In this study, we investigated integration site selection after infection with an ASV-based retroviral vector. We used a vector system that was engineered for infection of human cells (1
), which allowed identification of target sites in the human genome and determination of the transcriptional states of the target genes prior to integration. The general experimental design and the mapping methods were as described previously and shown to be free from significant biases (34
). For ASV, we found that, in general, integration sites were not restricted to any particular regions or known chromosomal features, as measured by several parameters. One limitation of our study is the use of heterologous human host cells, which was dictated by the availability of human genome sequence and the ability to perform transcriptional profiling. If site selection is influenced by specific interactions between the avian retroviral preintegration complex and chicken-specific host factors, such targeting would be missed in our system. However, the fact that authentic and apparently efficient ASV integration occurs in mammalian cells argues against the existence of critical species-specific interactions. We note that our conclusions regarding global accessibility of integration sites are generally consistent with an earlier study that used a primer extension method to survey ASV integration sites in turkey cells (42
). This study detected local integration hotspots for ASV, likely due to the ability to measure a larger number of events. With the noted caveats, our study provides new information regarding how broad chromosomal features and transcriptional activity affect ASV integration site selection and provides a useful comparison with MLV and HIV-1.
Our results suggest that ASV integration favors genes (e.g., 39.8 versus 22.4% expected for random integration) (Table ). Schröder et al. (34
) and Wu et al. (43
) reported that genes are favored targets for both HIV-1 and MLV integration, with HIV-1 showing a stronger preference than both ASV and MLV (Table ). The bias for HIV-1 integration into genes was especially pronounced in genes activated in response to HIV-1 infection (34
). The preference for ASV (and MLV) integration into genes is modest and is dependent on the measurement of random integration, as discussed above. Another method for determining the random integration frequency is to perform in vitro integration reactions on naked cellular target DNA (34
); however, this method could produce other biases. The measured preferences for integration into genes by ASV and MLV might decrease or increase with future analyses, based on more accurate values for random integration. However, in a comparative sense, we can more definitively conclude that ASV and MLV show similar frequencies of integration into genes, as the same cell type, gene definition (RefSeq), and value for random integration were used in the two studies.
The previous MLV studies revealed a unique bias compared with HIV-1; gene promoter regions were five times more likely than random to be targets for MLV integration, and it is possible that this tendency may be relevant to activation of the LMO2 gene in human patients during retroviral gene therapy (43
). With respect to targeting over the length of genes, we found that, in contrast to MLV, ASV does not show a bias for transcriptional start sites (Fig. ).
Transcriptional analysis of ASV target genes revealed that their expression values fell very close to the median for all analyzed genes. Further analyses showed that integration site selection is essentially random with respect to the intensity of transcriptional activity of target genes (Fig. and ). It is possible that the preference for ASV integration into genes may reflect increased accessibility associated with some baseline transcriptional activity. Our results show that, although there is a bias for ASV integration into genes, these genes are expressed at average levels; highly active genes (greater than twofold above the median expression level) are not favored (Fig. ). Recently, it was reported that highly active transcription may inhibit ASV integration (41
), and our results are not inconsistent with this proposal. In contrast to our findings with ASV, a positive correlation between high transcriptional activity and integration site selection was seen with HIV-1 (34
Results from our study together with previous reports (34
) suggest that the measurable differences in integration preferences of HIV, MLV, and ASV vectors may result from differences in the integration site selection mechanisms of these retroviruses. Determinants of selection might include virus-specific properties of IN proteins (18
), preintegration complexes, or virus-specific interactions with cellular cofactors (4
). The yeast retrotransposons Ty3 and Ty5 have evolved mechanisms of site selection that depend on physical interaction between components of the integration complex and different cellular factors bound to DNA, resulting in different targeting specificities (reviewed in reference 5
). Although retroviral integration appears to lack specificity with respect to target DNA sequence, similar host factor-based targeting mechanisms may play a role in site selection (5
). If such targeting mechanisms are indeed operative, they may be tissue or cell type specific. For example, putative chromatin-bound targeting proteins might be expected to be expressed differentially in various cell types. Furthermore, it is possible that the lack of targeting to transcription start sites that we observed could be due to interspecies infection.
Elaboration of the determinants of site selection will be necessary to completely describe the early events in retroviral replication; with such knowledge, virus-specific differences may be incorporated into retroviral vector design. The predominant retroviral vectors currently in use are based on MLV and HIV-1. It is possible that virus-specific features of integration site selection (such as an apparent lack of preference for transcriptional start sites, highlighted here for ASV) may provide practical advantages (43
). We (17
) and others (12
) recently demonstrated that transduction by ASV vectors is not limited to dividing cells, further highlighting the prospect of exploiting a broader array of retroviral vectors for gene therapy in differentiated cells. Thus, the choice of retroviral vectors might be tailored for specific needs. It has also been suggested that preferences for gene targeting might be exploited for insertional mutagenesis (34
Beyond providing a more detailed description of the early events of replication, and the possible implications for vector design, our results contribute to an understanding of how genomes are shaped in evolution, as a large percentage of the human genome comprises integrated retroviral sequences (19
). Technical refinements, which should eventually allow higher-throughput analysis of integration sites, may allow use of retroviral integration as an in vivo probe of chromatin structure and genome organization, providing an experimental model for genome shaping.