Foamy Virus Vector Carries a Strong Insulator in Its Long Terminal Repeat Which Reduces Its Genotoxic Potential

Abstract

Strong viral enhancers in gammaretrovirus vectors have caused cellular proto-oncogene activation and leukemia, necessitating the use of cellular promoters in "enhancerless" self-inactivating integrating vectors. However, cellular promoters result in relatively low transgene expression, often leading to inadequate disease phenotype correction. Vectors derived from foamy virus, a nonpathogenic retrovirus, show higher preference for nongenic integrations than gammaretroviruses/lentiviruses and preferential integration near transcriptional start sites, like gammaretroviruses. We found that strong viral enhancers/promoters placed in foamy viral vectors caused extremely low immortalization of primary mouse hematopoietic stem/progenitor cells compared to analogous gammaretrovirus/lentivirus vectors carrying the same enhancers/promoters, an effect not explained solely by foamy virus' modest insertional site preference for nongenic regions compared to gammaretrovirus/lentivirus vectors. Using CRISPR/Cas9-mediated targeted insertion of analogous proviral sequences into the LMO2 gene and then measuring LMO2 expression, we demonstrate a sequence-specific effect of foamy virus, independent of insertional bias, contributing to reduced genotoxicity. We show that this effect is mediated by a 36-bp insulator located in the foamy virus long terminal repeat (LTR) that has high-affinity binding to the CCCTC-binding factor. Using our LMO2 activation assay, LMO2 expression was significantly increased when this insulator was removed from foamy virus and significantly reduced when the insulator was inserted into the lentiviral LTR. Our results elucidate a mechanism underlying the low genotoxicity of foamy virus, identify a novel insulator, and support the use of foamy virus as a vector for gene therapy, especially when strong enhancers/promoters are required.IMPORTANCE Understanding the genotoxic potential of viral vectors is important in designing safe and efficacious vectors for gene therapy. Self-inactivating vectors devoid of viral long-terminal-repeat enhancers have proven safe; however, transgene expression from cellular promoters is often insufficient for full phenotypic correction. Foamy virus is an attractive vector for gene therapy. We found foamy virus vectors to be remarkably less genotoxic, well below what was expected from their integration site preferences. We demonstrate that the foamy virus long terminal repeats contain an insulator element that binds CCCTC-binding factor and reduces its insertional genotoxicity. Our study elucidates a mechanism behind the low genotoxic potential of foamy virus, identifies a unique insulator, and supports the use of foamy virus as a vector for gene therapy.

Keywords: CCCTC-binding factor (CTCF); CRISPR/Cas; foamy virus; gene insulator; gene therapy; genotoxicity.

FIG 1

The immortalization frequency and replating efficiency of FV are significantly lower than those of LV and GV vectors. (A) Schematic representation of the proviral forms of the vectors. An SFFV LTR-driven GV vector, RSF91.eGFP.pre (SFFV-GV), and an MSCV LTR-driven GV, MSCV.eGFP.pre, are shown. The U3 LTR region of the SFFV-GV contains the enhancer/promoter elements of the SFFV and drives expression of eGFP cDNA. SIN lentiviral vectors, RRL.ppt.SFFV.eGFP.pre (SFFV-LV), have a 400-bp LTR deletion and are driven by the enhancer-promoter elements from the SFFV U3 LTR region placed internally, upstream of eGFP; similarly, RRL.ppt.MSCV.eGFP.pre (MSCV-LV) is driven by an internal MSCV enhancer/promoter element from the MSCV U3 LTR region. The ΔΦ series of vectors represent FV vectors with a 582-bp LTR deletion and are driven by the internal enhancer-promoter elements derived from SFFV and MSCV. All the vectors carry the eGFP cDNA. Pr-less is a promoterless vector (ΔΦ.eGFP). The wPRE is present downstream of eGFP for all vectors except ΔΦMSCV.eGFP. Δ represents an LTR with a U3 deletion. The vectors are not drawn to scale. (B and C) The replating frequencies of immortalized clones were assessed at 2 and 5 weeks. The x axis represents replating frequencies at 2 weeks (left) and 5 weeks (right) normalized to VCN. The y axis represents vectors tested using the IVIM assay. The immortalization potentials of SFFV-FV (B) and MSCV-FV (C) were compared to that of SFFV-GV. The replating frequencies of SFFV-LV and MSCV-LV carrying internal SFFV or MSCV enhancer/promoter elements were also compared to that of SFFV-GV. Mock transductions were done without addition of virus and were negative controls for each experiment. A promoterless FV was also included as a negative control. Each symbol represents the replating frequency normalized for VCN from one independent transduction experiment using the specified vector. The data points on the left of the vertical line represent independent transductions with no replating clones. The replating frequency was calculated based on Poisson statistics using L-Calc software and was normalized to the mean VCN of the Lin⁻ bulk culture population prior to replating. Fold reductions in the frequencies of immortalized mutants are indicated in the graphs. Statistical significance between vector backbones is indicated by P values; ***, P < 0.001.

FIG 2

CRISPR/Cas9 facilitated insertion of GV, LV, and FV proviral sequences into a known locus previously shown to increase expression of LMO2. (A) General outline of CRISPR/Cas9 insertion of the proviral sequences into the LMO2 gene. The gRNA/Cas9 ribonucleoprotein complex creates a double-strand break (DSB) near the insertion site. The DSB is generally repaired by nonhomologous end joining (NHEJ) or by homologous recombination (HR) if a donor DNA, encoding the designed genetic modification flanked by homology arms (HA), is provided. (B) gRNA target sequences (purple hatched arrows) were chosen that target the LMO2 locus near the insertion site clinically found to be associated with insertional LMO2 transactivation and leukemogenesis. Insertion of proviral sequences occurred at the location indicated by the red arrow. (C) Donor templates for HR were plasmids constructed to insert vector sequences for GV, LV, and FV into the LMO2 gene at this locus. Viral sequences are flanked by HA corresponding to the region on either side of the insertion site. Each HA is ∼600 bp in length. All three vectors utilize the SFFV enhancer/promoter and carry an eGFP cDNA. Δ represents an LTR with a U3 deletion.

FIG 3

Establishing HeLa LMO2 clones. Following transfection and successful insertion of the proviral sequence via CRISPR/Cas9-mediated HR, cells express GFP. (A) Editing efficiency of HeLa cells assessed by GFP expression at 2 weeks posttransfection. GFP-positive cells were sorted into single cells to establish clones. (B) Schema for the two PCRs performed on each clone to detect homology-directed repair. (C) PCR 1 amplicon, which bridges the 5′ HA. (D) PCR 2 amplicon, which bridges the 3′ HA. PCR 2 used a different FWD primer for FV. Only correctly integrated sequences produced amplicons.

FIG 4

FV induces LMO2 mRNA expression to a lesser extent than either GV or LV. cDNA was generated from GV, LV, and FV clones, and LMO2 mRNA expression was determined using RT-PCR. Depicted are data from experiments using two different primer-probe sets and two different endogenous controls. (A) Hs001534473_m1 primer-probe set and GAPDH endogenous control. n = 8 GV, 9 LV, and 11 FV clones. (B) Hs001534473_m1 primer-probe set and PPIA endogenous control. n = 7 GV, 8 LV, and 8 FV clones. (C) Hs00277106_m1 primer-probe set and GAPDH endogenous control. n = 8 GV, 9 LV, and 11 FV clones. (D) Hs00277106_m1 primer-probe set and PPIA endogenous control. n = 7 GV, 8 LV, and 8 FV clones. The error bars indicate SEM.

FIG 5

FV induces LMO2 protein expression to a lesser extent than either GV or LV. Western blot analysis for LMO2 expression (top) was performed on SFFV-GV, SFFV-LV, and SFFV-FV clones. The clones used are indicated by the letter and number designations above each row. Untransduced HeLa cells (−) and K562 cells (+) served as negative and positive controls, respectively. Endogenous GAPDH expression (bottom) was used as a loading control.

FIG 6

High number of CTCF insulator binding sites in the sequence of FV. (A) Proviral sequences (except SFFV, eGFP, and wPRE sequences) of GV, LV, and FV were analyzed in silico for potential CTCF binding motifs. The approximate locations of predicted binding motifs are indicated by asterisks above the construct. Multiple asterisks indicate multiple overlapping predicted binding motifs. (B) PWM scores for the predicted CTCF binding motifs.

FIG 7

CTCF binding of FV proviral sequence. (A) CTCF-ChIP of the FV A2 clone, followed by qualitative PCR, was performed to interrogate in-cell binding of CTCF to the predicted binding sites. PCR was performed on the ChIP input for HeLa control cells and the FV A2 clone and on the ChIP product for the FV A2 clone. The amplicons for PCRs of H19, FV1, FV2, FV5, FV6, and FV7 were 165, 157, 188, 110, 115, and 155 bp, respectively. (B) Fluorescently labeled probes corresponding to predicted CTCF binding sites in FV and LV proviral sequences were allowed to bind recombinant CTCF protein and were resolved by EMSA, demonstrating binding between the FV2 probe and CTCF. (C) Competitive binding assay between the FV2 probe and unlabeled H19 probe. H19 was provided at the indicated molar excess. (D) Sequence of the FV2 probe (top) with predicted CTCF binding sites indicated in red and blue. Mutant FV2 probes (1 to 6) are listed, with mutated regions underlined in green. (E) EMSA utilizing mutant probes. The original FV2 probe was used as a positive control.

FIG 8

Induction of LMO2 mRNA expression by FV is increased when the insulator is removed, and induction of LMO2 mRNA expression by LV is decreased when the insulator is added to the LV LTR. cDNA was generated from LMO2-modified clones containing FV, LV, FV with no insulator (ins.), and LV with the FV insulator placed in the LTR. LMO2 mRNA expression was determined using qRT-PCR. The Hs001534473_m1 primer-probe set and PPIA endogenous control were used to acquire the data. n = 5, 6, 17, and 9 clones, respectively. The error bars indicate SEM.

FIG 9

LMO2 copy number analysis. (A) FISH was performed for a region overlapping the targeted LMO2 loci, revealing 4 LMO2 alleles in HeLa control cells. FISH was performed using an RP11-1006P23 FISH probe (Empire Genomics, Buffalo, NY) recognizing chr11 (33,736,494 to 33,907,488). (B) Copy number analysis was performed across the insertion site of LMO2 (intron 1). The numbers of nontargeted/WT LMO2 alleles were calculated relative to unedited HeLa cells. There were relatively similar nontargeted/WT LMO2 copy numbers between clones. n = 3 for each clone. The HeLa control sample is represented by the white bar. (C) PCR amplification of the region bridging the gRNA target site. Three LV clones (indicated by asterisks [panel B] and arrows [panel C]) were found to have a 261-bp deletion upon sequencing of the PCR product. One LV clone and three FV clones did not amplify.