Transposon-mediated BAC transgenesis in human ES cells

Generation of large reporter constructs and BAC reporters

The large constructs were made by subcloning from the respective BACs a region of 19 kb for hOCT4 gene and 25 kb for hNANOG into a plasmid with p15A origin of replication using recombineering technology (Supplementary Figure S1) (2,3). For the generation of large construct or BAC reporters, the green fluorescent protein (GFP) or Cherry cassettes were inserted directly after the initiating methionine (ATG) of the respective gene using recombineering. The PiggyBac or SB terminal repeats were inserted into different positions of the BAC backbone using a universal recombineering strategy applicable to most of the common used BAC vectors (Supplementary Figure S2). The recombineering details and list of oligos are presented in Supplementary Experimental Procedures.

hESC culturing

H7.S6 and H9 hESCs were cultured on mouse embryonic fibroblasts (MEFs) in DMEM/F12 medium supplemented with 20% Knockout Serum Replacement (Invitrogen) and 4 ng/ml basic fibroblast growth factor (bFGF) (Peprotech) and passaged using 1 mg/ml collagenase IV (Invitrogen) adding 10 µM Rho-associated kinase (ROCK) inhibitor Y-27632 (40). For transfections and differentiation assays, the cells were transferred to feeder-free conditions on Matrigel (BD Biosciences) in MEF-conditioned hESC medium, and propagated using TrypLE (Invitrogen).

Transfections of hESCs

Electroporation of large constructs into hESCs was performed according to the standard protocol at 320 V and 250 µF (15). BAC transfection was performed either by nucleofection (20) or lipofection. hOCT4-GFP, hNANOG-GFP, hPAX6-GFP and hGATA4-GFP BACs were prepared using Nucleobond BAC 100 kit (Macherey-Nagel). Nucleofection was done in 100 µl of solution V using program B-016 according to manufacturer protocol (Amaxa). 5 × 10⁶ of cells were nucleofected with 5 µg of the BAC and 300 ng of the transposase expression or control vector.

For lipofection, hESCs were split to Matrigel-coated dishes in the ratio 1:3, 1 day before transfection. 3, 10, 30 or 50 µg of BAC and 3 or 10 µg of the transposase expression or control vector were used for lipofection of a 10 cm dish with hESCs using Lipofectamine LTX (Invitrogen) according to manufacturer protocol.

Selection with G418 (100 µg/ml; Invitrogen), puromycin (0.5 µg/ml; Sigma) or blasticidin (2 µg/ml; Invitrogen) started 2 days after transfection. After 14 days of selection, stable resistant clones were picked to 96-well plates and expanded.

Polymerase chain reaction analysis of hESC clones

Genomic DNA from the hESCs clones was prepared directly in 96-well plates and used for screening by polymerase chain reaction (PCR) for the presence of transposon inverted repeats and loss of ampicillin/spectinomycin cassette that occurs during transposition. The clones that contained a BAC integrated by transposase according to PCR analysis (PB5⁺ Amp⁻ PB3⁺ or SB5⁺ Spec⁻ SB3⁺) were checked for the BAC copy number by quantitative PCR (qPCR). Five to six pairs of primers were designed along each BAC randomly with the distance ∼30–40 kb between primer pairs (listed in Supplementary Experimental Procedures). The copy number was calculated by normalization of C_t-values for each primer pair to GAPDH gene relatively to wild-type cells, which contain two allelic copies of each genomic region.

Splinkerette PCR

The integration sites of BACs in the clones were determined using splinkerette PCR (41). Genomic DNA digested with ApoI or BstYI was ligated to 75 nM splinkerette adaptor (SPLK-A and SPLK-ApoI/BstYI). The 5′- and 3′-junctions were amplified using nested PCR and sequenced.

Differentiation of hESCs

For differentiation of H7.S6 hOCT4-GFP clones, bFGF was removed from the cultured medium for 10 days and then GFP expression was analysed by flow cytometry.

H7.S6 hPAX6-GFP clones and wild-type cells were differentiated to neural epithelial cells (42,43). Embryoid bodies (EBs) were formed in N2B27 medium with 10 µM transforming growth factor beta (TGFβ) receptor inhibitor SB431542 (Tocris). After 8 days, EBs were spread to tissue culture dishes coated with 100 µg/ml polyornithine and 10 µg/ml laminin (Sigma–Aldrich). Expression of PAX6 and GFP was checked by flow cytometry or immunocytochemistry on Day 14 of differentiation.

H7.S6 hGATA4-GFP clones and wild-type cells were differentiated to definitive endoderm (44). RPMI supplemented with B27 (1×), 1 mM sodium butyrate, 100 ng/ml activin A and 25 ng/ml Wnt3a (both from Peprotech) was used for the first day of differentiation. Next day, Wnt3a was omitted from the medium, and the cells were cultured further in RPMI with B27 (1×), 0.5 mM sodium butyrate and 100 ng/ml activin A. On Day 7, the cells were analysed for GFP and CXCR4 expression by flow cytometry, and for markers expression by qPCR.

Immunostaining and microscopy

The incubation with primary antibodies was for 1 h at room temperature with mouse anti-Oct4 (1:50, sc-5279; Santa Cruz), rabbit anti-Nanog (1:30, AB5731; Chemicon) or overnight at +4°C with mouse anti-Pax6 (1:30; Developmental Studies Hybridoma Bank). The cells were incubated with secondary antibodies diluted 1:500 (FITC goat anti-rabbit and TRITC goat anti-mouse; Jackson Immunoresearch Laboratories and Alexa633 goat anti-mouse; Molecular Probes, Invitrogen) for 1 h at room temperature. Fluorescence images were taken using Leica SP5 laser scanning confocal microscope.

Flow cytometry

The cells were dissociated and fixed in phosphate buffered saline (PBS) with 1% formaldehyde. For antibody staining, 10⁶ of live cells were incubated with R-Phycoerythrin (PE)-conjugated anti-CD184 or immunoglobulin G (IgG) isotype control (BD Biosciences) diluted 1:100 in PBS with 2% fetal calf serum (FCS) for 30 min at +4°C. The cells were analysed with flow cytometer LSR II (Becton Dickinson) using FACSDiva software. The data were processed with FloJo software.

Generation of OCT4 and NANOG fluorescent reporter H7.S6 cells using large constructs

To study pluripotency, lineage commitment and differentiation pathways in hESCs, we aimed to create a panel of reporter cell lines based on the expression of fluorescent proteins under control of selected promoters. Using recombineering, genomic regions containing the OCT4 and NANOG genes (19 and 25 kb, respectively) were subcloned from BACs and GFP or mCherry IRES neomycin cassettes were inserted at the initiating methionine codon (Figure 1a and Supplementary Figure S1). Stable H7.S6 hESC clones were established after electroporation and G418 selection. However, we failed to obtain clones with uniform expression of the reporters, despite the fact that OCT4 and NANOG are expressed in undifferentiated hESCs. All clones (n = 8) showed mosaic expression (59.9–85.0% positive cells by flow cytometry), which was further reduced when G418 selection pressure was removed. However, the non-fluorescent cells were not differentiated, as shown by staining with OCT4 and NANOG antibodies (Figure 1b). We also generated double stable reporter lines after a second round of electroporation. The double reporter clones H7.S6 OCT4-mCherry/OCT4-GFP and NANOG-mCherry/NANOG-GFP also displayed mosaic expression of both fluorescent transgenes, which only partially overlapped (Figure 1c), indicating that even these relatively large transgenes undergo random silencing.

Generation of hESC BAC reporters using PiggyBac transposition

To circumvent silencing, we decided to use BAC transgenes. BACs containing OCT4 and NANOG genes were modified with the GFP-IRES-neo-pA reporter cassettes and stable hESC clones were established using nucleofection (20). However, once again we observed mosaic expression (data not shown). Because small transposons have been shown to mediate efficient gene transfer in mESC, hESC and human cell lines (25,29,34–36), we decided to evaluate whether transposition could be applied to BAC transgenesis. Among several transposons, we first chose PiggyBac based on its reported efficiencies in mammalian cells (26,45). The OCT4 and NANOG BACs were further recombineered to insert a cassette into the BAC backbone that contained the PiggyBac ITRs (5′-313 bp, 3′-235 bp (46)); flanking an ampicillin resistance gene. For this purpose, we built a recombineering cassette in an R6K plasmid so that PacI or PacI/AscI restriction digestion releases a fragment that will recombine with most common human and mouse BAC vectors; pBACe3.6, pBeloBAC11, pTARBAC1, pTARBAC1.3, pTARBAC2, pTARBAC2.1 and pTARBAC6 (Figure 2a and Supplementary Figure S2). Then, the OCT4-GFP and NANOG-GFP BACs were transfected into H7.S6 hESCs using either nucleofection or lipofection, with or without co-transfected codon optimized PiggyBac transposase expression plasmid, mPBase (47). Without co-transfected mPBase, both transfection methods produced a similar number of resistant colonies (22 and 16 clones for OCT4-GFP and 5 and 4 clones for NANOG-GFP per 10⁷ transfected cells; Table 1 and Supplementary Table S1). With nucleofection, co-transfection of mPBase did not increase the colony number (29 resistant clones for OCT4-GFP and 3 clones for NANOG-GFP). However, co-lipofection of mPBase produced more colonies for both BACs (413 for OCT4-GFP and 14 for NANOG-GFP).

To check whether the BACs had been integrated by transposition, we established a PCR-based strategy for colony screening based on primers directed to the PiggyBac ITRs (PB5 and PB3) and the ampicillin (Amp) gene (Figure 2b). Transposition should integrate the ITRs yet exclude the ampicillin gene, whereas random integration with or without fragmentation could give any combination of PCR signals. Most of the colonies established by nucleofection, with or without mPBase, contained either the whole PiggyBac cassette (PB3+ Amp+ PB5+) or none of it (PB3− Amp− PB5−). The same was observed for the lipofection of the BACs alone (Table 1 and Supplementary Table S1). Notably, only lipofection with mPBase resulted in clones that had the signature of transposition (PB3+ Amp− PB5+; 18.8% of OCT4-GFP and 75% of NANOG-GFP clones). Characterization of clones obtained from nucleofection with or without mPBase indicated that all contained only limited pieces of the BAC and none of them was due to transposition, suggesting that nucleofection breaks BACs.

Given these encouraging results, we generated BAC transposon GFP reporters for the lineage-specific genes PAX6 and GATA4 and lipofected them into H7.S6 cells with or without mPBase. Because these genes are not expressed in hESCs, the IRES could not be used to drive neomycin resistance so we used the PGK promoter, which consequently provided similar efficiencies for both transfections (Table 1 and Supplementary Table S1). Application of the mPBase resulted in 1.7-fold increase of the colony number for both GATA4 and PAX6 BACs, and about one-third of those clones gave the transpositional signature. The use of the hyperactive version of PBase, hyPBase (48), further increased the total number of colonies and the proportion of transpositional events (61.7%). These lipofection results were essentially reproduced using another hESC line, H9 (Table 1 and Supplementary Table S1), which is more difficult to transfect (our unpublished data). The optimum ratio of BAC transgene to PBase expression vector was evaluated (Supplementary Table S2). In most cases, 10–30 µg BAC with 3–10 µg PBase per 10 cm dish (∼3 × 10⁶ cells), which corresponds to molar ratios of BAC to PBase from 1:9 to 1:30, gave the most colonies.

Furthermore, we generated a PAX6-GFP reporter BAC that contained PiggyBac ITRs flanking the vector backbone (ca. 9.5 kb apart from each other). This version also contained the Blasticidin resistance gene driven by the ubiquitin C promoter at one end of the genomic sequence next to the PiggyBac 3′-ITR (Supplementary Figure S2). Transposition by PBase (mPBase or hyPBase) resulted in clones that contained integrated BAC without the vector backbone, albeit with apparent lower efficiency. Thus, PiggyBac transposition can be used to exclude integration of the prokaryotic vector sequences into the genome.

Copy number of BAC integrations

To further characterize the BAC integrations, we analysed OCT4-GFP, GATA4-GFP and PAX6-GFP hESC clones for transgene copy number using qPCR on genomic DNA. Primer pairs were selected across the full-length BAC at a distance of 30–40 kb from each other (Figures 2b and 3). The C_t-values were normalized to wild-type DNA to calculate the number of additional copies arising from the integrated transgenes.

We analysed 17 clones positive for the transpositional signature (PB5+ Amp− PB3+; Figure 3). Most clones showed single-copy signals for all primer pairs indicative of a single transpositional event (12/17; 70.5%). A further three clones showed full-length, multi-copy integrations (three, two and four copies) suggesting multiple transpositional events. The remaining two clones showed signs of a partial integration of a second copy in addition to a single full-length copy, suggesting a combination of a single transpositional event and a partial random event. In addition, we characterized 19 clones that did not present the transpositional signature (either PB5+ Amp+ PB3+ or PB5− Amp− PB3−; Figure 3b). Most of these clones had missing or inconsistent signals for at least one primer pair indicating random integration of fragments.

Analysis of integration sites of the BACs

The integration sites were examined by splinkerette PCR and sequencing (41). We analysed 21 H7.S6 clones (including OCT4-GFP, OCT4-mCherry, NANOG-GFP, PAX6-GFP, SOX1-GFP and GATA4-GFP) and 4 H9 clones that were positive for the transpositional signature. In all cases, the integration locus was continuous on the 5′- and 3′-sides of a duplicated TTAA sequence (Table 2), which is characteristic of PiggyBac insertion sites (49). As controls we analysed the junctions of PiggyBac ITRs in several PB3+ Amp+ PB5+ clones that were established by transfection of the BAC without mPBase. As expected, the PiggyBac ITRs were followed by the ampicillin gene sequence and not genomic DNA, consistent with random integration of the BAC (data not shown).

Previous studies showed that the PiggyBac transposon can integrate into any chromosome with a preference for AT-rich sequences within 10 kb from transcription units (25,29,34). We analysed the sites of integrations of the BACs and they were either located in intergenic (n = 14) or intronic regions (n = 11) (Table 2) without any obvious chromosomal preference or DNA consensus around the TTAA site of the integrations (Supplementary Figure S3).

Functionality of the reporter clones

The integrity of the reporter BAC transgenes after transposition into hESCs can be functionally evaluated. Various clones were examined for the stability and pattern of expression before and after differentiation. Using single-copy OCT4-GFP clones integrated by mPBase (n = 7), we found that all clones uniformly expressed GFP (99%) in the undifferentiated state in the presence of G418 and notably after >1 month of culture without G418 selection pressure (96.13%; Figure 4a). The clones were induced to differentiate by bFGF removal (50,51) leading to the loss of GFP expression after 8 days. Hence, the OCT4-GFP BAC transgenes integrated by transposition are functional, reliable and not prone to position effects.

To validate lineage reporters for the genes that are not expressed in hESCs, we used PAX6-GFP and GATA4-GFP BAC transfected clones, created by transposition. H7.S6 PAX6-GFP cells were differentiated into neural epithelium cells in N2B27 through EBs in the presence of TGFβ receptor inhibitor (42,43). The attached EBs formed neural rosettes, which is the characteristic phenotype for neural progenitor cells. 74.81% of the cells were GFP positive and co-expressed GFP with endogenous PAX6 as shown by flow cytometry and immunostaining (Figure 4b).

H7.S6 GATA4-GFP BAC reporter cells were used to generate definitive endoderm by treatment with a high concentration of activin A (44). Quantitative reverse transcription–PCR (qRT–PCR) confirmed up-regulation of the endodermal markers GATA4, SOX17 and FOXA2 after 7 days of differentiation (Figure 4c). Most (83.67%) cells were GFP positive and co-expressed the marker of definitive endoderm CXCR4 shown by flow cytometry. Interestingly, the dynamic of the reporter expression closely reproduced recent observations on definitive endoderm differentiation, which showed that GATA4 expression preceeds CXCR4 (52) (data not shown). These data support the conclusion that the BAC transposon reporters are functional and faithfully reproduce endogenous patterns of gene expression during differentiation.

Sleeping Beauty BAC transposition

To determine whether the ability to transpose a very large cargo was a feature of PiggyBac, we undertook BAC transposon experiments with SB (Supplementary Figure S4). SB ITRs, separated by a spectinomycin resistance gene, were inserted into the PAX6-GFP BAC vector and co-lipofected with SB100xco transposase (a human codon optimized version provided by Zsuzsanna Izsvak). The combination of BAC and transposase expression plasmid resulted in >20-fold increase in the colony number compared with the control co-transfection of BAC with catalytic mutant of transposase (D3 construct). PCR screening identified the clones with transpositional signature (23/32; 72%) and most of them contained a single copy of the BAC. We mapped the transgene positions in three clones and confirmed a TA integration site that is characteristic for SB transposition (Supplementary Figure S4).