Lens epithelium-derived growth factor fusion proteins SEE COMMENTARY redirect HIV-1 DNA integration Andrea L. Ferrisa,1, Xiaolin Wub,1, Christina M. Hughesc, Claudia Stewartb, Steven J. Smitha, Thomas A. Milnec, Gang G. Wangc, Ming-Chieh Shund, C. David Allisc, Alan Engelmand, and Stephen H. Hughesa,2 aHIV Drug Resistance Program, National Cancer Institute, Frederick, MD 21702; bLaboratory of Molecular Technology, SAIC-Frederick, Inc., Frederick, MD 21702; cLaboratory of Chromatin Biology and Epigenetics, The Rockefeller University, New York, NY 10065; and dDepartment of Cancer Immunology and AIDS, Dana-Farber Cancer Institute, Boston, MA 02115 Communicated by John M. Coffin, Tufts University School of Medicine, Boston, MA, December 17, 2009 (received for review November 13, 2009) Lens epithelium-derived growth factor (LEDGF) fusion proteins can by treatment with siRNA. Relatively small amounts of residual direct HIV-1 DNA integration to novel sites in the host genome. The LEDGF can support HIV-1 DNA integration (18, 22). If the C terminus of LEDGF contains an integrase binding domain (IBD), PWWP domain of LEDGF is removed, the truncated protein and the N terminus binds chromatin. LEDGF normally directs in- cannot support HIV-1 replication. However, fusion proteins in tegrations to the bodies of expressed genes. Replacing the N ter- which the PWWP domain of LEDGF is replaced with the linker minus of LEDGF with chromatin binding domains (CBDs) from other histone H1 or with a peptide derived from the chromatin binding proteins changes the specificity of HIV-1 DNA integration. We chose Kaposi’s sarcoma herpes virus LANA protein do support effi- two well-characterized CBDs: the plant homeodomain (PHD) finger cient HIV-1 replication (23). This result means that these fusion from ING2 and the chromodomain from heterochromatin binding proteins are able to bind both chromatin and the PIC in a way protein 1α (HP1α). The ING2 PHD finger binds H3K4me3, a histone that supports viral integration. We show that replacing the mark that is associated with the transcriptional start sites of ex- PWWP domain and AT hooks of LEDGF with other chromatin pressed genes. The HP1α chromodomain binds H3K9me2,3, histone binding domains (CBDs) supports viral integration and redirects marks that are widely distributed throughout the genome. A fusion the viral DNA integration to sites in the genome that reflect the protein in which the ING2 PHD finger was linked to the LEDGF IBD chromatin binding patterns of the CBDs. The ability to redirect directed integrations near the start sites of expressed genes. A HIV-1 DNA integration may help solve one of the problems similar fusion protein in which the HP1α chromodomain was linked associated with using retroviral vectors in gene therapy: the to the LEDGF IBD directed integrations to sites that differed from unintentional activation of host cell genes by the nearby insertion both the PHD finger fusion–directed and LEDGF-directed integra- of vector DNA (13, 24). tion sites. The ability to redirect HIV-1 DNA integration may help solve the problems associated with the activation of oncogenes Results when retroviruses are used in gene therapy. CBD–IBD Fusions Can Functionally Replace LEDGF to Support Efficient HIV-1 DNA Integration. We chose two CBDs that are well defined chromatin | histone marks | lentiviral vectors both structurally and functionally, selecting one that binds pri- marily to euchromatin and one that can bind to more tightly etroviruses can integrate their DNAs at many sites in host packed heterochromatin. The first CBD was the plant homeo- RDNA (1), but different retroviruses have different integration domain (PHD) finger from ING2, which is known to bind his- site preferences (2–9). HIV-1 and simian immunodeficiency virus tone 3 lysine 4 trimethylations (H3K4me3) (25, 26). The other DNAs preferentially integrate into expressed genes, murine leu- CBD was the chromodomain from heterochromatin binding α α kemia virus (MLV) DNA preferentially integrates near tran- protein 1 (HP1 , also known as Cbx5), which has been shown to bind to histone 3 that is either di- or trimethylated on lysine 9 scriptional start sites (TSSs), and avian sarcoma leukosis virus MICROBIOLOGY (ASLV) and human T cell leukemia virus (HTLV) DNAs inte- (H3K9me2,3). There is evidence that H3K9me2,3 marks are grate nearly randomly, showing a slight preference for genes. This widely distributed throughout the genome and are found both in suggests that the preintegration complexes (PICs) of the different poorly expressed and actively expressed genes (27). retroviruses interact with different factors in host cell chromatin We generated fusion proteins in which the two CBDs were and that the integration site preferences are determined by the joined to the LEDGF IBD. To avoid problems from the expression of small amounts of LEDGF that can occur in siRNA knockdowns, distribution of the different host cell factors. Studies of the spe- – fi cificity of integration of yeast retrotransposons, which are dis- we expressed the CBD IBD fusions in a mouse embryo broblast tantly related to retroviruses, provide strong support for this cell line that was derived from a LEDGF knockout mouse (MEF- interpretation (10–12). The host factors that MLV, ASLV, and KO) (19). As a positive control, we expressed human LEDGF (huLEDGF). The expression plasmid contains a CMV promoter, a HLTV PICs interact with are not known. DNA segment encoding huLEDGF or the CBD–IBD fusion, and HIV-1 integrase (IN) binds specifically to lens epithelium- an internal ribosome entry site (IRES) followed by the coding derived growth factor (LEDGF). A crystal structure shows that the region for GFP (19). The expression plasmid was introduced into integrase binding domain (IBD), located in the C-terminal portion of LEDGF, makes important contacts with both the catalytic core and N-terminal domains of IN (13). The N-terminal portion of Author contributions: A.L.F., X.W., and C.M.H. designed research; A.L.F., X.W., C.S., and LEDGF contains a PWWP domain and AT hooks that bind S.J.S. performed research; C.M.H., T.A.M., G.G.W., M.-C.S., C.D.A., and A.E. contributed chromatin and DNA, respectively (14) (Fig. S1). The PWWP do- new reagents/analytic tools; A.L.F., X.W., and C.M.H. analyzed data; and A.L.F., X.W., main is the primary chromatin binding determinant, although the C.M.H., C.D.A., A.E., and S.H.H. wrote the paper. protein or modification to which it binds is currently unknown (15, The authors declare no conflict of interest. 16). LEDGF promotes the association of the PIC with the host See Commentary on page 2735. genome, enhancing the integration of HIV-1 DNA (16–20). 1A.L.F. and X.W. contributed equally to this work. In assays that depend on the integration of viral DNA, viral 2To whom correspondence should be addressed. E-mail: [email protected]. – titer is reduced 5 100-fold in the absence of LEDGF (18, 19, 21). This article contains supporting information online at www.pnas.org/cgi/content/full/ In some of these experiments, the level of LEDGF was reduced 0914142107/DCSupplemental. www.pnas.org/cgi/doi/10.1073/pnas.0914142107 PNAS | February 16, 2010 | vol. 107 | no. 7 | 3135–3140 Downloaded by guest on September 25, 2021 MEF-KO cells by nucleofection, and cells that expressed similar gene expression in the MEF-KO cells. A total of 17,306 well- levels of GFP were selected by FACS (Fig. S2). The huLEDGF and characterized RefSeq Genes were sorted into groups of 100 ac- the CBD–IBD fusions were HA tagged. Expression of the fusion cording to their RNA levels, and the integration data were sorted proteins was confirmed by Western blot. Judged by the Western on the basis of these data. As expected, huLEDGF directed in- blots, the sorted cells contained similar amounts of huLEDGF and tegrations to expressed genes. The most highly expressed genes the CBD–IBD proteins. GFP-expressing cells were infected with an have been reported to be less efficient targets than genes that are HIV-1 vector that expresses luciferase. Although there is some moderately well expressed. Because huLEDGF-directed integra- variation in the assay, the titer in MEF-KO cells that expressed high tions can occur anywhere in a gene, simply counting the integra- levels of huLEDGF was ≈100-fold higher than the titer in tions in each gene favors larger genes. When we normalized the unmodified parental MEF-KO cells, which agrees with the results integration data according to the size of the genes (discussed reported by Shun et al. (19). below), there was a simple positive correlation (r = 0.94) between fi Titer, measured by luciferase activity, depends both on the ef - the level of expression and the number of huLEDGF-directed ciency of integration (the number of integrated proviruses) and on integrations (Fig. 1). their level of expression. Expression of the ING2 PHD finger IBD fusion (ING2–IBD) in MEF-KO cells increased the titer of the ING2 PHD Finger Fusion Directs HIV-1 DNA to Integrate Near TSSs. A HIV-1 vector by ≈60-fold (three separate experiments showed in- total of ≈15,000 independent ING2–IBD-directed integrations in α creases of 67-, 58-, and 59-fold), and the HP1 chromodomain IBD single-copy DNA were obtained from a combination of Sanger α– ≈ fusion (HP1 IBD) by 80-fold (two separate experiments showed sequencing and two completely independent experiments ana- increases of 77- and 80-fold), demonstrating that integration was lyzed by 454 pyrosequencing. Approximately half (50.3%) of the fi fi ef cient and that a signi cant fraction of the proviruses were ING2–IBD-directed integration sites were within 2.5 kb of a TSS.