Search

The recent recognition that recombinant retrovirus vectors can induce oncogenic transformation has stimulated much interest in the pattern of vector integration sites. We report here on the integration pattern of a gammaretrovirus reporter vector following transduction and ex vivo culture of primary mouse bone marrow progenitor cells in the absence and presence of drug selection. Using a novel method of cloning junction fragments, we observed no bias for integrations within genes, but did observe a bias for integrations within gene-dense regions and especially near transcriptional start sites of highly active genes, similar to previous reports in other cell types. We also document a novel bias for integrations within or near a class of genes that encode nuclear-localized proteins. We found that drug selection resulted in an increase in the frequency of recovered integration events that were located within the beginning of genes, integration events that were located in less gene-dense regions, and integration events that were oriented in an antisense direction relative to flanking gene transcription. Taken together, these studies provide new insights into the nature of retrovirus vector integration patterns in primary cells and demonstrate that selection based on vector expression can bias the integration site repertoire.

The expression of integrated oncoretrovirus vectors is subject to the inhibitory effects of surrounding chromatin. A previous report from our laboratory indicated that such position effects can be overcome by flanking a reporter vector with the cHS4 chromatin insulator. To characterize this activity more thoroughly, we switched the promoter-gene combinations in the reporter vector and analyzed expression of these vectors flanked with the cHS4 fragment in both orientations following bone marrow transduction and transplantation in mice. The results indicate that the cHS4 fragment can function in both orientations and can insulate both the virus long-terminal-repeat (LTR) promoter and an internal phosphoglycerate kinase (Pgk) promoter. However, insulation of the LTR promoter diminished when the orientation of the cHS4 fragment placed the CTCF-binding core element immediately proximal to the U3 region, suggesting a minimal distance requirement. Moreover, placement of the cHS4 fragment in the U3 region of the 3' LTR dramatically decreased the level of expression from an internal Pgk promoter, presumably by blocking interaction with the 3' LTR enhancer. Finally, sorting studies suggest that the severity of position effects or autonomous promoter silencing increases as transduced progenitors differentiate into mature progeny. These findings have direct implications for the use of chromatin insulators such as cHS4 in oncoretrovirus vectors.

Although gene transfer vectors based on recombinant retroviruses have been known to have the potential to lead to insertional mutagenesis, that integrations can lead to oncogenic transformation of transduced target cells in a gene therapy setting has led to intensive research on integration site choice. Several groups have reported that oncoretroviral vectors have a slight preference for landing within genes, a strong preference for integration within the region immediately around transcriptional start sites, and a preference for actively transcribed genes. Using primary mouse bone marrow progenitors as target cells for an oncoretroviral vector, we addressed three questions: 1) Are genes preferentially the targets; 2) Are regions surrounding transcriptional starts favored; and 3) Does selection for the vector bias the repertoire of integration sites? We obtained bone marrow from 5-FU treated mice, prestimulated the cells for two days, transduced by co-cultivation on irradiated ecotropic producer lines at a low multiplicity of infection. The retrovirus has a bacterial drug resistance cassette for Zeocin in the long terminal repeat, an MSCV LTR transcribing a GFP reporter, and a separate internal Pgk-Neo cassette (used to select for expression). Progenitor cell colonies were recovered in methylcellulose +/|[minus]| G418 seven days after transduction. Integration junctions were cloned directly from genomic DNA with the Zeocin cassette present in the LTRs, rather than by a PCR based strategy, enabling us to obtain larger genomic segments. Junction sequences were mapped by BLAST to build 33 of the mouse genome in the ENSEMBL database. We identified 130 integration sites from unselected pools and 130 integration sites from selected pools of progenitor cells. We also mapped an equivalent number of simulated integrations; this dataset was treated as the experimental dataset was, with unmappable sites (such as those that landed entirely in gaps or repetitive elements) removed. We have found that MSCV integrations do not occur more frequently into genes (defined as within known transcripts introns and exons, 42%) than random (41%), differing from what was previously reported for MLV integration into HeLa cells. Comparing the integrations that are within 5 kb upstream of a transcription start site, there is a highly significant difference between our experimental dataset (18%) and the simulated dataset (4%), P = < .0001. With the progenitor colonies having undergone selection, the frequency of landing within introns has increased from 35% to 43%. When the overall distribution of integrations is compared to gene density mapped onto the mouse chromosomes, it appears that oncoretroviruses favor gene-rich regions of the mouse genome. Though MSCV retroviruses do not have a preference for integrating within genes, they favor the 5 kb region upstream of transcriptional starts, 18% in this setting, versus 8.5% observed previously in HeLa cells. These results suggest that there may be less of a bias for oncoretrovrus integration within genes in mouse bone marrow progenitor cells compared to other cells types tested to date, and that the use of selection can bias the repertoire of integration sites that are observed following oncoretrovirus transduction.

Recombinant retroviruses are being developed as gene transfer vectors for a wide range of research and therapeutic applications. However, vector expression is subject to the inhibitory effects of chromatin surrounding the sites of integration, and vector sequences can activate oncogenes surrounding the site of integration. In order to address these limitations, we have been investigating a class of DNA elements called chromatin insulators. We have previously demonstrated that a prototypic insulator cHS4 can increase the frequency of expression for reporter genes and human γ-globin in mice transplanted with transduced marrow, and that this activity may be influenced by distance. In order to test the role of distance directly, the frequency of GFP expression from a vector flanked with various forms of the cHS4 element was assessed in myeloid progenitor colonies. Linking the cHS4 core to a 950bp native sequence or 720bp spacer increased the frequency of expression 2.1±0.3 fold (p 590bp from the target promoter in order to provide insulating activity. In order to further test the efficacy of the cHS4 element, we analyzed an insulated vector for human γ-globin in mouse models of β-thalassemia. In the heterozygous Hbbth-3/+ model there was a modest improvement in RBC counts (3.8±0.1×106/ul vs 2.6±1.1 for mock) and total hemoglobin (6.7±0.2 g/dl vs. 4.7±1.9 for mock) for mice with ≥25% g(+) RBC, although this effect was only transient. In the homozygous Hbbth-3/Hbbth-3 model there was increased survival to 34±20 days vs 16±7 days for mock (p