating? Could other tissues be targeted, especially those in which cell proliferation normally dilutes out episomal vector genomes? An obvious area for improvement is in designing expression cassettes that function well in the nucleolus. Protein-encoding genes can be expressed as RNA polymerase I transcripts if translation initiates at an internal ribosome entry site,¹⁷ and functional short hairpin RNAs can be expressed from RNA polymerase I promoters.¹⁸ Incorporating these designs into AAV vectors may allow for selective expression in the nucleolus, which would increase transgene expression and also reduce the genotoxic risks of integration in other loci.

Finally, the observation that rDNA sequences present in a vector promote nonhomologous integration into chromosomal rDNA repeats suggests that other loci might be targeted the same way. For example, adenovirus vectors that contained the β-globin locus control region were found to preferentially integrate into nonhomologous regions of the β-globin locus in an erythroleukemic cell line.¹⁹ Thus, it may be possible to target diverse chromosomal loci by incorporating homologous cis-acting sites into the vector genome that attract sequence-specific DNA-binding proteins and localize to the nuclear regions where these loci are expressed.