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

Vector design for comparative genotoxicity.

The spleen focus-forming virus (SFFV) vector, a GV, was previously reported to generate a high frequency of immortalized clones in the in vitro immortalization (IVIM) assay (8, 9), which was correlated with the occurrence of leukemia in mice (9) and with a high incidence (80 to 100%) of myelodysplastic syndrome and leukemia in the CGD and WAS gene therapy trials. We constructed analogous enhanced green fluorescent protein (eGFP)-encoding FV and LV vectors carrying the internal enhancers/promoters from the SFFV LTR or FV and LV vectors carrying internal enhancers/promoters from the murine stem cell virus (MSCV) LTR (also a GV): (i) ΔΦSF.eGFP (SFFV-FV) carries an internal SFFV enhancer/promoter; (ii) ΔΦMSCV.eGFP (MSCV-FV) carries an internal MSCV enhancer/promoter. These vectors were compared to two analogous LV vectors, RRL.ppt.SF.eGFP.pre (SFFV-LV), which carries an internal SFFV enhancer/promoter, and RRL.ppt.MSCV.eGFP.pre (MSCV-LV), which carries an internal MSCV enhancer/promoter, and to two analogous GV vectors, SF91-eGFP.pre (SFFV-GV), which is driven by the SFFV LTR, and MSCV.eGFP.pre, which is driven by the MSCV LTR, as positive controls with known high genotoxic potential (Fig. 1A). A promoterless (Pr-less) FV vector, ΔΦ.eGFP, and mock transductions were included as negative controls.

FV vectors showed significantly less immortalization of primary mouse HSPCs than GV and LV vectors.

To compare the genotoxic potentials, the above-mentioned viral vectors were used in the IVIM assay. This assay is widely used as a preclinical screening tool and is particularly sensitive for relative quantitative detection of myeloid lineage-related genotoxicity (37).

Lineage-negative (Lin⁻) cells from bone marrow of wild-type (WT) (C57BL/6J) mice were transduced with the SFFV- and MSCV-GV, -LV, and -FV vectors using optimized transduction protocols for each vector (5, 21, 38). The cells were expanded for 2 weeks and then cloned, as previously described (39). By 2 weeks, untransduced/mock-transduced Lin⁻ cells terminally differentiated and died. If vector insertion conferred a proliferative potential, clonal outgrowth occurred, creating immortalized clones. The replating frequencies (immortalization frequencies) of cells transduced with GV, LV, and FV vectors were assessed at 2 weeks and at 5 weeks. All wells with immortalized cells were picked for expansion. Cells transduced with SFFV-driven GV or LV vectors expanded robustly, even at 5 weeks. Transduction efficiencies (measured by GFP marking) of Lin⁻ cells in SFFV-GV-, MSCV-GV-, SFFV-LV-, MSCV-LV-, SFFV-FV-, and MSCV-FV-transduced Lin⁻ cells were 91%, 70%, 89%, 99%, 93%, and 75%, respectively. The average VCN in Lin⁻ cells transduced with the SFFV-GV, MSCV-GV, SFFV-LV, MSCV-LV, SFFV-FV, and MSCV-FV vectors were 8 ± 2, 9 ± 0.3, 10 ± 2, 27 ± 2, 7 ± 0.8, and 8.5 ± 0.7 (mean ± standard error of the mean [SEM]), respectively (Table 1).

The fitness of immortalized clones (i.e., the ability to be replated and expanded) after transduction with GV or LV vectors with SFFV/MSCV enhancers at 5 weeks was similar to that at 2 weeks. The number of immortalized clones with SFFV-GV and SFFV-LV was consistent with previously reported studies (40). Notably, however, the immortalization frequencies of SFFV-FV and MSCV-FV were remarkably lower, by more than 2 orders of magnitude, than those of the analogous SFFV-GV and MSCV-GV (P < 0.01) (Fig. 1B and C and Table 1). The analogous SFFV-LV and MSCV-LV showed a 10- to 14-fold reduction in immortalization frequency compared to SFFV-GV, consistent with prior reports (40). In addition, the immortalized clones derived from FV transduction were not as fit as those derived from LV or GV transduction, as they had a lower expansion potential than clones with SFFV-GV and SFFV-LV insertions and therefore lower 5-week replating frequencies. Mean Sca-1 and c-Kit expression trended lower for SFFV-FV and MSCV-FV clones (69.6 and 57.5%, respectively) than for SFFV-GV, SFFV-LV, and MSCV-LV clones (89.0, 76.6, and 70.5%, respectively), but the differences were not statistically significant (Mann-Whitney U test, one tailed). The percentage of unique insertions from FV and LV vectors with respect to gene transcriptional units and nongenic/repeat sequences demonstrated a unique integration profile for FV compared to LV insertions. Frequencies of integration sites for FV within genes, in nongenic/repeat sequences, and nonassignable were 27, 62, and 11%, respectively; for LV, they were 60, 6, and 34%, respectively. The numbers of unique insertions for SFFV-FV and MSCV-FV were 121 and 65, respectively. The number of unique insertions identified for MSCV-LV was 270.

To exclude differences in immortalization frequency due to VCN, the immortalization frequency of each vector was normalized to the VCN (Fig. 1B and C), allowing comparative analysis of the relative genotoxicity. When normalized for VCN, the MSCV-GV vector had a 3-fold-lower immortalization potential than the SFFV-GV vector. Importantly, at 2 weeks, the SFFV-FV and MSCV-FV vectors showed a 110-fold- and 156-fold-lower immortalization potential, respectively, than the SFFV-GV vector. By 5 weeks, the immortalization potentials of FV vectors declined even further, resulting in 155-fold- and 414-fold-lower immortalization potentials of SFFV-FV and MSCV-FV vectors, respectively, than of the SFFV-GV vector. The SFFV-LV and MSCV-LV vectors showed 12- and 14-fold reductions in immortalization frequency compared to the SFFV-GV vector, consistent with prior reports (9, 40). Supporting the notion that immortalization in this assay occurred secondarily to vector integration, (i) there were no detectable immortalized clones in the mock-transduced progenitor cultures and (ii) promoterless FV had live wells at 2 weeks, but they contained fewer cells that were lost by 5 weeks, suggesting that they were not truly immortalized.

CRISPR/Cas9-based targeted insertional genotoxicity.

The SFFV enhancer has been shown in the IVIM assay, in mice, and in human trials to be one of the most genotoxic enhancers (5, 10, 40). The remarkably reduced genotoxicity (150- to 400-fold less) in the IVIM assay from the SFFV/MSCV enhancers in an FV vector, as shown here, could not be fully explained by the reported 2-fold-higher propensity of FV to integrate in nongenic regions, especially when FV tends to integrate near TSS, like GV. We hypothesized that the FV backbone may have an enhancer-blocking/insulator effect. To assess the potential enhancer-blocking functionality of the vector backbone without the confounding effects of the enhancer/promoter, transgene, or integration site, we targeted proviral forms of SFFV-GV, SFFV-LV, and SFFV-FV into the LMO2 gene at the retroviral integration site (RIS), known to cause multiple cases of secondary leukemia (1–3). In order to isolate the genotoxic effects of viral vector backbone sequences from integration site effects, a CRISPR (clustered regularly interspaced short palindromic repeat)/Cas9-based assay was devised that allowed integration of the proviral sequences of GV, LV, and FV, all carrying an eGFP transgene driven by the SFFV enhancer/promoter, at precisely the same locus within LMO2 and in the same direction (Fig. 2A). The insertion site for the viral-vector sequences was based on a previous report of secondary leukemia in a patient following gene therapy for SCID-X1 using a GV vector (41).

Five potential guide RNA (gRNA) target sequences, each with low predicted off-target activity and in close proximity to the insertion site, were identified (Fig. 2B). LMO2 gRNA 5 was the most efficient, with an indel generation efficiency of 24.1% (efficiencies for gRNAs 1 to 4 were 0, 13.3, 0, and 1.5%, respectively), and was used for subsequent experiments. Donor plasmids containing proviral cassettes for GV, LV, or FV were cloned (Fig. 2C). Sequences were constructed in reverse orientation to match the directionality of the insertion described previously (41). HeLa cells, which have very low LMO2 mRNA expression and lack LMO2 protein (42), were simultaneously transfected with the gRNA/Cas9 plasmid and one of the donor (provirus-containing) plasmids. Approximately 2 weeks after transfection, 9.7%, 28.2%, and 12.1% of GV-, LV-, and FV-transfected pools showed GFP transgene expression when green fluorescent protein-positive (GFP⁺) cells were sorted into single cells (Fig. 3A). Clones were then harvested and screened by PCR for homology-directed insertion of the proviruses (Fig. 3B to D).

When editing with CRISPR/Cas9, it is possible to edit more than one allele in a given clone, especially in highly transfectable cell lines. In our case, this would either result in proviral sequences being integrated into multiple LMO2 alleles in a given HeLa cell (provirus-targeted alleles) or the double-strand break (DSB) would be repaired by nonhomologous end joining (NHEJ), creating a small indel (termed “edited nontargeted” alleles here). Since HeLa cells have very low LMO2 mRNA expression and do not express LMO2 protein and the proviruses are targeted to an intron of the LMO2 gene, LMO2 expression would be influenced by the virus enhancers only in the provirus-targeted alleles. Moreover, the reading frame of LMO2 (and its mRNA expression) would not be altered by an edited nontargeted allele. The proviral plasmid could also randomly integrate into the genome in HeLa cells, resulting in GFP expression, but would not affect LMO2 expression. Due to the potential for random integration, we could not directly determine whether the clones had more than one provirus-targeted LMO2 allele by quantitative PCR (qPCR) of the proviral sequences. We therefore took several steps to calculate the targeted-allele copy numbers indirectly (see Materials and Methods and Table 2). Briefly, fluorescence in situ hybridization (FISH) for the LMO2 locus on control HeLa cells showed four LMO2 alleles. Next, copy number analysis was used to detect unedited alleles and alleles containing small indels. PCR across the target site with a larger amplicon, followed by gel electrophoresis and sequencing, was used to detect larger indels that would not be detected by copy number analysis. One LMO2 allele with a 261-bp deletion was found in three of the LV clones that initially showed only one nontargeted/WT allele on copy number analysis. The number of targeted alleles for a given clone equals the number of LMO2 loci with the number of nontargeted alleles detected by copy number analysis and PCR subtracted (Table 2). Overall, comparable numbers of LMO2 alleles were targeted (2 or 3 targeted alleles in most clones) with FV, LV, or GV.

The advantage of using the HeLa cell line without significant endogenous LMO2 expression is that editing events that abrogate gene expression does not significantly affect the overall increase in LMO2 expression due to directed proviral insertion events. We determined LMO2 mRNA expression in the generated HeLa clones by quantitative real-time (qRT) PCR with two probe and primer sets. The primers were selected to detect all spliced transcript variants expressed from the LMO2 promoter from both modified and WT alleles. Data from both primer sets using two different loading controls were very similar (Fig. 4A to D). Overall, we found that the SFFV enhancer in GV demonstrated the greatest fold increase in LMO2 mRNA expression (median increase, 280- ± 23-fold over unmodified HeLa cells), followed by the SFFV enhancer in LV (median increase, 200- ± 27-fold). It should be noted that SFFV-GV provirus had two copies of the enhancer at either LTR, while SFFV-LV (and SFFV-FV) had only one copy of the SFFV enhancer. However, the same SFFV enhancer in FV showed a remarkably lower (median, 45- ± 7-fold) increase in LMO2 mRNA expression, 4- and 6-fold lower expression than was seen with the SFFV enhancer in LV and GV, respectively. We then performed a Western blot analysis to detect LMO2 protein expression from three representative clones among the ones used for qRT-PCR (Fig. 5 and Table 2). We could not detect any LMO2 expression in SFFV-FV clones, which was similar to baseline in mock-transduced (nonedited) HeLa cells. However, significantly more LMO2 protein was detectable in GV and LV clones. Taken together, the qRT-PCR and Western blot analyses confirmed that the FV backbone/cis elements have a strong enhancer-blocking or insulator effect, which likely contributes to the reduced ability of the SFFV enhancer to upregulate the expression of LMO2.

In silico insulator analysis.

A likely mechanism for the observed enhancer-blocking effect of the FV backbone is that it contains one or more insulator elements. To test our hypothesis, we performed an in silico analysis for CTCF binding sites, the main insulator in vertebrates. The proviral sequences of GV, LV, and FV vectors (excluding the SFFV enhancer/promoter, eGFP, and woodchuck hepatitis virus posttranscriptional regulatory element [wPRE] sequences) were analyzed for predicted CTCF binding sites, or consensus sequences, using the CTCFBSDB 2.0 database (http://insulatordb.uthsc.edu/) (32) to identify core motifs for CTCF binding, represented as position weight matrices (PWMs). The algorithm searches for identified core motifs for CTCF binding sites and represents the motifs as PWMs. PWM scores correspond to the log odds of the observed sequence being generated by the motif versus being generated by the background. A PWM score of >3.0 is suggestive of a significant match. A limitation of the prediction tool is that it returns only the best match for a given PWM in a sequence. Therefore, other putative CTCF consensus sequences for a given motif within the same analyzed sequence are not revealed. To partially account for this, sequences were divided into several fragments, and each fragment was analyzed separately.

Analysis of the vector backbone sequences identified totals of 4, 6, and 26 motifs with PWM scores of >3 for GV, LV, and FV, respectively. The locations of the motifs in the vector backbone are depicted in Fig. 6A and Table 3 and their PWM scores in Fig. 6B. Besides the number of significant PWMs, the PWM scores for GV were lower in general, ranging from 3.3 to 3.4, while the PWM scores for LV ranged from 5.1 to 9.9 and those for FV ranged from 3.4 to 12.6. Motifs for FV were dispersed both in the LTR and the cis sequences compared to those for GV. Overall, the in silico analysis suggested that FV had a greater number of predicted CTCF binding motifs and motifs with higher PWM scores than GV and LV.

Mapping the enhancer-blocking element in the FV backbone.

To assess the binding of CTCF to the proviral sequences within HeLa cells, chromatin immunoprecipitation (ChIP) purification of CTCF-bound DNA was performed, followed by qualitative PCR for predicted binding sites within the FV proviral sequence (Fig. 7A). Using HeLa control cells, analysis of the ChIP input material showed amplification of only the H19-Igf2 locus, a known CTCF binding site. However, both the input material and the ChIP-purified DNA from one of the FV clones (FVA2) showed amplification of five tested sites with a high level of predicted binding. While ChIP analysis showed the presence of CTCF binding, the close proximity of the assayed regions limited the resolution between sites by this ChIP-PCR assay. Regardless, the ChIP-PCR assay demonstrated in-cell binding of CTCF to the FV proviral sequence.

An electrophoretic mobility shift assay (EMSA) was then used to map the predicted CTCF binding sites within LV and FV. Eighty- to 90-bp DNA fragments corresponding to predicted CTCF binding sites by in silico analysis, labeled at both 5′ ends with fluorescent dye, were used as the EMSA probe. Five of the six predicted binding sites in LV were probed. All predicted binding sites in FV were probed. Probes LV1, LV2, and FV8 contained two predicted CTCF binding motif sequences. EMSA was conducted using recombinant human CTCF. An H19 oligonucleotide containing a consensus sequence known to bind CTCF with high affinity was the positive control.

None of the LV probes demonstrated any binding to CTCF. However, probe FV2, corresponding to the sequence ATATCACTAGATGTCTCCCT (located in the LTR and containing four motifs with PWM scores of 4.1, 8.7, 8.0, and 5.9) demonstrated a significant band shift (Fig. 7B). Additionally, the labeled probe could be competed off with unlabeled H19 probe (Fig. 7C). The sequence of the FV2 probe was analyzed in silico for predicted CTCF binding sites. In addition to the previously predicted site, a second site was identified, 5′-TGTAGTTCA-3′, with a score of 6.8. The central region of the FV2 probe was divided into four regions (1 to 4). Region 1 contained the newly identified predicted binding site. Regions 2 and 3 contained the original predicted binding site. Six mutant probes were designed that replaced one or more of the four regions with a scrambled DNA sequence (Fig. 7D). The sequences were analyzed to ensure that no new predicted CTCF binding sites were created. Mutating region 1, 2, or 3 reduced CTCF binding (EMSA band intensities are shown in Fig. 7E). Mutating region 4 appeared to have no effect. Mutating regions 1 and 3 together or 2 and 3 together further reduced CTCF binding. Therefore, we found that CTCF binds the 36-bp sequence defined by regions 1 to 3 of the FV2 probe. A BLAST search (https://blast.ncbi.nlm.nih.gov/) using the defined sequence did not reveal matches to any sequences other than those of foamy virus.

Verifying insulator function.

To verify the insulator function of the defined CTCF binding sequence, (i) the 36-bp sequence was precisely excised from the proviral SFFV-FV sequence, leaving the rest of the sequence intact, and (ii) the proposed insulator was inserted into the LTR of the SFFV-LV proviral sequence. The modified proviral sequences were then inserted into the LMO2 gene using our CRISPR/Cas9-based targeted insertional genotoxicity assay, as before. Expression of LMO2 relative to control HeLa cells and LV and FV clones used previously was determined by qPCR using the Hs001534473_m1 primer-probe set and PPIA endogenous control (Fig. 8). Removing the insulator from the FV LTR resulted in a >5-fold increase in relative LMO2 expression (9.4 to 47.4; P = 0.001). Inserting the sequence into the LV LTR resulted in a >13-fold drop in relative LMO2 expression (21.0 to 1.6; P = 0.0002). Interestingly, placing the insulator into the LTR of SFFV-LV resulted in LMO2 expression that was only 1.6-fold higher than that of control HeLa cells and significantly less than that of the original SFFV-FV containing the insulator (P = 0.002). The overall lower relative expression seen in this assay compared to the prior assay using GV, LV, and FV was due to higher observed expression of LMO2 in the control HeLa cells. Based on our prior assay, the number of inserted proviral sequences did not correlate well with LMO2 expression. The copy number, determined by qPCR only, predicted that all of the new FV clones without the insulator had three correctly placed proviral sequences. Of the new LV clones containing the insulator, 6, 1, and 2 clones contained 1, 2, and 3 copies of the proviral sequence, respectively. Again, the number of inserted proviral sequences did not seem to correlate with expression levels.

Taking the data together, we show that FV LTRs contain a strong 36-bp CTCF binding motif with potent CTCF binding that produces an enhancer-blocking effect and serves to protect nearby genes from the enhancer activity of a delivered transgene. These data provide novel insight into the remarkably low immortalization potential of SFFV (and MSCV) in FV vectors, demonstrating a previously unreported and significant mechanism contributing to the lower genotoxicity of FV carrying strong viral enhancers.

Vector design and production.

The vectors SFFV-GV (RSF91.eGFP.pre) and SFFV-LV (RRL.ppt.SF.eGFP.pre) used in this study have been described previously (8, 57). MSCV-GV (MSCV.eGFP.pre) has also been described previously (58). MSCV-LV (RRL.ppt MSCV.eGFP.pre) was generated in the laboratory of Dennis Hickstein (National Cancer Institute, Bethesda, MD). A 388-bp region of the MSCV promoter (identical in sequence to the MSCV LTR enhancer/promoter in ΔΦMSCV CD18 [21]) was obtained through PCR amplification with artificial XhoI and AgeI ends and then cloned into XhoI/AgeI restriction enzyme-digested pRRLSIN.cPPT.PGK.eGFP.WPRE (Addgene, Cambridge, MA) to create pRRL.ppt.MSCV.eGFP.pre (M. J. Hunter and D. D. Hickstein, unpublished results). All FVs were in the ΔΦ backbone. The FV ΔΦMSCV.eGFP has been described previously (59, 60). The SFFV promoter replaced the MSCV promoter in the ΔΦMSCV.eGFP FV vector to create ΔΦSF.eGFP. All the vectors carry eGFP cDNA. The promoterless FV was derived from the ΔΦMSCV.eGFP vector by removing the MSCV enhancer/promoter and religation.

Ecotropic GV supernatants were produced in 293T cells by transient transfection, as described previously, and titers were determined on NIH 3T3 fibroblasts (American Type Culture Collection [ATCC]) (39). Virus titers were in the range of 10⁶ to 10⁷ infectious units (IU)/ml. The LV vectors SFFV-LV and MSCV-LV were produced by transient cotransfection of 293T cells (ATCC) as described previously (38). SFFV-FV, MSCV-FV, and Pr-less FV were produced by four-plasmid (pCiES [Env], pCiGSΔΨ [Gag], pCiPs [Pol], and vector [pΔΦ]) transient transfection, as described previously (59). pCiGSΔΨ is the Gag expression cassette (D. W. Russell, unpublished data) with a more complete deletion in the packaging signal. pΔΦ is a deleted FV backbone with a polylinker to insert the transgene cassette.

FVs were resuspended in Stemspan (Stem Cell Technologies, Vancouver, BC, Canada) containing 2% heat-inactivated fetal bovine serum (FBS) (HyClone, Logan, UT) and 5% dimethyl sulfoxide (DMSO) (Sigma, St. Louis, MO) and were stored frozen in 5% DMSO until use. Titers of Pr-less FV were determined by measuring the genome copy number of transduced HT1080 cells by quantitative real-time PCR using primers that recognize the wPRE, while titers of the other vectors were determined by quantifying GFP expression by fluorescence-activated cell sorter (FACS). The titers of FV were in the range of 3 × 10⁷ IU/ml to 1 × 10⁸ IU/ml.

Isolation of Lin⁻ cells.

Bone marrow from C57BL/6J mice was used for the isolation of Lin⁻ cells using biotinylated lineage-specific antibodies (lineage cell depletion kit; BD Biosciences, San Jose, CA) as described previously (39, 40). The biotin-labeled Lin⁻ cells were incubated with anti-biotin microbeads (Miltenyi Biotech), followed by magnetic sorting of unlabeled Lin⁻ cells. The isolated Lin⁻ cells were prestimulated for viral transductions in Stemspan medium (Stem Cell Technologies) containing 1% penicillin-streptomycin, 50 ng/ml mouse stem cell factor (mSCF), 100 ng/ml human interleukin 11 (hIL-11), and 10 ng/ml mIL-3.

In vitro immortalization assay.

Lin⁻ cells were prestimulated overnight in Stemspan medium containing 1% penicillin-streptomycin, 50 ng/ml mSCF, 100 ng/ml hIL-11, and 10 ng/ml mIL-3. On day 2, 100,000 Lin⁻ cells were used for each LV vector transduction. The Lin⁻ cells were transduced at a multiplicity of infection (MOI) of 20 twice at 8-h intervals using concentrated LV vector supernatants. For GV transduction, lineage-negative cells were prestimulated for 2 days in Stemspan-cytokine cocktail. GV transductions were performed on day 3 and day 4 on RetroNectin recombinant human fibronectin fragment (TaKaRa Bio Inc., Kusatsu, Shiga, Japan)-coated 24-well dishes preloaded with retroviral vectors SFFV-GV and MSCV-GV at an MOI of 20. After the final transductions, the transduced Lin⁻ cells were washed and expanded as bulk cultures in Stemspan-cytokine cocktail for 19 days.

FV stocks were rapidly thawed by adding warm prestimulation medium, and the Lin⁻ cells were transduced once with FV in 48-well plates precoated with RetroNectin recombinant human fibronectin fragment at a concentration of 8 μg/cm² at an MOI of 50. A higher MOI was chosen for transduction, as foamy virus transductions were done only once compared to two transductions with lentiviral vectors. After 16 h, the cells were washed and expanded in the Stemspan-cytokine cocktail. We observed toxicity in Lin⁻ cells (50 to 60% viability) following foamy virus vector transductions even at ≤1% DMSO (final concentration) during transduction. At day 4 after the final transduction, transgene expression from the transduced bulk cultures was analyzed using flow cytometry. In bulk cultures with lower gene transfer efficiency, GFP⁺ cells were sorted using a BD FACS Aria II (BD Biosciences) and expanded until plated. Bulk cultures with higher gene transfer efficiency were also sorted for GFP⁺ cells, and the replating frequencies of the sorted pools and unsorted transduced pools were compared. During expansion, the transduced bulk cultures were maintained at a concentration of 2 × 10⁵ to 5 × 10⁵ cells/ml. After expansion, the cells were plated in 96-well plates at a density of 100 cells/well. After 2 weeks of plating, the 96-well plates were examined and scored for the presence of wells with proliferating cell populations. Under these conditions, the mock-transduced cells/untransduced cells barely survived. The positive wells were further expanded for molecular analysis. At 5 weeks, some of the clones expanded at 2 weeks had terminally differentiated and died by 5 weeks. The replating frequency of each vector tested at 2 weeks and 5 weeks was calculated based on Poisson statistics using L-Calc software (Stem Cell Technologies).

Immortalized clones derived from SFFV-GV could be replated at the same frequency at 2 weeks and 5 weeks, allowing comparison of the genotoxicities of our vectors relative to the highly genotoxic SFFV-GV vector. In contrast, clones derived from vectors with low genotoxic potential showed initial growth and replating potential at 2 weeks but terminally differentiated thereafter and lost their replating frequency by 5 weeks. To be able to compare immortalization frequencies/VCN of sorted and unsorted populations, a portion of the transduced bulk cultures from SFFV-GV-, MSCV-GV-, SFFV-LV-, MSCV-LV-, SFFV-FV-, and MSCV-FV-transduced Lin⁻ cells were sorted for GFP expression and showed proportional immortalization before and after sorting, validating this modification to give a similar immortalization readout. Immortalization frequencies/VCN before sorting of SFFV-GV-, SFFV-LV-, and SFFV-FV-transduced cells were 0.001755, 0.000176, and 0.000006, respectively. After sorting, the frequencies were 0.002016, 0.000128, and 0.000023, respectively. For the vector-transduced group negative for replating clones, calculations were based on the assumption that a replating clone would be detected if 97 wells were plated instead of 96 wells (9).

Phenotypic analysis of immortalized clones.

Immortalized clones were labeled with antibodies that recognize the cell surface markers Sca-1 phycoerythrin (PE) (clone D7; catalog number 553108) and c-Kit allophycocyanin (APC) (clone 2B8; catalog number 553356) from BD Biosciences and analyzed using a FACS Canto (BD Biosciences).

Vector copy number analysis.

Quantitative real-time PCR was performed to assess the gene transfer efficiencies of GV, LV, and FV vector-transduced bulk cultures. For GV vectors, LV vectors, and FV vectors, primers that recognize the wPRE region were used to measure the VCN in bulk cultures. Genomic DNA from a single-copy NIH 3T3 cell clone carrying a single copy of the MM13 vector was used as a standard for copy number analysis. The MM13 plasmid has been described previously (61). Primers in the FV backbone were used to measure the copy number, as well (FV backbone forward primer, 5′-AATCCTTTACATGGAGAAGTTATAGGTCTT-3′, and reverse primer, 5′-TGGCCAAATCCATAGCCTTAGA-3′). PCR was carried out with TaqMan probe (5′-ATCTGAAATCTCTCAATTTGTCCCCACCA-3′) with tetramethyl-6-carboxyrhodamine dye as a quencher. The FV- or wPRE-specific signal was normalized to mouse ApoB in each sample. Genomic DNA (50 ng) from a single-copy murine erythroleukemia (MEL) cell clone transduced with FV was diluted with untransduced MEL DNA to generate copy number standards. Quantitative PCR was performed with an Applied Biosystems 7900HT real-time PCR system (Thermo Fisher, Grand Island, NY) using a thermocycler protocol for 96-well plates, according to the manufacturer's instructions.

LAM PCR to determine insertion sites in immortalized clones.

For ligation amplification-mediated (LAM) PCR, the junction sequences between the viral LTR and the mouse genome were linearly amplified twice with 100 ng of genomic DNA from FV immortalized clones using 0.25 pmol of the FV-specific 5′-end-biotinylated primer (5′-GAACCTTGTGTCTCTCATCCC-3′) and 2.5 units of Taq polymerase (Qiagen, Hilden, Germany), with cycling conditions as follows: initial denaturation at 95°C for 3 min, 50 cycles of amplification (95°C for 30 s, 55°C for 30 s, and 72°C for 1 min), and a final extension at 72°C for 3 min. After DNA enrichment of the biotinylated DNA, hexanucleotide primer extension was carried out using Klenow enzyme (Promega, Madison, WI), and the primer-extended product was digested with TasI (New England BioLabs [NEB], Ipswich, MA). Following TasI digestion, the DNA was ligated to TasI-specific double-stranded linkers (5′-GACCCGGGAGATCTGAATTCAGTGGCACAGCAGTTAGG-3′/5′-AATTCCTAACTGCTGTGCCACTGAATTCAGATC-3′). The first exponential amplification of the linked products was performed using 12.5 pmol each of FV-specific primer (5′-GTCTATGAGGAGCAGGAGTA-3′) and the linker cassette-specific primer (5′-GACCCGGGAGATCTGAATTC-3′). Eight percent of the first exponential PCR was then used as the template for a second exponential nested-PCR amplification using 12.5 pmol each of nested FV-specific primer (5′-CCTCCTTCCCTGTAATACTC-3′) and nested linker cassette-specific primer (5′-AGTGGCACAGCAGTTAGG-3′) under the same conditions as the first PCR. To detect the insertion sites from the MSCV-GV-immortalized clones, 100 ng of genomic DNA was linearly amplified using MSCV LTR-specific 5′-end-biotinylated primer LTR1 (5′-CTGGGGACCATCTGTTCTTGGCCCT-3′), enriched with Dynabeads M-280 streptavidin (Thermo Fisher), digested with Tsp5091 (NEB), and linked to an asymmetric linker cassette (5′-AATTCTCTAGTATGCTACTCGCACCGATTATCTCCGCTGTCAGT-3′ and 5′-ACTGACAGCGGAGATAATCGGTGCGAGTAGCATACTAGAG-3′). The ligation products were then amplified with LTR- and linker-specific primer LTR2 (5′-GACTTGTGGTCTCGCTGTTCCTTGG-3′) and a linker cassette primer, LC1 (5′-ACTGACAGCGGAGATAATCG-3′) (first exponential PCR). The second exponential PCR was carried out with primers LTR3 (5′-GGTCTCCTCTGAGTGATTGACTACC-3′) and LC2 (5′-GTGCGAGTAGCATACTAGAG-3′) (62).

Next-generation sequencing of LAM PCR products.

The products from the second exponential PCR were processed for next-generation DNA sequencing at the Cincinnati Children's Hospital Medical Center (CCHMC) DNA Sequencing Core. LAM second exponential PCR products were purified using a QIAquick PCR purification kit (Qiagen) and then rendered blunt ended by end repair with T4 DNA polymerase, Klenow enzyme, and T4 PNK (Promega) in the presence of 10 mM deoxynucleoside triphosphates (dNTPs) (Thermo Fisher). The blunt-ended products were randomly concatenated by treatment with T4 Quick Ligase (NEB) at room temperature for 15 min. Next-generation sequencing libraries compatible with the Illumina system were prepared using a Nextera in vitro transposition kit (Epicentre, Madison, WI) according to the manufacturer's recommendations and amplified using a different molecular barcode for each sample. After another round of PCR purification, all 10 libraries were quality checked on an Agilent Bioanalyzer (Agilent, Santa Clara, CA) and then mixed in equal amounts in a single pool. Sequencing was conducted on an Illumina HiSeq2000 (Illumina, San Diego, CA) in single-read mode with indexing, producing 100-base-long sequences.

After demultiplexing of all the sequences in the pool and assignment to their respective samples, reads were processed and aligned to the mm9 mouse reference assembly using the CASAVA 1.8 package. The results were generated in the QSEQ SORTED file format so that alignments could be visualized using the ChIP sequencing (ChIP-seq) module of Illumina's Genome Studio software. Insertion sites detected by LAM PCR are characterized by an LTR sequence upstream of the insertion and an adapter sequence downstream. While the aligner was configured to position reads that contained only mouse genome sequence, reads that contained some LTR or some adapter sequence along with a majority of mouse sequence were also positioned. By zooming in to the base level display in Genome Studio (Illumina), it was possible to determine the edge of the covered regions and sides that matched the adapter sequence and the sides that matched the LTR sequence, allowing the determination of the insertion point and the direction in which the provirus integrated. All identified insertions were compared to the National Center for Biotechnology Information (NCBI) mouse build 37 genome database (http://www.ncbi.nlm.nih.gov).

CRISPR/Cas9 insertion of proviral sequences. (i) gRNA development.

The reference sequence used for the initial description (41) of the LMO2 integration site (Homo sapiens chromosome 11 [chr11] clone RP1-22J9 map of p12-14.1; GenBank accession number AL135799.8) was obtained from NCBI. It corresponds to GRCh38.p2 chr11:33890271. Genomic DNA was isolated from Jurkat cells, and the region around the insertion site was PCR amplified using Q5 polymerase (NEB) and sequenced by the CCHMC DNA Sequencing and Genotyping Core. The PCR primers were LMO2 FWD PCR (5′-TTTAGGTTGCCCTGAAAAGGTG-3′) and LMO2 REV PCR (5′-GCCAAACACTCCTAGGCTCTTG-3′). The sequencing primers were LMO2 FWD PCR, LMO2 REV PCR, and LMO2 seq1 (5′-GTCTCTCGCAGCCACATGGG-3′). The region around the insertion site was analyzed for potential gRNA target sites using the CRISPR design program (Benchling, Inc., San Francisco, CA). Five gRNAs were chosen on the basis of proximity to the planned insertion site and low predicted off-target effects. A plasmid containing both a gRNA and a Cas9-T2A-eGFP expression cassette (pX458m) was a kind gift from Yueh-Chiang Hu, Transgenic Animal and Genome Editing Core at CCHMC. eGFP cDNA was first replaced with an mCherry reporter (pX458m-mCherry). Site-directed mutagenesis was performed using a QuikChange II XL site-directed mutagenesis kit (Agilent) to remove a BbsI site within the mCherry sequence (primers 5′-CCCGTAATGCAGAAGAAAACCATGGGCTGGGAGGC-3′ and 5′-GCCTCCCAGCCCATGGTTTTCTTCTGCATTACGGG-3′). DNA oligonucleotides for cloning the target sequences into the pX458m-mCherry vector were designed and obtained from Integrated DNA Technologies (IDT) (Coralville, IA). The oligonucleotides used to generate gRNA 1 with targeting sequence GATACCAATAGATATCAATC were LMO2 gRNA 1 FWD (5′-CACCGGGATACCAATAGATATCAATC-3′) and LMO2 gRNA 1 REV (5′-AAACGATTGATATCTATTGGTATCCC-3′). The oligonucleotides used to generate gRNA 2 with targeting sequence ATCACCAGATTGATATCTAT were LMO2 gRNA 2 FWD (5′-CACCGGGATCACCAGATTGATATCTAT-3′) and LMO2 gRNA 2 REV (5′-AAACATAGATATCAATCTGGTGATCCC-3′). The oligonucleotides used to generate gRNA 3 with targeting sequence AATTGCATAGTCGTGAAGTC were LMO2 gRNA 3 FWD (5′-CACCGGGAATTGCATAGTCGTGAAGTC-3′) and LMO2 gRNA 3 REV (5′-AAACGACTTCACGACTATGCAATTCCC-3′). The oligonucleotides used to generate gRNA 4 with targeting sequence ATTGCATAGTCGTGAAGTCA were LMO2 gRNA 4 FWD (5′-CACCGGGATTGCATAGTCGTGAAGTCA-3′) and LMO2 gRNA 4 REV (5′-AAACTGACTTCACGACTATGCAATCCC-3′). The oligonucleotides used to generate gRNA 5 with targeting sequence TCGTGAAGTCAGGGCTTCTA were LMO2 gRNA 5 FWD (5′-CACCGGGTCGTGAAGTCAGGGCTTCTA-3′) and LMO2 gRNA 5 REV (5′-AAACTAGAAGCCCTGACTTCACGACCC-3′).

pX458m-mCherry was digested with FastDigest BbsI (Thermo Fisher) and simultaneously dephosphorylated with FastAP (Thermo Fisher). The digested product was then gel purified. Oligonucleotide pairs were phosphorylated and annealed in a reaction mixture of 100 μM each oligonucleotide and T4 polynucleotide kinase (NEB), placed in a Veriti 96-well fast thermal cycler (Thermo Fisher) at 37°C for 30 min and 95°C for 5 min, and then ramped down to 25°C at 5°C/min. The annealed oligonucleotides were then ligated into the cleaved pX458m plasmid and transformed into PX5-α competent cells (Protein Express, Cincinnati, OH). A plasmid was subsequently prepared using an EndoFree Plasmid Maxi kit (Qiagen); 2 μg of each gRNA/Cas9 plasmid was transfected into 2.4 × 10⁵ Jurkat cells in a 24-well plate using Lipofectamine 3000 (Thermo Fisher) according to the manufacturers' protocol. At day 7, cells were harvested and genomic DNA was purified. The area around the target site was amplified using Q5 polymerase and sequenced (PCR primers, LMO2 FWD v3, 5′-GCTTGGGTTTTACACGTCTTC-3′, and LMO2 REV v3, 5′-TCAGCTAGAAAACAAGTACTTGC-3′; sequencing primer, LMO2 seq1, 5′-GTCTCTCGCAGCCACATGGG-3′). The gRNA efficiency was determined using the tracking of indels by decomposition (TIDE) assay (63).

(ii) Donor vector templates for homology-directed repair (HDR).

After sequencing the LMO2 region in Jurkat cells, ∼600-bp homology arms were designed with a multiple-cloning site region at the chosen insertion site. The homology vector was ordered as a plasmid in the pUC57 backbone from GenScript. The foamy virus pΔΦ.SF.eGFP.PRE was cut at the LTRs with XbaI and EcoNI and inserted between NheI and EcoNI sites. The lentivirus pRRL.PPT.SF.eGFP.PRE was cut at the LTRs with BsaI and PsiI and inserted between BbsI and NaeI sites. Retrovirus pRSF91.eGFP.PRE was cut at the LTRs between XhoI and HindIII and inserted between BsmFI and XhoI sites. The resulting clones were checked by restriction digestion and sequenced for verification. For GV, the entire LTR sequences, which contain SFFV enhancers/promoters, were contained in the cloned sequence. To facilitate cloning of LV and FV, sequences from the R region of the 5′ LTR through the entire 3′ LTR were cloned from viral production plasmids. For the LV vector, the 40 bp of the 5′ ΔU3 region (left after deletion of the U3 enhancer/promoter in the LTR) were not part of the cloned “proviral” construct. This small region does not have enhancer/promoter activity and therefore was inconsequential for the purpose of studying genotoxicity. Similarly, in the case of the FV vector, the omitted 5′ U3 region contains a 582-bp deletion that removes the U3 TATA box and transcriptional enhancer sites of the LTR, leading to silencing of the LTR (64). Here, we use the term “provirus” for these LV and FV constructs.

After identifying the 36-bp insulator region in the FV LTR, the LMO2 donor containing pΔΦ.SF.eGFP.PRE was modified at the 5′ LTR and 3′ LTR to remove the identified CTCF binding sites. The initial sequence of this region was AGT AAA AGG ATT TGT ATA TTA GCC TTG CTA AGG GAG ACA TCT AGT GAT ATA AGT GTG AAC TAC ACT TAT CTT AAA TGA TG to AGT AAA AGG ATT TGT ATA TTA GCC TTG CTA AGC ACA TTC GAT AGT GAT ATA AGA GGC TTT ATA TCT TAT CTT AAA TGA TG (the insulator sequence is underlined). For the 3′ LTR of the proviral sequence, a gene block containing the modified insulator sequence was ordered from IDT to replace the ∼800-bp region between EcoNI and MluI sites. For the 5′ LTR, a gene synthesis product was ordered from GenScript to replace the ∼550-bp region between PacI and AvrII sites. The resulting plasmid was confirmed by sequencing.

To add the 36-bp insulator sequence to the lentiviral LTRs, the following sequence was added ahead of the R regions of the LTRs in the LMO2 donor containing pRRL.PPT.SF.eGFP.PRE: AAG GGA GAC ATC TAG TGA TAT AAG TGT GAA CTA CAC. Two gene blocks were ordered from IDT to replace the ∼1-kb region between BsiWI and MluI sites encompassing the 3′ LTR and the ∼900-bp region between BspEI and MfeI sites encompassing the 5′ LTR. The resulting plasmid was confirmed by sequencing.

(iii) Generation of HeLa clones.

On day −1, 5 × 10⁴ cells were seeded into a 24-well plate. The cells were transfected with the LMO2 gRNA 5 plasmid, as well as with one of the three (GV, LV, or FV) LMO2 donor plasmids. Five hundred nanograms of total DNA was transfected, divided at an approximate molar ratio of 1:2 (LMO2 gRNA 5 plasmid to donor plasmid). Transfection was performed using 1.5 μl of Lipofectamine 3000 (Thermo Fisher) according to the manufacturer's recommendations. Successful transfection was verified on day 2 by analyzing a portion of the cells for expression of both eGFP (donor plasmid) and mCherry (LMO2 gRNA 5 plasmid) using a FACS Canto (BD Biosciences). At 2 weeks, the cells were reanalyzed for eGFP and mCherry by FACS. GFP-positive and mCherry-negative cells were sorted as single cells into a 96-well plate by the Research Flow Cytometry Core at CCHMC using a BD FACS Aria II.

After reaching at least 80% confluence, a portion of the cells were harvested. DNA was purified by resuspending the cell pellet in 20 μl of QuickExtract DNA extraction solution (Epicentre) and incubated at 65°C for 15 min, at 68°C for 15 min, and at 98°C for 10 min. The purified DNA was then screened for correct integration of the donor sequence by PCR using primer sets flanking the homology arms. The first PCR for ensuring correct 5′ homology used primers LMO2 FWD v3 and Viral FWD (5′-CGAGCGTTGGTAAGAGAAGC-3′). The second PCR for ensuring correct 3′ homology used primers LMO2 REV v3 and either Viral REV1 (5′-GAGATCTGTCCCGCTAGCA-3′) for GV and LV or Viral REV2 (5′-GGATAATTTACAAATAAACCCGACTTATATTCG-3′) for FV. Clones that had correct bands for both PCRs were considered to have correctly inserted viral sequences by HDR. The steps to estimate targeted-allele copy numbers were as follows. (i) HeLa cells have been reported to contain 3 or 4 copies of chromosome 11p, on which the LMO2 gene resides. We first confirmed by fluorescent in situ hybridization that our unedited/WT HeLa cells had 4 LMO2 alleles (Fig. 9A). (ii) We determined the number of edited nontargeted or WT LMO2 alleles, using primers in the LMO2 gene that flank the Cas9 DSB/proviral insertion site, so that only WT/edited nontargeted LMO2 would be detected and LMO2 loci containing a provirus sequence insertion would not be amplified. WT/nontargeted LMO2 copy number analysis showed that all evaluable clones had 1 or 2 WT/nontargeted LMO2 copies (Fig. 9B). (iii) Clones were further interrogated by PCR amplifying across the gRNA target site, followed by sequencing (the PCR/sequencing primers were the same as those used for the TIDE assay) to assess for the presence of large deletions and indels in each clone (Fig. 9C).

LMO2 expression analysis by qRT-PCR.

RNA was prepared by lysing cells in RNA Stat-60 (AMS Biotechnology, Abingdon, United Kingdom) and passing them over a QIAshredder column (Qiagen). RNA was isolated by chloroform phase separation. The aqueous layer was precipitated with isopropanol, and the resulting pellet was washed with 75% ethanol. The RNA pellets were resuspended in nuclease-free molecular-grade water and dissolved by incubation at 55°C. The RNA was quantified using a NanoDrop 1000 spectrophotometer (Thermo Fisher). cDNA was prepared with purified RNA using a high-capacity cDNA reverse transcription kit (Thermo Fisher). We determined LMO2 mRNA expression in HeLa clones by qRT-PCR with two probe and primer sets, LMO2 TaqMan genomic assays Hs00277106 and Hs00153473 (Thermo Fisher), corresponding to two different regions in LMO2 cDNA, and using two different validated loading controls for HeLa cells, human PPIA and GAPDH (glyceraldehyde-3-phosphate dehydrogenase) (Thermo Fisher) (4). The LMO2 regions amplified are found in all reported transcript variants. Both probe-primer sets bridge exons, so only mRNA is amplified. For LMO2 transcript variant 1 (NM_005574.3), the probe-primer sets bridge exons 4 and 5 and exons 5 and 6, respectively. qRT-PCR mixtures were prepared with iTaq Universal Probes Supermix (Bio-Rad, Hercules, CA). The ABI 7900HT real-time PCR system was used to run the qRT-PCR. Data from the RT-PCR was analyzed by relative quantification by the 2^−ΔΔCT method (65) using human GAPDH/PPIA to normalize the results and to determine fold induction. Unedited HeLa samples were used as calibrators.

WT LMO2 copy number analysis.

Genomic DNA was isolated from HeLa clones. Reactions were prepared with iTaq Universal Probes Supermix (Bio-Rad). The human ApoB gene was used as an endogenous control gene. The CFX Connect real-time PCR detection system (Bio-Rad) was used to run the qRT-PCR. Data from the RT-PCR were analyzed by relative quantification by the 2^−ΔΔCT method (65) using a K562 cell line and unedited HeLa cells to normalize the results and to determine copy numbers. The primer and probe sets were as follows: LMO2-CN-FW (5′-TGGGGAACAAGTACAATTTTGTG-3′), LMO2-CN-RV2 (5′-CAATGTGGTGATATCAATCTGGTG-3′), LMO2-CN-Probe (5′-ACAAGCGTAAATTGCATAGTCGTGA-3′), hApoB-CN-FW (5′-CTTGGTTTATGAATCTGGCTC-3′), hApoB-CN-RV (5′-GCCTTTAGCAGTTAGAACAC-3′), and hApoB-CN-Probe (5′-ACATGCTGGGAATCGACTTGTGAT-3′).

Western blot analysis of LMO2 protein expression.

Cells were lysed in RIPA lysis buffer (Santa Cruz, Dallas, TX), and 20 μl of protein lysate (containing between 30 and 87 μg protein per sample) was separated on a 4 to 15% Mini-Protean TGX precast protein gel (Bio-Rad) and transferred onto a nitrocellulose polyvinylidene difluoride (PVDF) membrane (Bio-Rad). The membrane was blocked with Odyssey blocking buffer (Li-Cor Biosciences, Lincoln, NE) and probed with a primary human LMO2 antibody and secondary anti-goat IRDye 800CW antibody (Li-Cor Biosciences). The membrane was then reprobed with a primary human GAPDH antibody (Fitzgerald, Acton, MA) and secondary anti-mouse IRDye 680LT (Li-Cor Biosciences). Signals were visualized using an Odyssey 9120 infrared imager (Li-Cor Biosciences).

Electrophoretic mobility shift assay.

Oligonucleotides corresponding to predicted CTCF binding sites in the LV and FV proviral sequences were designed. Oligonucleotides labeled with IRDye 700 at the 5′ end were ordered from IDT. LV1 probe, made by annealing oligonucleotides 5′-ACAAGATAGAGGAAGAGCAAAACAAAAGTAAGACCACCGCACAGCAAGCGGCCGCTGATCTTCAGACCTGGAGGAGGAGATATGAGGGA-3′ and 5′-TCCCTCATATCTCCTCCTCCAGGTCTGAAGATCAGCGGCCGCTTGCTGTGCGGTGGTCTTACTTTTGTTTTGCTCTTCCTCTATCTTGT-3′, contained the GGAAGAGCA and CTCCTCCTCCAGGT sequence motifs. LV2 probe, made by annealing oligonucleotides 5′-GATACCTAAAGGATCAACAGCTCCTGGGGATTTGGGGTTGCTCTGGAAAACTCATTTGCACCACTGCTGTGCCTTGGAATGCT-3′ and 5′-AGCATTCCAAGGCACAGCAGTGGTGCAAATGAGTTTTCCAGAGCAACCCCAAATCCCCAGGAGCTGTTGATCCTTTAGGTATC-3′, contained the TCCCCAGGAGCTGTTGATCC and GGCACAGCA sequence motifs. LV3 probe, made by annealing oligonucleotides 5′-GTCGGGGAAGCTGACGTCCTTTCGAATTCGATATCAAGCTGTACCTTTAAGACCAATGACTTACAAGGCAGCTGTAGATC-3′ and 5′-GATCTACAGCTGCCTTGTAAGTCATTGGTCTTAAAGGTACAGCTTGATATCGAATTCGAAAGGACGTCAGCTTCCCCGAC-3′, contained the GGTACAGCT sequence motif. FV1 probe, made by annealing oligonucleotides 5′-TCCATTAACACTCTGCTTATAGATTGTAAGGGTGATTGCAATGCTTTCTGCATAAAACTTTGGTTTTCTTGTTAATCAAT-3′ and 5′-ATTGATTAACAAGAAAACCAAAGTTTTATGCAGAAAGCATTGCAATCACCCTTACAATCTATAAGCAGAGTGTTAATGGA-3′, contained the AGCATTGCA sequence motif. FV2 probe, made by annealing oligonucleotides 5′-AGTAAAAGGATTTGTATATTAGCCTTGCTAAGGGAGACATCTAGTGATATAAGTGTGAACTACACTTATCTTAAATGATG-3′ and 5′-CATCATTTAAGATAAGTGTAGTTCACACTTATATCACTAGATGTCTCCCTTAGCAAGGCTAATATACAAATCCTTTTACT-3′, contained the ATATCACTAGATGTCTCCCT and overlapping sequence motifs. FV3 probe, made by annealing oligonucleotides 5′-TCGGGTTTATTTGTAAATTATCCCTAGGGACCTCCGAGCATAGCGGGAGGCATATAAAAGCCAATAGACAATGGCTAGCA-3′ and 5′-TGCTAGCCATTGTCTATTGGCTTTTATATGCCTCCCGCTATGCTCGGAGGTCCCTAGGGATAATTTACAAATAAACCCGA-3′, contained the AGCATAGCG sequence motif. FV4 probe, made by annealing oligonucleotides 5′-GGCATCAGCCTACAAATACCAGTATTCATACTGAAGGCAATGCCCTAGCAGATAAGCTTGCCACCCAAGGAAGTTATGTA-3′ and 5′-TACATAACTTCCTTGGGTGGCAAGCTTATCTGCTAGGGCATTGCCTTCAGTATGAATACTGGTATTTGTAGGCTGATGCC-3′, contained the GGCATTGCC sequence motif. FV5 probe, made by annealing oligonucleotides 5′-CGCAACTGTTAAATCTCTCAATGTACTCACTAGTATTGCAATTCCAAAGGTGATTCACTCTGATCAAGGTGCAGCATTCA-3′ and 5′-TGAATGCTGCACCTTGATCAGAGTGAATCACCTTTGGAATTGCAATACTAGTGAGTACATTGAGAGATTTAACAGTTGCG-3′, contained the GGAATTGCA sequence motif. FV6 probe, made by annealing oligonucleotides 5′-CTCGTTCCTGGTCTCCTGTTGTTGGCCAATTGGTCCAGGAGAGGGTGGCTAGGCCTGCTTCTTTGAGACCTCGTTGGCAT-3′ and 5′-ATGCCAACGAGGTCTCAAAGAAGCAGGCCTAGCCACCCTCTCCTGGACCAATTGGCCAACAACAGGAGACCAGGAACGAG-3′, contained the TGGTCCAGGAGAGGGTGGCT and overlapping sequence motifs. FV7 probe, made by annealing oligonucleotides 5′-ATGAGGCACTTCAGAATACAACAACTGTGACTGAACAGCAGAAGGAACAAATTATACTGGACATTCAAAATGAAGAAGTA-3′ and 5′-TACTTCTTCATTTTGAATGTCCAGTATAATTTGTTCCTTCTGCTGTTCAGTCACAGTTGTTGTATTCTGAAGTGCCTCAT-3′, contained the TGAACAGCAGAAGGAACAAA and overlapping sequence motifs. FV8 probe, made by annealing oligonucleotides 5′-TATGGAAGCTTATGGACCTCAGAGAGGAAGTAACGAGGAGAGGGTGTGGTGGAATGCCACTAGAAACCAGGGAAAACAAG-3′ and 5′-CTTGTTTTCCCTGGTTTCTAGTGGCATTCCACCACACCCTCTCCTCGTTACTTCCTCTCTGAGGTCCATAAGCTTCCATA-3′, contained TAACGAGGAGAGGGTGTGGT, GGCATTCCA, and overlapping sequence motifs. Similarly, FV2 mutant 1 probe was made by annealing oligonucleotides 5′- AGTAAAAGGATTTGTATATTAGCCTTGCTAAGGGAGACATCTAGTGATATAAGaggctttatatcTTATCTTAAATGATG-3′ and 5′- CATCATTTAAGATAAgatataaagcctCTTATATCACTAGATGTCTCCCTTAGCAAGGCTAATATACAAATCCTTTTACT-3′. FV2 mutant 2 probe was made by annealing oligonucleotides 5′- AGTAAAAGGATTTGTATATTAGCCTTGCTAAGGGAGACATCaggctttatatcTGTGAACTACACTTATCTTAAATGATG-3′ and 5′- CATCATTTAAGATAAGTGTAGTTCACAgatataaagcctGATGTCTCCCTTAGCAAGGCTAATATACAAATCCTTTTACT-3′. FV2 mutant 3 probe was made by annealing oligonucleotides 5′- AGTAAAAGGATTTGTATATTAGCCTTGCTaggctttatatcTAGTGATATAAGTGTGAACTACACTTATCTTAAATGATG-3′ and 5′- CATCATTTAAGATAAGTGTAGTTCACACTTATATCACTAgatataaagcctAGCAAGGCTAATATACAAATCCTTTTACT-3′. FV2 mutant 4 probe was made by annealing oligonucleotides 5′- AGTAAAAGGATTTGTATaggctttatatcAAGGGAGACATCTAGTGATATAAGTGTGAACTACACTTATCTTAAATGATG-3′ and 5′- CATCATTTAAGATAAGTGTAGTTCACACTTATATCACTAGATGTCTCCCTTgatataaagcctATACAAATCCTTTTACT-3′. FV2 mutant 5 probe was made by annealing oligonucleotides 5′- AGTAAAAGGATTTGTATATTAGCCTTGCTaagcacattcgaTAGTGATATAAGaggctttatatcTTATCTTAAATGATG-3′ and 5′- CATCATTTAAGATAAgatataaagcctCTTATATCACTAtcgaatgtgcttAGCAAGGCTAATATACAAATCCTTTTACT-3′. FV2 mutant 6 probe was made by annealing oligonucleotides 5′- AGTAAAAGGATTTGTATATTAGCCTTGCTaagcacattcgaaggctttatatcTGTGAACTACACTTATCTTAAATGATG-3′ and 5′- CATCATTTAAGATAAGTGTAGTTCACAgatataaagccttcgaatgtgcttAGCAAGGCTAATATACAAATCCTTTTACT-3′. Lowercase letters represent sequences divergent from the original FV2 probe.

Of note, one LV predicted binding site was not interrogated. A probe corresponding to the H19–Igf2 locus (H19), previously shown to bind CTCF, was used as a positive control (66). Unlabeled H19 probe was used for competition assays. Oligonucleotides were annealed in duplex buffer (IDT). Purified full-length human recombinant CTCF protein (Abnova, Taipei City, Taiwan) and labeled oligonucleotides were incubated at room temperature for 30 min in 20 mM HEPES (pH 7.5), 50 mM KCl, 5 mM MgCl₂, 3.3 μM ZnSO₄, 1 mM dithiothreitol, 0.3 mg/ml bovine serum albumin (BSA), 0.5 μg poly(dI-dC), 5% glycerol, and 0.5% Triton X-100 (67). Binding reaction mixtures were then resolved on 6% Novex Tris-borate-EDTA (TBE) gels (Thermo Fisher) using 0.5× TBE running buffer (Thermo Fisher). The gels were imaged using an Odyssey 9120 infrared imager.

Chromatin immunoprecipitation.

Briefly, HeLa cell clones (1 × 10⁷ to 2 × 10⁷ cells) from transfected FV and LV and untransfected control HeLa cells were treated with formaldehyde (1% final concentration) and incubated at 37°C for 10 min to cross-link histones to DNA. The formaldehyde was neutralized with 2.5 M glycine (final concentration, 0.25 M) for 5 to 10 min at room temperature and centrifuged for 5 min at 2,000 rpm. The cells were pelleted and stored at −80°C. For ChIP, cells were thawed and the pellet was resuspended in 200 μl of SDS lysis buffer (Millipore, Billerica, MA) and incubated on ice for 10 min, and protease inhibitors (Pierce protease inhibitor; Thermo Fisher) were added (1 mM phenylmethylsulfonyl fluoride [PMSF], 1 μg/ml aprotinin, and 1 μg/ml pepstatin A) to the cell lysate. The lysate was sheared under optimized conditions (in a Covaris TM S220 [Covaris, Woburn, MA] for 70 s at 4°C [peak power, 105; duty factor, 10; and cycles/burst, 200]) to generate cross-linked DNA fragments 200 to 1,000 bp in length. The sonicated samples were centrifuged for 10 min at 13,000 rpm at 4°C. The supernatant was diluted 10-fold with ChIP dilution buffer (Millipore), and protease inhibitors were added as described above. A portion (1%) of the sample was retained as the input sample. The diluted samples were then precleared with 75 μl of salmon sperm DNA-protein A agarose, 50% slurry (Millipore) for 30 min at 4°C with agitation. The agarose was pelleted by brief centrifugation, and 2 μl anti-CTCF antibody (Millipore) was added to 2 ml of precleared supernatant and incubated overnight at 4°C with constant rotation. The next day, 60 μl of salmon sperm DNA-protein A agarose, 50% slurry was added for 1 h at 4°C with rotation to collect the antibody-histone complex. The agarose was pelleted by gentle centrifugation at 1,000 × g for 1 min, and the supernatant with unbound chromatin was discarded. The protein A agarose-antibody-chromatin complex was washed sequentially with 1 ml each of low-salt immune complex (Millipore), high-salt immune complex (Millipore), and LiCl immune complex (Millipore) and twice with Tris-EDTA (TE) buffer (Millipore) for 5 min at 4°C with rotation. The TE wash buffer was removed, and the protein A agarose-antibody-chromatin complex was resuspended in 250 μl of fresh elution buffer (1% SDS, 0.1 M NaHCO₃) and incubated at room temperature for 15 min with rotation. The agarose beads were spun down. The process was repeated twice, and the eluates were combined. The cross-links were reversed by adding 20 μl of 5 M NaCl (Millipore) to the combined eluates and heating at 65°C for 4 h, followed by the addition of 10 μl of 0.5 M EDTA (Millipore), 20 μl of 1 M Tris-HCl, pH 6.5 (Millipore), and 2 μl of 10-mg/ml proteinase K to the eluates. This mixture was incubated for 1 h at 45°C. DNA was recovered from the eluate by using a PCR clean-up kit (Qiagen). Following ChIP purification, the eluted products were analyzed by qualitative PCR. The following primers, corresponding to predicted CTCF binding sites, were utilized: FV1, 5′-CGAGACTCTCCAGGTTTGGTAA-3′ and 5′-GGTTCTCGAATCAAGTCGGTTT-3′; FV2, 5′-AACCGACTTGATTCGAGAACCT-3′ and 5′-GTTGGGCGCCAATTGTCAT-3′; FV5, 5′-ACTAAGGCTCCTTCTACTAGCG-3′ and 5′-GTTGAAGAAGTGAATGCTGCAC-3′; FV6, 5′-TTATACCATCCATCCACCCCTC-3′ and 5′-GTTTATGCCAACGAGGTCTCAA-3′; and FV7, 5′-GCATGAGGCACTTCAGAATACA-3′ and 5′-AGGCCAATACTCTTGAGCTAGT-3′. The H19 locus was again used as a positive control, using primers 5′-CCCATCTTGCTGACCTCAC-3′ and 5′-AGACCTGGGACGTTTCTGTG-3′. The sizes of the amplicons for H19, FV1, FV2, FV5, FV6, and FV7 were 165, 157, 188, 110, 115, and 155 base pairs, respectively. ChIP input for HeLa control cells and a HeLa FV clone (FVA2), as well as the ChIP product for the HeLa FV clone, were assayed.

Statistical analysis.

Two-tailed Student unpaired t tests, using GraphPad software (GraphPad Software, Inc., La Jolla, CA), were used to calculate the statistical differences between groups in the immortalization assay. The immortalization frequencies of the SFFV and MSCV series of FV and LV vectors were compared to the replating frequency of SFFV-GV. Data are presented as means ± standard errors of the mean (SEM), and differences with a P value of <0.05 were considered statistically significant. RT-PCR data were analyzed using a one-tailed Mann-Whitney U test.