Modified Lentiviral LTRs Allow Flp Recombinase–mediated Cassette Exchange and In Vivo Tracing of “Factor-free” Induced Pluripotent Stem Cells

Abstract

Methods for generating induced pluripotent stem cells (iPSCs) for disease modeling and cell therapies have progressed from integrating vectors to transient delivery of reprogramming factors, avoiding permanent genomic modification. A major limitation of unmodified iPSCs is the assessment of their distribution and contribution to adverse reactions in autologous cell therapy. Here, we report that polycistronic lentiviral vectors with single Flp recombinase (Flp) recognition target (FRT) sites can be used to generate murine iPSCs that are devoid of the reprogramming cassette but carry an intergenic 300-bp long terminal repeat sequence. Performing quantitative polymerase chain reaction on this marker, we could determine genetic identity and tissue contribution of iPSC-derived teratomas in mice. Moreover, we generated iPSCs carrying heterospecific FRT twin sites, enabling excision and recombinase-mediated cassette exchange (RMCE) of the reprogramming cassette for another expression unit of choice. Following screening of iPSCs for “safe harbor” integration sites, expression cassettes were introduced by RMCE into various previously silenced loci of selected single-copy iPSCs. Analysis of DNA methylation showed that RMCE reverted the local epigenetic signature, which allowed transgene expression in undifferentiated iPSCs and in differentiated progeny. These findings support the concept of creating clonotypically defined exchangeable and traceable pluripotent stem cells for disease research and cell therapy.

Introduction

The first generation of induced pluripotent stem cells (iPSCs), introduced by Takahashi and Yamanaka in 2006,¹ was a breakthrough in the field of regenerative medicine. The possibility of creating pluripotent cells from somatic cells—which can then give rise to cells from all three germ layers—allows for the use of patient-specific pluripotent stem cells and their differentiated derivatives in disease modeling and drug research, as well as for the development of new cell therapies.² In the past 7 years, the methods for reprogramming somatic cells have evolved from simple integrating of γ-retroviral vectors transferring individual transcription factors to lentiviral vectors with polycistronic expression cassettes,^3,4 and even to transient methods such as episomal plasmid transfection, Sendai virus–mediated transduction, mRNA transfer, and delivery of proteins and small molecules.^5,6 These latter methods aim at yielding transgene-free iPSCs to circumvent the risk of insertional mutagenesis, which has been observed in clinical studies targeting hematopoietic stem and progenitor cells using integrating vectors with strong promoter/enhancer sequences.⁷ The use of unmodified stem cells and their differentiated progeny as sources of transplantable cell populations may create the basis for human cell and tissue products that are virtually indistinguishable from donor cells. Because cells derived from iPSCs or embryonic stem cells (ESCs) carry the risk of tumor formation,⁸ discrimination between extrinsic and intrinsic tumors is of high importance, especially for autologous transplantation strategies. We⁹ have proposed that, in such cases, genetically modified, identifiable, and traceable iPSCs with unique genetic IDs are beneficial for clinical applications. However, any DNA sequence that marks iPSCs should not interfere with cellular gene expression and thus ideally should be devoid of enhancer or promoter elements that could lead to activation of neighboring protooncogenes.¹⁰ Furthermore, the integration sites in iPSCs should not be located within genes or in close proximity to gene-enriched regions. To meet these criteria, selection of single-copy safe-harbor integrations is necessary.¹¹ Finally, iPSCs used for clinical settings should also be devoid of exogenous reprogramming factors to exclude residual expression, which can be oncogenic and might counteract differentiation.^12,13

Following this concept, we have developed a Flp recombinase (Flp)-excisable lentiviral reprogramming vector that yields bona fide iPSCs from different genetic backgrounds that do not contain the reprogramming factors but carry a small (~300 bp) promoterless and enhancerless sequence tag that allows the determination of iPSC-derived tissue contribution via quantitative polymerase chain reaction (qPCR). Together with the integration site of the lentiviral vector, the sequence tag creates a clone-specific genetic ID for every generated iPSC line. Clonal cell lines can then be screened and selected for safe-harbor integrations using ligation-mediated PCR. We expanded the versatility of our system by introducing heterospecific twin Flp recognition target (FRT) sites into our vector, enabling Flp-mediated excision and exchange of the reprogramming factors by an expression cassette of choice. After successful exchange into a previously silenced safe-harbor locus, we could show that the newly introduced enhancer and promoter sequences remained nonmethylated and enabled stable long-term expression of a hygromycin–thymidine kinase fusion protein (hygtk) in undifferentiated iPSCs and differentiated progeny. Thus, we developed a novel vector system for the generation of safe, clonally defined, and traceable iPSCs with an exchangeable transgene in a defined, epigenetically flexible locus for various clinical applications.

Results

Optimized vector design and culture conditions allow efficient generation of excisable single-copy iPSCs from fibroblasts derived from various murine disease models

We used a previously validated lentiviral reprogramming vector expressing human codon–optimized transcription factors Oct4, Klf4, and Sox2, plus wild-type (WT) c-Myc and the fluorescent marker dTomato.^4,14 Similar to γ-retroviral vectors previously described by our group,¹⁵ this lentiviral vector contains a single WT FRT site (“F”) in the ΔU3 region of the 3′ long terminal repeat (LTR). Because the 3′ ΔU3 region is copied to the 5′ end during reverse transcription, the resulting integrated proviral DNA is flanked by two identical FRT sites (Figure 1a, “excisable vector”). Lentiviral vectors produced by transient transfection of 293T yielded supernatants with high titers (~1 × 10⁷ transducing units/ml, data not shown). To generate iPSCs for disease phenotype studies, we isolated adult fibroblasts from a variety of murine disease models for granulocytic defects, including congenital neutropenia (p14^f/f conditional knockout (KO) mice),¹⁶ Wiskott–Aldrich syndrome (Wasp KO mice),¹⁷ and chronic granulomatous disease (gp91phox KO mice).¹⁸ Additionally, embryonic fibroblasts were isolated from mice with a KO of the gene Lef1 (Lef1 KO mice), which plays an essential role in granulopoiesis.¹⁹ For reprogramming, isolated fibroblasts were transduced with low amounts of the excisable lentiviral vector (multiplicity of infection of 0.2). To enhance iPSC generation, the medium used for the reprogramming process was supplemented with ascorbic acid (VitC) and valproic acid (VPA), which have been shown to increase reprogramming efficiency and quality (Figure 1b).^20,21,22 About 2 weeks after transduction, 12–16 iPSC colonies with ESC-like morphology (Figure 1c) from four different genetic backgrounds were picked and expanded on irradiated murine embryonic fibroblasts (MEFs). qPCR analysis for the vector-internal posttranscriptional regulatory element (PRE) revealed low vector copy numbers for all obtained iPSC clones (~50% of the generated clones contain only one copy of the reprogramming vector per genome) (Figure 1d).

Figure 1.

Optimized lentiviral vector design allows generation of induced pluripotent stem cells (iPSCs) with low vector copy number. (a) Schematic vector description (provirus depicted). We constructed a lentiviral reprogramming vector expressing human codon–optimized reprogramming factors linked by 2A processing sites (P2A, T2A, and E2A) and an internal ribosomal entry site (IRES)–driven fluorescence marker dTomato as previously described. We inserted a single wild-type (WT) Flp recognition target (FRT) (F) or dual heterospecific F+F3 twin sites (linked by a NotI restriction enzyme site) into the ΔU3 region of the lentiviral long terminal repeat (LTR) to generate excisable and exchangeable vector variants, respectively. The identity-specifying eight bases of each FRT site are underlined. (b) Illustration of the reprogramming protocol. To enhance the reprogramming efficiency, ascorbic acid (VitC) and valproic acid (VPA) were added to the medium (MEF, medium for fibroblasts, ES, medium for embryonic stem cells (ESCs)/iPSCs) during the reprogramming process. Medium was changed from MEF to ES on day 4 posttransduction (TD), and cells were split onto a feeder layer of irradiated C3H murine embryonic fibroblasts (MEFs) on day 8. ESC-like colonies were picked around day 16 to establish clonal iPSC lines. (c) After transduction with the reprogramming vectors, morphological changes in adult fibroblasts could be observed (left panel), which finally led to generation of iPSC lines with typical dome-shaped structure of murine ESCs (right panel). (d) We established >95 single iPSC lines from adult fibroblasts of the following genetic backgrounds: p14f/f (conditional loxP-flanked p14 knockout (KO), p14 can be deleted by Cre recombinase expression, equals WT), Lef1 KO (here, MEFs were used for reprogramming), Wasp KO, gp91 KO, C57BL/6 WT, and Hax1 KO. We used the excisable construct for reprogramming, except for C57/BL6 WT and Hax1 KO adult fibroblasts. Here, the adult fibroblasts were transduced with the recombinase-mediated cassette exchange (RMCE) vector carrying heterospecific FRT twin sites (F+F3). Quantitative polymerase chain reaction analysis for copy number of the vector-internal posttranscriptional regulatory element (PRE) revealed a high amount (~50%) of iPSCs containing a single copy of the reprogramming vector (n = 3). Values <1 are due to undepleted feeder cells in the genomic DNA preparation. Bars display mean values. Flp, Flp recombinase; SFFV, spleen focus–forming virus promoter.

Retroviral protein transfer of Flp recombinase mediates vector excision in murine iPSCs that maintain pluripotency marker expression, and hematopoietic differentiation potential

To exclude potential reactivation of the integrated reprogramming factors, we performed Flp recombinase–mediated excision of the reprogramming vector cassette. This was accomplished using a retroviral protein transduction technique, transferring Flp recombinase as part of the retroviral polyprotein precursor.¹⁴ This method has the advantage that no nucleic acids are transferred and that Flp activity is restricted to the half-life of the protein. For each of the four genotypes mentioned above, two single-copy iPSC clones were treated twice with retroviral protein transduction particles, and single cells were sorted and screened for excision of the lentiviral reprogramming cassette. To achieve this goal, we used a three-primer multiplexing PCR amplifying (i) a part of the reprogramming vector cassette (size: 290 bp) and (ii) the enhancer- and promoter-deprived LTR (size: 170 bp; Figure 2a). Successful excision of the reprogramming vector is detected by loss of the 290-bp fragment (Figure 2b). We detected recombined, reprogramming factor–excised subclones with a success rate of 10% (Supplementary Table S1) and verified the successful excision of the reprogramming factor sequences by validating loss of the vector-internal PRE element in genomic DNA of the excised subclones using qPCR (Figure 2c). Single vector copies and clonality of parental iPSC lines, as well as Flp recombinase–mediated excision, were further confirmed by Southern blot analysis (Supplementary Figure S1). Vector-excised iPSCs from four different genetic backgrounds maintained ESC-like morphology and stained positive for pluripotent stem cell markers, such as Oct4, Nanog, and SSEA1, as determined by immunohistochemistry and flow cytometry (Supplementary Figure S2a). Quantitative reverse transcription–PCR for a panel of various stem cell–specific proteins revealed expression levels similar to control C57BL/6 ESCs (Supplementary Figure S2b). To detect possible genotoxicity of the Flp recombinase treatment, such as chromosomal deletions, we performed comparative genome hybridization analysis (array-based comparative genomic hybridization; resolution: 246 kb; sensitivity: 0.5) for all excised subclones, which detected loss of an X chromosome in clones p14 #1 EX, p14 #2 EX, and gp91 #1 EX2, as well as three aberrations in clone gp91 #1 EX2 (Supplementary Figure S3 and Table S3). To test the functionality of “factor-free” iPSCs, we performed an in vitro differentiation into embryoid bodies (EBs). iPSCs readily formed cystic EBs, which contained a high percentage of CD41⁺ c-kit⁺ early hematopoietic progenitors (Supplementary Figure S4a).²³ Furthermore, the hematopoietic differentiation potential of two excised IPSC clones was demonstrated by generating Gr-1⁺ CD11b⁺ mature granulocytes using granulocyte colony-stimulating factor–containing media and coculturing with OP9 stromal cells (Supplementary Figure S4b and Methods).

Figure 2.

Transfer of Flp recombinase by retroviral protein transfer (RPT) particles mediates excision of the reprogramming cassette from single-copy induced pluripotent stem cells (iPSCs) without selection. (a) We applied our established RPT for excision of the whole vector cassette, leaving behind a single residual long terminal repeat (LTR). Flp-mediated recombination of the two homologous F sites was identified by multiplexing polymerase chain reaction (PCR) with three primers detecting the vector (primers 1+3; product size: 290 bp) and the residual LTR (primers 2+3; product size: 170 bp). (b) Multiplex PCR analysis of RPT-treated single-cell-derived subclones of Wasp KO iPSC clone Wasp #1, showing two excised subclones (marked EX1 and EX2). Loss of the upper 290-bp band indicates successful excision of the reprogramming vectors. Applying this method, we established excised iPSC subclones of four different genetic backgrounds p14f/f, gp91 KO, Lef1 KO, and Wasp KO. (c) Quantitative PCR detecting the vector-internal posttranscriptional regulatory element (PRE) element for determination of the absolute vector copy number (VCN) of the reprogramming vector in nontransduced fibroblasts and iPSCs (n = 3). Nontreated parental iPSC clones show VCN of 1.0–1.5. In excised iPSC subclones (termed EX), VCN is reduced to background levels, validating the generation of iPSCs devoid of the reprogramming vector cassette. Error bars display mean + SD. Flp, Flp recombinase; KO, knockout; SFFV, spleen focus–forming virus promoter.

qPCR detection of the intergenic residual LTR allows determination of iPSC-derived tissue contribution and identification of tumor origin in vivo

Next, we mapped the locations of the residual LTR in five excised iPSC clones using ligation-mediated PCR.²⁴ Genomic positions were identified by sequencing the PCR products. Encouragingly, four out of five analyzed clones contained potentially neutral intergenic integration sites. None of the five integrations sites fully matched the safe-harbor criteria defined by Papapetrou et al. (Supplementary Table S2). However, in this case, these criteria might be too stringent because undesired transactivating effects of the integrated enhancer- and promoter-deprived sequences are not expected (as it has been shown for similar enhancer and promoterless LTRs).¹¹ Therefore, in accordance with Papapetrou et al., selection of iPSC lines with intergenic residual LTRs is appropriate.

The 300-bp intergenic marker can serve as a clonotypic identifier to detect clonal contaminations in cell culture. An even more relevant feature is the potential to trace iPSC-derived progeny in transplantation settings. To test this hypothesis, we developed a qPCR protocol that specifically detects the residual LTR sequence. qPCR analysis of serially diluted DNA from clone p14 #2 EX (WT clone containing a single residual LTR) in DNA of control immunodeficient nonobese diabetic/severe combined immunodeficiency (NOD/SCID) γ chain^−/− (NSG) mice revealed a linear correlation between dilution factor and detected copy number, with a detection limit of ~1 LTR-marked iPSC in 3,000 cells (Figure 3a). To simulate a transplantation of undifferentiated cells leading to an iPSC-derived tumor, we performed teratoma formation assays in NSG mice, using different amounts (10⁶, 10⁵, and 10⁴) of p14 #2 EX iPSCs for subcutaneous injection. After 4 weeks, we observed teratomas (containing cells derived from all germ layers, Supplementary Figure S5) in all three mice that had been injected with 10⁶ cells and in one mouse injected with 10⁵ cells. At this point, all mice were sacrificed, and genomic DNA was prepared from teratomas, blood, kidneys, spleens, livers, and blood drawn on day 14 after injection. Two mice that had been injected with high doses of iPSCs (10⁶ cells) had developed a malignant ascites surrounding the teratoma, from which also samples were taken for genomic DNA preparation. Next, qPCR analyses for the residual LTR were performed on all genomic DNA samples, and tissue contributions derived from iPSCs were calculated (Figure 3b). As expected, analyzed teratomas contained a high percentage of iPSC-derived cells (55–75%), simulating a solid tumor derived from transplanted cells. We also detected iPSC contribution to a large degree (25–52%) in samples from the surrounding ascites and—with a frequency of 1 in 500 cells—in the blood of one teratoma-bearing mouse, suggesting extensive dissemination of iPSC-derived teratoma cells or iPSC-derived hematopoiesis.²⁵ In the mouse that received 10⁵ cells and developed a teratoma, we found a minor contribution of iPSCs to the kidney tissue, possibly due to retention and accumulation of teratoma-derived cells from the blood. Thus, qPCR of the 300-bp residual LTR sequence allowed detection of iPSC-derived cells after in vivo transplantation and identification of the iPSC origin of a solid tumor.

Figure 3.

The residual long terminal repeat (LTR) sequence for tracing of “factor-free” induced pluripotent stem cells (iPSCs) in vivo. (a) We performed quantitative polymerase chain reaction (qPCR) detecting the residual LTR and a control gene (PTBP2). Genomic DNA of iPSC clone p14 #2 EX (containing one copy of a residual LTR) was serially diluted with genomic DNA from spleen of control NOD/SCID γ chain^−/− (NSG) mice. qPCR revealed a linear correlation between detected residual LTR copy number and dilution factor, allowing detection of 1 iPSC in 3,000 cells. Detection threshold was set according to background signal in NSG control DNA (n = 6). (b) Seven NSG mice were subcutaneously injected with different amounts of p14 #2 EX iPSCs (mice 10⁶1, 10⁶2, and 10⁶3 with 10⁶ cells; 10⁵1 and 10⁵2 with 10⁵ cells; 10⁴1 and 10⁴2 with 10⁴ cells). Mice were sacrificed after 4 weeks, when teratomas were detectable in mice 10⁶1, 10⁶2, 10⁶3, and 10⁵1. Mice 10⁶1 and 10⁶2 had developed ascites surrounding the teratomas. Genomic DNA was prepared from teratomas, ascites, liver, kidney, spleen, and blood from days 14 and 28. We performed qPCR on all samples and calculated iPSC-derived tissue contribution. Teratomas contained high proportions (55–75%) of iPSC-derived tissue. Furthermore, LTR-marked cells were detected in ascites samples from mice 10⁶1 and 10⁶2, in the blood of mouse 10⁶1, and in the kidney of mouse 10⁵2, suggesting shedding of teratoma cells. b.t., below threshold; n.d., not detectable.

Generation of single-copy iPSCs with heterospecific “twin” F+F3 sites for recombinase-mediated cassette exchange

As mentioned before, the reprogramming process with integrating vectors requires selection of iPSC clones with single neutral intergenic insertions to avoid perturbation of physiological gene expression. Because the reprogramming cassette is obsolete after acquisition of the iPSC state, we wanted to explore the possibility of exchanging the reprogramming locus for a transgene of choice by recombinase-mediated cassette exchange (RMCE). This provides further benefit from careful locus selection and reduces the number of cumbersome screening approaches.²⁶ To enable RMCE with our reprogramming cassette vector system, we optimized the vector platform by introducing a second, mutated FRT site (“F3”) into the 3′ LTR of the reprogramming vector construct (Figure 1a). Hence, the resulting integrated provirus of this vector is flanked by F+F3 “twin sites,” which allow excision (in case Flp recombinase is delivered alone) or, alternatively, RMCE—if Flp recombinase is delivered together with donor DNA flanked by the same F and F3 sites (Figure 4a).²⁷ Transduction of adult fibroblasts isolated from C57BL/6 WT (BL6 WT) mice and from a Hax1 KO congenital neutropenia mouse model using low amounts of the F+F3 virus (multiplicity of infection of 0.1) resulted in successful establishment of ~30 iPSC lines with ESC-like morphology. PRE-specific qPCR detected a high proportion of single-copy iPSC clones (~50%, Figure 1d), of which 14 were chosen for insertion site mapping of the reprogramming vector by ligation-mediated PCR. We screened eight WT and six Hax1 KO iPSC clones for “safe” insertions and were able to identify a WT clone (“WT30”) that complied with the refined, stringent standards for safe-harbor loci (Supplementary Table S2). For further experiments, we chose three WT iPSC clones (WT21, WT22, and WT30) with single reprogramming vector integrations (confirmed by Southern blot, Supplementary Figure S6) and cultured them in embryonic stem (ES) cell medium supplemented with inhibitors for glycogen synthase kinase (GSK)3β and mitogen-activated protein kinase kinase (MEK) (termed “2i”), which allows feeder-free culture of pluripotent stem cells.²⁸ iPSCs cultured with 2i showed ESC-like expression levels of pluripotency-associated genes (Supplementary Figure S7). Array-based comparative genomic hybridization analysis (resolution: 13 kb; sensitivity: 0.5) confirmed a normal karyotype and detected two aberrations already present in parental fibroblasts as well as one additional aberration in clone WT21 (Supplementary Figure S8 and Table S4).

Figure 4.

Flp recombinase–mediated exchange of the reprogramming factor cassette in induced pluripotent stem cells (iPSCs) with a spleen focus–forming virus (SFFV) promoter–driven hygtk expression cassette. (a) Schematic representation of recombinase-mediated cassette exchange (RMCE) with F+F3 twin sites. Flp recombinase mediates exchange of the reprogramming cassette (“4-in-1” cassette) for a plasmid-derived SFFV-driven hygtk expression unit flanked by single F and F3 sites by site-specific recombination between two homologous sites. (b) RMCE workflow. To check whether the “4-in-1” cassette of iPSCs can be exchanged by RMCE, iPSC clones WT21, WT22, and WT30 were transfected with 3 µg of the donor plasmid pF.SF.hygtk.F3 and 1 µg of a codon-optimized Flp (Flpo-puro) expression plasmid, if not indicated otherwise. After hygromycin selection for 14 days, hygromycin-resistant iPSCs were harvested, and genomic DNA was prepared to verify successful RMCE in the polyclonal sample. For each experiment, single cell–derived iPSC subclones were derived via flow cytometric single-cell sorting. (c) RMCE in C57BL/6 wild-type iPSC cell lines WT21, WT22, and WT30. Exchange of the reprogramming cassette for the SFFV-driven hygtk selection marker in the unsorted polyclonal sample (P) was confirmed by polymerase chain reaction (PCR). Genomic DNA from single cell–derived iPSC clones was subjected to PCR analysis for detection of RMCE, random integration (RI) of the donor plasmid (RI-donor), and RI of the Flp expression vector (RI-Flp). In total, we detected five clones (WT21-SF1, WT22-SF2, WT30-SF1, WT30-SF7, and WT30-SF8), which were positive for RMCE and negative for RI of the donor and the Flp expression vector. For subclones WT30-SF8–SF10, 5 µg donor plasmid and 2 µg Flpo plasmid were used for transfection (marked with bold letters). (d) Quantitative PCR for the posttranscriptional regulatory element (PRE) element in the reprogramming vector validated loss of the PRE-containing “4-in-1” cassette in RMCE-positive iPSC clones WT21-SF1, WT22-SF2, WT30-SF7, and WT30-SF8, indicating successful cassette exchange (n = 3). Error bars display mean + SD. Flp, Flp recombinase; WT, wild-type.

Recombinase-mediated exchange of the reprogramming vector cassette for a gene of interest at different genomic positions in murine iPSCs

Next, we exchanged the reprogramming cassettes in iPSCs by Flp-RMCE. To see whether the tagged reprogramming vector remains accessible at different genomic locations, we performed RMCE with three iPSC clones containing different integration sites: WT21 (intergenic integration <10 kb distance from a neighboring gene), WT22 (intragenic integration), and WT30 (safe-harbor integration, see Supplementary Table S2). To this end, we transfected 2.5 × 10⁵ cells of these iPSC clones with a Flp recombinase expression plasmid and different amounts of donor DNA, which contains a hygtk selection marker under control of the spleen focus–forming virus (SFFV) promoter. This expression cassette is flanked by an FRT WT site (F) and an F3 mutant site (F3) to allow RMCE at the genomic locus after transfection (Figure 4a). After 2 weeks of hygromycin selection, single cells were sorted to obtain clonal iPSC lines without further antibiotic selection (Figure 4b). Using exchange-specific PCR and Southern blot analysis, we detected successful RMCE in genomic DNA from the polyclonal pool (PCR detection before sorting) and from clonal iPSC populations derived from all three input clones, which shows that Flp-mediated recombination is feasible at various genomic locations (Figure 4c, Supplementary Figure S6). PCR analysis of single clones for random integration of donor and Flp expression plasmids excluded random integration in five of six RMCE-positive subclones. Furthermore, loss of the reprogramming vector in selected RMCE-positive clones was validated by qPCR (Figure 4d) and by Southern blot for the vector-internal PRE element (Supplementary Figure S6). These clones maintained ESC-like morphology and expression of pluripotency markers (Supplementary Figure S7), demonstrating the successful generation of “factor-free” iPSCs with a gene of interest at a selected genetic location without random plasmid integration.

RMCE changes the dynamics of enhancer/promoter CpG methylation during iPSC differentiation

Considering the paradigm that reprogramming factors are silenced after completion of the reprogramming process,^4,29 we investigated whether the RMCE of a silenced DNA fragment for a segment of plasmid DNA can reverse this process and whether the exchanged locus permits active transcription during long-term culture or differentiation into EBs in absence of antibiotic selection (of note: we initially applied 14 days of hygromycin selection to establish RMCE-positive subclones). Therefore, we compared transgene expression and methylation levels of enhancer- or promoter-specific CpG islands by bisulfite sequencing in genomic DNA of the nonexchanged safe-harbor clone WT30 (containing the silenced reprogramming cassette) and its exchanged derivative WT30-SF7 (expressing hygtk with an SFFV promoter). These analyses were performed during short- and long-term (30 days) stem cell culture under 2i conditions and after differentiation into EBs. Although the SFFV enhancer/promoter sequences in the nonexchanged clone WT30 were only partially methylated in naive iPSCs (probably due to global DNA hypomethylation in 2i culture),³⁰ the proportion of methylated CpGs approached 100% on day 6 of EB differentiation (Figure 5a,b). This coincided with low residual expression of the reprogramming factors in naive WT30 iPSCs (detected by reverse transcription–qPCR, not detectable by fluorescence-activated cell sorting (data not shown)), which reached the limit of detection in EBs at day 6 (Figure 5c). In contrast, CpG methylation of the SFFV enhancer/promoter region in the exchanged subclone WT30-SF7 remained very low during EB formation and long-term stem cell culture (30 days; Figure 5d,e), correlating with a high and stable expression of the hygtk transgene (Figure 5f). To validate these findings, we repeated RMCE experiments and expression/methylation analyses using a donor cassette with a hygtk gene driven by an ubiquitous chromatin opening element (UCOE)³¹ and obtained similar data (showing even less CpG methylation of the UCOE promoter, Supplementary Figure S9). In summary, these results indicate that RMCE into the previously methylated safe-harbor locus induces a reset of the epigenetic state at the integration site, allowing active transcription even during differentiation of iPSCs.

Figure 5.

Recombinase-mediated cassette exchange (RMCE) reverses the epigenetic signature in transgene-silenced induced pluripotent stem cells (iPSCs) and allows stable expression of a newly introduced transgene in differentiated progeny. Transgene expression and CpG methylation of spleen focus–forming virus (SFFV) enhancer/promoter sequences were determined in the safe-harbor clone WT30 and its RMCE-positive subclone WT30-SF7 in the naive undifferentiated state (at least three passages after selection), after extended culture in medium supplemented with inhibitors for GSK3β and MEK (“2i”) and on day 6 of differentiation into embryoid bodies (EBs). As control for reprogramming factor expression, dTomato-sorted adult fibroblasts transduced (TD) with the F+F3 reprogramming vector were used. (a) Illustration of SFFV enhancer/promoter CpG methylation in transduced fibroblasts and in undifferentiated and differentiated iPSC clone WT30 containing the reprogramming vector with F+F3 sites (open circle, nonmethylated CpG; filled circle, methylated CpG). As expected, DNA methylation of the SFFV enhancer/promoter sequences in the reprogramming vector was nearly absent in dTomato-sorted fibroblasts. (b) Following reprogramming and clonal culture in naive stem cell conditions (ES medium plus inhibitors for GSK3β and MEK (2i)), SFFV enhancer and promoter sequences in iPSC clone WT30 showed modest CpG methylation levels (22 and 56%, respectively), which increased to nearly 100% after 6 days of differentiation into EBs (n = 6). (c) This coincided with low residual reprogramming factor expression detected by quantitative reverse transcription–polymerase chain reaction (RT–PCR) in naive iPSCs (3 per 1,000, compared with transduced (TD) fibroblasts), which was barely detectable in differentiated EBs (n = 3). (d) The RMCE-positive clone WT30-SF7 containing the SFFV-driven hygtk transgene displayed low CpG methylation of SFFV enhancer and promoter sequences in short- and long-term (30 days) stem cell culture and in differentiated EBs. (e) Unlike in the parental clone, differentiation of WT30-SF7 into EBs did not lead to full methylation of the SFFV enhancer and promoter but to a minor increase from 7 and 2% in naive iPSCs of 18 and 20%, respectively (n = 6). Furthermore, CpG methylation did not increase during long-term (30 days) culture of naive iPSCs. (f) Stable expression of the hygtk transgene was detected by quantitative RT–PCR during long-term culture without hygromycin selection and after differentiation of WT30-SF7 into EBs (n = 3). Error bars display mean + SD. PRE, posttranscriptional regulatory element; WT, wild-type.

Discussion

iPSCs have proven to be a valuable cell source for elucidating disease mechanisms, screening drug candidates, and developing innovative cell therapies.^32,33 Although transient methods to deliver reprogramming factors have been developed, integrating vector systems are still widely used for generating iPSCs. Advantages of integrating vectors include the ability to reprogram sensitive input material (e.g., cells that are difficult to transfect), to easily trace gene-marked cells, and to obtain higher efficiencies in establishing bona fide iPSC lines. However, the existence of integrated exogenous reprogramming factors creates the risk of uncontrolled reactivation, possibly leading to inhibition of differentiation and increased tumorigenesis.^12,13 Currently used integrating vector systems, therefore, include excisable platforms, which allow Cre recombinase–mediated excision of the integrated reprogramming genes due to inserted loxP sites.³⁴ The Flp/FRT–based reprogramming vectors introduced in this study have several advantages relative to Cre/loxP–based and transient technologies. First, mammalian genomes contain cryptic loxP sites that can mediate illegitimate recombination and genomic alterations in cells expressing Cre recombinase.³⁵ Existence of such “pseudosites” has not been shown for Flp recombinase, which recognizes larger target sequences (48 bp) than Cre recombinase (34 bp), providing higher specificity and therefore less potential for genotoxicity (the loss of X chromosomes in some of our generated iPSC lines is probably due to cell culture artifacts of female cells and not induced by Flp recombinase activity or the reprogramming process).^36,37,38 Second, the efficient generation of single-copy iPSCs with our optimized lentiviral vectors combined with excision of the reprogramming factors by protein transfer of Flp recombinase allows fast creation of “tagged” but “factor-free” iPSCs from various genetic backgrounds. These iPSCs are marked by single residual LTR sequences, which, as we show, are beneficial as they permit quantitative tracing of iPSC cells and differentiated progeny. This feature is important to assess the performance of iPSC-based cell therapies (e.g., by measuring chimerism after infusion of in vitro–generated blood cells or by determining contribution of iPSC-derived cardiomyocytes to repaired heart muscle tissue).³⁹ Moreover, the genetic marker allows identification of autologous iPSCs as a tumor source. The risk of residual undifferentiated stem cells in suboptimally differentiated transplants not only requires development of strategies to eliminate tumor-forming cells (e.g., by introduction of inducible suicide gene cassettes) but highlights the need to discriminate between extrinsic and intrinsic tumors.⁸ In cancer patients, minimal residual disease markers are frequently used as sensitive methods to assess the tumor burden in a proportion of tissue samples and these represent a powerful prognostic indicator.⁴⁰ The ability to detect adverse events by minimal residual disease monitoring will be critical for risk assessment and to guide decisions on treatment strategies during early phase trials with iPSC-based cellular therapeutics.

As a further refinement, we show that a pair of heterospecific FRT sites (F+F3 twin site) in the LTRs enables efficient cassette exchange in iPSCs when Flp recombinase is delivered together with a compatible donor sequence, permitting targeted insertion of an expression unit of choice into a defined genomic location. Insertion of therapeutic transgenes for genetic correction of hematopoietic diseases normally requires mass transduction of large populations of effector or progenitor cells (e.g., hematopoietic stem cells) by integrating vectors. This creates the risk of insertional mutagenesis, which can be caused by transactivation of proto-oncogenes through semi-randomly integrated vector-derived enhancer/promoter sequences.⁴¹ The associated risk can be circumvented in iPSC-based therapies because the potential for unlimited proliferation allows in vitro selection of clonal iPSC lines with defined integration sites and without genetic alterations. In an extensive study, including lentiviral reprogramming of patient cells, Cre-mediated vector excision, and secondary lentiviral gene transfer, Papapetrou et al. showed that such a selection of iPSC clones with single vector integrations of a lentiviral β-globin expression vector is feasible. They further defined novel, stringent properties for safe-harbor integration sites by quantifying activating effects of the integrated β-globin vector via analyzing expression levels of genes surrounding the integration sites in various clones.¹¹ Correspondingly, we analyzed the integration sites of the reprogramming vector in our single-copy murine iPSCs and identified a single clone (out of 19 analyzed) that fully met these criteria. The low frequency of safe-harbor integrations might be due to the specific lentiviral integration pattern and selection processes during the establishment of single-copy iPSCs from transduced fibroblasts, although this has not been observed in human iPSC clones with multiple integrations.⁴² Importantly, by using an RMCE-compatible vector, which allows deletion of the reprogramming factors and simultaneous addition of a cassette of choice at a predefined locus, the number of screening steps required to obtain locus-defined gene-modified iPSCs can be reduced to two: the selection of iPSCs with single exchangeable vectors in safe-harbor loci and the subsequent clonal analysis after cassette exchange. Making use of this streamlined process, Grabundzija et al. recently introduced an interesting Sleeping Beauty transposon-based vector system for RMCE in iPSCs.⁴³ However, difficulties in generating iPSCs with a single vector copy hinder efficient recovery of subclones compatible with single-locus RMCE in transposon-based reprogramming experiments. By contrast, our lentiviral vector system allows highly efficient generation of iPSCs containing single integrations (>50% of generated clones, Figure 1d)—a prerequisite for successful cassette exchange.

The process of RMCE provides the possibility of deriving various iPSC lines—expressing transgenes of choice—from one single well-characterized, clonal iPSC source. This will permit comparison of different therapeutic transgene cassettes in disease-specific iPSCs and progeny. Additionally, introduction of inducible cassettes with genes controlling differentiation processes can enhance the efficiency of in vitro production of desired cells and tissues for autologous donors.⁴⁴ Similarly, RMCE can be exploited to integrate other traceable transgenes for imaging purposes, such as the sodium–iodide symporter for imaging of transplanted cell types in vivo via single-photon emission computed tomography.³⁹ New technologies, such as “RMCE multiplexing,” further expand the options for creation of (i) multiple targetable loci in the genome⁴⁵ and (ii) vector cassettes that enable serial RMCE-based modifications at a single locus (allowing step-by-step insertion of different promoters, enhancers, or insulators to test their performance in vitro).

Because generation of differentiated cell types is the main application of iPSCs, it is of great interest to exclude epigenetic silencing at the former reprogramming locus during directed differentiation to achieve stable expression of the therapeutic transgenes. This requires an “open” epigenetic environment at a locus, that is, in our case, silenced during establishment of iPSCs. By analysis of CpG DNA methylation profiles of enhancer and promoter sequences with bisulfite PCR, we demonstrate that in iPSCs, the epigenetic state at this locus can be reversed by RMCE. There might be two reasons contributing to this observation: (i) RMCE introduces a “naked,” epigenetically unmodified part of nonmethylated plasmid DNA. In an earlier study, Schübeler et al. analyzed gene expression after RMCE using either nonmethylated or in vitro–methylated donor plasmids encoding enhanced green fluorescent protein (eGFP).⁴⁶ They observed long-term eGFP expression (10 weeks) in RMCE-positive cells, for which they had used nonmethylated donor DNA. Hence, the insertion of “naked” nonmethylated DNA by RMCE in our study might prevent de novo methylation of the previously silenced locus and thereby enables stable transgene expression. (ii) Epigenetic reversion could be due to the initial antibiotic selection (Figure 4) for hygtk-expressing cells and not per se be a result of the RMCE process itself. However, all analyzed RMCE-modified clones in this study underwent epigenetic reversion and allowed stable long-term transgene expression in undifferentiated cells and differentiated progeny even after selection was withdrawn. Therefore, the modified locus is not prone to be silenced again but rather remains permissive for transcriptional activity. This finding provides a basis for studies using RMCE-based gene correction of patient-derived iPSCs, wherein silencing of the transgene cassette would hinder therapeutic application.

In summary, we demonstrate that improved reprogramming vector design adapted for the use of Flp recombinase allows efficient generation of karyotypically normal murine iPSCs devoid of exogenous reprogramming factor sequences. Furthermore, we provide proof of principle for the traceability of iPSCs and derived progeny (including tumors) in an in vivo setting using qPCR, taking advantage of a sequence tag left behind after reprogramming cassette excision. Finally, an optimized vector containing heterospecific FRT twin sites enabled targeted cassette exchange of a single transcriptionally silent lentiviral reprogramming cassette in iPSCs by an expression cassette of choice at a safe-harbor locus, which remained nonmethylated and transcriptionally active during long-term culture and differentiation.

We therefore propose our lentiviral twin F+F3 reprogramming vector system as a promising candidate for generation of iPSCs that alleviate screening for effects of different transgene cassettes and which serve as traceable sources in preclinical and clinical studies for autologous cell and tissue transplantation therapies.

Materials and Methods

Plasmid construction. The “4-in-1” reprogramming vector pRRL.PPT.SF.hOKSMco.idTom.PRE, coexpressing human codon–optimized transcription factors and the fluorescent marker dTomato, has been previously described.⁴ To generate versions that allow Flp recombinase–mediated excision (pRRL.PPT.SF.hOKSMco.idTom.PRE.FRT) and cassette exchange (pRRL.PPT.SF.hOKSMco.idTom.PRE.F+F3), we introduced a WT FRT site or two heterospecific FRT sites (F+F3) into the promoter-deprived U3 region of the lentiviral vector construct, respectively. The plasmids Flpo-Puro and pF3-eGFP-F have been described previously.⁴⁵ To create plasmid pF.hygtk.F3, the WT F site and the mutated F3 site were exchanged for each other, and subsequently, the cDNA for eGFP was replaced with a cDNA encoding a hygtk. To generate the RMCE-donor plasmid pF.SF.hygtk.F3, a 432-bp SFFV promoter fragment was introduced between F and hygtk of plasmid pF.hygtk.F3. Plasmid pF.UCOErev.hygtk.F3 was generated by inserting a 1.5-kb A2UCOE sequence⁴⁷ (kindly provided by C. Brendel and M. Grez, Georg-Speyer-Haus, Frankfurt, Germany) into plasmid pF.hygtk.F3. All constructs were validated by sequencing. Cloning details are available on request.

Reprogramming adult fibroblasts into iPSCs. On day −1, fibroblasts (50,000 cells per well) were transduced in six-well plates with low amounts of viral supernatants (multiplicity of infection: 0.01–0.5) in 1 ml of MEF medium (low-glucose Dulbecco's modified Eagle's medium, 15% fetal bovine serum, 2 mmol/l l-glutamine, 100 units/ml penicillin/100 μg/ml streptomycin (all from PAA, Pasching, Austria), 0.1 mmol/l nonessential amino acids (Gibco, Karlsruhe, Germany) and 100 μmol/l β-mercaptoethanol (Sigma-Aldrich, Steinheim, Germany)) in the presence of 4 µg/ml protamine sulfate. Cells and vector particles were spinoculated for 1 hour at 2,000 rpm (863g) to enhance transduction and were incubated overnight. The next day (day 0) and on day 2, medium was changed to 2 ml MEF medium supplemented with 50 µg/ml 2-phospho-VitC and 2 mmol/l VPA. On day 4, medium was switched to 100% ES medium (KO Dulbecco's modified Eagle's medium (Gibco)), 15% ES-tested fetal bovine serum, 2 mmol/l l-glutamine, 100 units/ml penicillin/100 μg/ml streptomycin (all from PAA), 0.1 mmol/l nonessential amino acids (Gibco), 100 μmol/l β-mercaptoethanol (Sigma), and 10³ units/ml leukemia inhibitory factor (kindly provided by the Department of Technical Chemistry, Leibniz University Hannover, Germany)⁴⁸ supplemented with VitC+VPA. On day 8, cells were trypsinized (0.05% trypsin–ethylenediaminetetraacetic acid) and split 1:30 on plates with irradiated (30 Gy) MEFs prepared from pregnant (d13.5) C3H mice (Charles River, Sulzfeld, Germany). ES medium with VitC+VPA was changed every 2 days until the emerging iPSC colonies were harvested around day 16 posttransduction. Harvested colonies were expanded in ES medium on MEFs.

Transfections for RMCE. iPSCs (2 × 10⁵) were cotransfected with 3–5 µg of donor vectors F.SF.hygtk.F3 or F.UCOErev.hygtk.F3 and 1–2 µg of Flpo-puro (or 1–2 µg of BSpac-Δp as negative control)⁴⁹ according to the Metafectene transfection protocol (Biontex, Martinsried, Germany). After transfection, cells were cultured in T25 flasks (Sarstedt, Nuembrecht, Germany) with 5 ml of ES medium supplemented with inhibitors for GSK3β and MEK (“2i”). One day posttransfection, the culture medium was changed to ES medium supplemented with 150 U/ml hygromycin b (hygB). Cells were cultivated under selection conditions for ~14 days. Genomic DNA of the hygromycin-resistant cell pool was prepared and subjected to PCR analyses to confirm RMCE. Single-cell clones were obtained from the cell pool via single-cell sorting (MoFlo, Beckman-Coulter, Krefeld, Germany). After expansion, genomic DNA was prepared and subjected to PCR analyses. To assess RMCE, primer RMCECHECK5, which binds in the donor backbone, and primer SINFRTREV, which binds at the residual SIN LTR, were used to generate a 392-bp amplicon. To investigate random genomic integration of the donor plasmids F.SF.hygtk.F3 and F.UCOErev.hygtk.F3, primers RMCECHECK5 and JS Seq #5 forw were used. A 432-bp amplicon was detected in samples with random integration of these donors. To assess random integration of the Flpo-Puro plasmid, primers Flpo5 and Flpo3 were used to generate a 349-bp amplicon from the Flpo coding sequence.

Additional methods are described in Supplementary Material and Methods.

SUPPLEMENTARY MATERIAL Figure S1. Southern blot analysis probing the vector-internal PRE element confirms generation of single copy iPSC clones from different genetic backgrounds as well as reprogramming vector excision after protein transfer of Flp recombinase. Figure S2. Excised “factor-free” iPSCs show ESC-like morphology and expression of pluripotency-specific genes. Figure S3. Array-based comparative genomic hybridization (array-CGH) analysis of reprogramming factor-excised iPSCs from different genetic backgrounds. Figure S4. “Factor-free” iPSCs from different genetic backgrounds form embryoid bodies containing hematopoietic progenitors and differentiate into neutrophils. Figure S5. Teratoma sections of clone p14 #2 EX contain cells originating from all three germ layers. Figure S6. Southern Blot analysis probing the PRE element of the reprogramming vector and the HYG gene in the RMCE vector confirms generation of single copy iPSC clones and successful RMCE. Figure S7. C57BL/6 wildtype iPSCs WT21, WT22 and WT30 as well as the RMCE-positive subclones WT30 SF7 and WT30 SF8 show ESC-like morphology and expression of pluripotency-specific genes. Figure S8. Array-based comparative genomic hybridization (array-CGH) analysis of C57BL/6 adult fibroblasts and iPSCs generated with the RMCE-compatible reprogramming vector. Figure S9. Unmethylated UCOE sequences allow constant gene expression after successful RMCE in C57BL/6 wildtype iPSC clone WT30 using a UCOE-hygtk donor. Table S1. Calculated efficiencies of successful reprogramming vector excision in single cell subclones after retroviral protein transfer of Flp recombinase. Table S2. Genomic locations of the residual LTR (EX clones) or the RMCE-compatible reprogramming vector in single copy iPSCs. Table S3. Array-CGH Knock-out miPSC Aberration Report. Table S4. Array-CGH BL6 WT miPSC Aberration Report. Material and Methods

Supplementary Material

Supplementary Figure S1

Southern blot analysis probing the vector-internal PRE element confirms generation of single copy iPSC clones from different genetic backgrounds as well as reprogramming vector excision after protein transfer of Flp recombinase.

Supplementary Figure S2

Excised “factor-free” iPSCs show ESC-like morphology and expression of pluripotency-specific genes.

Supplementary Figure S3

Array-based comparative genomic hybridization (array-CGH) analysis of reprogramming factor-excised iPSCs from different genetic backgrounds.

Supplementary Figure S4

“Factor-free” iPSCs from different genetic backgrounds form embryoid bodies containing hematopoietic progenitors and differentiate into neutrophils.

Supplementary Figure S5

Teratoma sections of clone p14 #2 EX contain cells originating from all three germ layers.

Supplementary Figure S6

Southern Blot analysis probing the PRE element of the reprogramming vector and the HYG gene in the RMCE vector confirms generation of single copy iPSC clones and successful RMCE.

Supplementary Figure S7

C57BL/6 wildtype iPSCs WT21, WT22 and WT30 as well as the RMCE-positive subclones WT30 SF7 and WT30 SF8 show ESC-like morphology and expression of pluripotency-specific genes.

Supplementary Figure S8

Array-based comparative genomic hybridization (array-CGH) analysis of C57BL/6 adult fibroblasts and iPSCs generated with the RMCE-compatible reprogramming vector.

Supplementary Figure S9

Unmethylated UCOE sequences allow constant gene expression after successful RMCE in C57BL/6 wildtype iPSC clone WT30 using a UCOE-hygtk donor.

Supplementary Table S1

Calculated efficiencies of successful reprogramming vector excision in single cell subclones after retroviral protein transfer of Flp recombinase.

Supplementary Table S2

Genomic locations of the residual LTR (EX clones) or the RMCE-compatible reprogramming vector in single copy iPSCs.

Supplementary Table S3

Array-CGH Knock-out miPSC Aberration Report.

Supplementary Table S4

Array-CGH BL6 WT miPSC Aberration Report.