In these in vitro studies of retroviral DNA integration, we took advantage of a DNA target that has been used previously to reconstitute extended nucleosome arrays that can be compacted into higher-order structures comparable to chromatin in living cells (8
). Our results showed that when an extended nucleosomal array is used as an integration target by ASV and HIV-1 integrases, the distribution of joining sites is distinct from that observed in the corresponding naked DNA, consistent with earlier findings (30
). However, further compaction of the extended array by histone H1 or a high NaCl concentration did not cause major changes in the joining pattern but had dramatically opposite effects on the efficiency of the reaction with the ASV and HIV-1 integrases. We found that ASV integrase-mediated joining of the cognate viral DNA was substantially more efficient with a histone H1-compacted DNA chromatin target than with naked DNA or an extended nucleosome array. A similar increase was also observed with NaCl-compacted target chromatin, suggesting a critical role for the structure of the DNA in compacted chromatin, rather than interactions between the ASV integration machinery and histone H1. In contrast to the results obtained with ASV integrase, a decrease rather than enhancement of integration into the histone H1-compacted chromatin target was observed in reactions catalyzed by HIV-1 integrase. These observations indicate that the integration machineries of these two retroviruses may be markedly different in their site selection preferences.
Extended chromatin and compacted chromatin are characteristic of actively transcribed or silent regions of cellular chromosomes, respectively. Actively transcribed regions include a number of bound components that could potentially reduce the efficiency of retroviral DNA integration. Our data suggest that transcription factors bound to the chromatin may block ASV DNA integration in actively transcribed regions. The HNF3 transcription factor binds to three sites in the DNA associated with nucleosome N1 in the extended nucleosome array, and such binding results in DNA protection and exposure of hypersensitive sites, as detected by DNase I footprinting (8
). In our ASV integrase-mediated joining reaction, accessibility was limited specifically at HNF3 binding sites in this target DNA but, interestingly, we did not observe any additional hot spots for viral donor DNA joining in the presence of this transcription factor. We also asked if HNF3 transcription factor binding to the histone H1-compacted target would affect ASV integrase-mediated joining in our system. Surprisingly, we observed no changes in the joining pattern when the compacted target was opened by HNF3 binding, even though our control experiment showed that sequences in the N1 nucleosome region became more accessible to DNase I digestion. These results underscore the preference of ASV for compacted chromatin, even when an open region is available on the same template.
The striking differences we observed between the target preferences of the ASV and HIV-1 integrases suggest that these proteins may interact differentially with extended and compacted chromatin in vivo. Although many factors may contribute to target site selection (11
), it is noteworthy that the joining preferences of the integrase proteins that we have uncovered in vitro are consistent with results reported for integration of ASV and HIV-1 DNAs in infected cells (34
). This observation underscores the importance of using more developed models of chromatin structure in vitro to replicate integration activities seen in vivo.
Retroviral integration site selection has become an important topic of study owing to the use of retroviral vectors in gene therapy for human disease. Several cases of insertional mutagenesis, believed to have been triggered by integration of the MLV vector DNA near the growth-promoting gene LMO2
were recently observed. In these clinical studies, 2 of 11 patients treated for severe combined immunodeficiency disease developed leukemias, with the malignant cells showing integrations at this locus (4
). A recent study with mice has also shown that integration of replication-deficient MLV vector DNA is associated with leukemia development (24
). Our results suggest that an ASV vector might have some advantage over MLV or HIV-1 vectors if integration into actively transcribed regions is to be avoided. On the other hand, integration into regions of compacted chromatin might cause silencing of an integrated ASV vector. Additional engineering, for example, introduction of insulator sequences, might help to alleviate this problem (10
). As ASV has in common with HIV-1 the ability to infect nondividing cells (17
), an ASV vector may be especially useful for particular therapeutic applications.
Integrases from different retroviruses produce unique and reproducible patterns of integration into naked DNA in vitro. The determinants for these patterns appear to be located within the core domain of the protein (19
). Here we show that the HIV-1 and ASV integrases also have different integration preferences with respect to compacted chromatin targets. The molecular determinants for these properties of integrases remain to be determined. Detailed knowledge about the mechanism of integration site selection in chromatin may contribute to the development of safer gene therapeutic tools.