Enhancers are major targets for murine leukemia virus vector integration

Abstract

Retroviral vectors have been used in successful gene therapies. However, in some patients, insertional mutagenesis led to leukemia or myelodysplasia. Both the strong promoter/enhancer elements in the long terminal repeats (LTRs) of murine leukemia virus (MLV)-based vectors and the vector-specific integration site preferences played an important role in these adverse clinical events. MLV integration is known to prefer regions in or near transcription start sites (TSS). Recently, BET family proteins were shown to be the major cellular proteins responsible for targeting MLV integration. Although MLV integration sites are significantly enriched at TSS, only a small fraction of the MLV integration sites (<15%) occur in this region. To resolve this apparent discrepancy, we created a high-resolution genome-wide integration map of more than one million integration sites from CD34(+) hematopoietic stem cells transduced with a clinically relevant MLV-based vector. The integration sites form ∼60,000 tight clusters. These clusters comprise ∼1.9% of the genome. The vast majority (87%) of the integration sites are located within histone H3K4me1 islands, a hallmark of enhancers. The majority of these clusters also have H3K27ac histone modifications, which mark active enhancers. The enhancers of some oncogenes, including LMO2, are highly preferred targets for integration without in vivo selection.

Importance: We show that active enhancer regions are the major targets for MLV integration; this means that MLV preferentially integrates in regions that are favorable for viral gene expression in a variety of cell types. The results provide insights for MLV integration target site selection and also explain the high risk of insertional mutagenesis that is associated with gene therapy trials using MLV vectors.

FIG 1

MLV vector integration sites are highly enriched at active TSS and promoter regions. (A) MLV vector and HIV lentivector integration frequencies in the region near TSS (±1 kb) are compared to random sites and represented as fold enrichment on the left axis; percentages of total integration sites are on the right axis. (B) Mapping of integration sites from MLV vector, HIV lentivector, and a random selection within 5,000 bp upstream and downstream of TSS. MLV integration sites peak at both the promoter region upstream of active TSS (−500 bp) and the region downstream of TSS (+500 bp). In contrast, no such peaks are observed near inactive TSS/promoters, which have few, if any, H3K4me3 marks (black). There is a sharp dip at the region immediately adjacent to the TSS. HIV integration is reduced near TSS and increases downstream of TSS in the gene body. (C) Promoters are sorted into bins of 100 each based on gene expression level in CD34⁺ cells. MLV vector integrations are counted in each bin with the ±1-kb region of promoters/TSS.

FIG 2

MLV vector integration sites form tight clusters that colocalize with enhancers and promoters. (A) MLV vector integration clusters on chr11. Integration sites are represented by small ticks with colors denoting the orientation of the provirus. Red represents positive-strand orientation, and blue represents negative-strand orientation. Some clusters colocalize with promoters/TSS (yellow box). Some are within introns (pink box). Others are intergenic (blue box). However, all MLV integration clusters colocalize with epigenetic marks for active enhancers and promoters, including H3K4me1, H3K27ac, and H3K4me3. (B) Percentage and fold enrichment of MLV and HIV vector integration sites within peaks of specific epigenetic marks identified by ChIP-seq in CD34⁺ cells. Fold enrichment was compared to frequencies of random (Rnd) integration sites. (C) Venn diagrams showing the overlap of MLV vector integration site peaks with H3K4me1 peaks or the overlap of random peaks with H3K4me1 peaks. The majority of the MLV vector integration site peaks (89%) overlap a subset of H3K4me1 peaks. Base pairs that overlap are shown in parentheses. (D) Percentage of integration sites associated with active/inactive enhancers identified by H3K4me1 alone or H3K4me1 and H3K27ac together. A total of 87% of MLV integration sites are within enhancers marked by both H3K4me1/H3K27ac or H3K4me1 alone. (E) Integration site distribution around the peaks of specific epigenetic marks. Integration frequencies at each base surrounding the peak summit (±2 kb) were calculated and are presented as heatmaps. MLV vector integration is associated with H3K4me1, H2AZ, and H3K4me3. Integration of the HIV lentivector is associated with H3K36me3. Both MLV and HIV vectors avoid repressed regions.

FIG 3

MLV integration sites are cell type specific. (A) MLV integration site enrichment in cell type-specific H3K4me1 and H3K27ac peaks. (Left) MLV vector integration in CD34⁺ cells is enriched in CD34⁺ cell-specific H3K4me1 peaks, whereas MLV integration in CD4⁺ cells is enriched only in CD4⁺ cell-specific H3K4me1 peaks. (Right) A similar cell type-specific preference was seen for H3K27ac peaks. (B) MLV integration sites are clustered at the enhancer for the LMO2 gene in CD34⁺ cells but not in CD4⁺ cells. In CD34⁺ cells, the enhancer region shows marks that are characteristic of active enhancers: high levels of H3K4me1, H3K27ac, and H3K4me3 marks and a low level of the repressive mark H3K27me3. However, in CD4⁺ cells, the levels of the active marks are much lower and the level of the repressive mark H3K27me3 is higher. (C) CD34⁺ cell-specific MLV vector integration site cluster near the HOXA10 gene. No MLV integration sites were found in this region in CD4⁺ cells. (D and E) CD4⁺ cell-specific MLV integration site clusters near genes expressed in CD4⁺ cells. The CD4⁺ cell-specific clusters have a much smaller number of integrations than the CD34⁺ cell-specific clusters. (E) No integration sites were found in the same window in CD34⁺ cells.

FIG 4

MLV integration targeting model. The MLV preintegration complex (PIC) is targeted to active enhancer and promoter regions through interaction with BET proteins (BRD2, BRD3, and BRD4), which interact with histone acetyl modifications in active enhancers and promoters.