Here we describe the sustained expression of transgenes introduced into human embryonic stem (ES) cells using self-inactivating lentiviral vectors. At low multiplicity of infection, vesicular stomatitis virus-pseudotyped vectors containing a green fluorescent protein (GFP) transgene under the control of a human elongation factor 1α promoter transduced human ES cells at high efficiency. The majority of the transduced ES cells, which harbored low numbers of integrated vectors, continued to express GFP after 60 days of culture. Incorporation of a scaffold attachment region (SAR) from the human interferon-β gene into the lentiviral vector backbone increased the average level of GFP expression, and inclusion of the SAR together with a chromatin insulator from the 5′ end of the chicken β-globin locus reduced the variability in GFP expression. When the transduced ES cells were induced to differentiate into CD34+ hematopoietic precursors in vitro, GFP expression was maintained with minimal silencing. The ability to efficiently introduce active transgenes into human ES cells will facilitate gain-of-function studies of early developmental processes in the human system. These results also have important implications for the possible future use of gene-modified human ES cells in transplantation and tissue regeneration applications.
Human embryonic stem (ES) cells have been shown to differentiate in vitro into hematopoietic, endothelial, cardiac, neural, pancreatic, and trophoblast cells, and probably have the ability to differentiate into any cell of the body [1–6]. Because human ES cells can be expanded without a known limit , they offer a scalable source of human cells for basic biology, drug discovery, and transplantation medicine. Efficient genetic manipulation is one of the essential techniques needed to establish human ES cells as a widely used research tool. Unlike mouse ES cells, standard transfection strategies have so far proved inefficient with human ES cells, with the most effective approaches yielding approximately 1/105 stable transfectants .
Transgenic mice were first generated by infection with murine leukemia virus (MLV) over 25 years ago , but silencing of MLV-based oncoretroviral vectors has precluded their application as a practical alternative for creating transgenic animals and has limited their use in ES cell research. Recently, lentiviral vectors based on the human immunodeficiency virus (HIV) have been developed , and these have been demonstrated to resist silencing in mouse ES cells and in transgenic mice [11–13]. Here we demonstrate the utility of these vectors for delivery of functional transgenes into human ES cells. We also report that inclusion of a scaffold attachment region (SAR) in the vector backbone improved transgene expression levels both during prolonged culture of undifferentiated cells and following in vitro differentiation into hematopoietic cells .
Materials and Methods
Maintenance and Differentiation of Human ES Cells
The human ES cell line H9 was derived and maintained as described in fibroblast-conditioned medium on Matrigel [1, 15]. To promote hematopoietic differentiation, the human ES cells were cocultured with the mouse bone marrow stromal cell line S17 and analyzed at 17 days. Media to support hematopoietic differentiation consisted of Dulbecco's modified Eagle's medium (Invitrogen Corp.; http://www.invitrogen.com) supplemented with 20% fetal bovine serum (HyClone; http://www.hyclone.com), 2 mM L-glutamine, 0.1 mM β-mercaptoethanol, and 1% nonessential amino acids .
Undifferentiated H9 cells were washed with phosphate-buffered saline (PBS) and dissociated with 0.05% trypsin/0.53 mM EDTA (Invitrogen Corp.). Dissociated cells were filtered through 85-μm Nitex mesh to remove remaining clumps. To collect GFP+ cells, dissociated cells were sorted on a FACSVantage SE equipped with DiVa electronics and software (BD Biosciences). The GFP signal was excited with an argon laser tuned to 488 nm at 200 mW of light and the emission was collected through a 530/30 bandpass filter. Cells were sorted at 12 psi using a 90-μm nozzle tip and collected in 100% fetal calf serum. To analyze CD34 expression, the differentiated cell mixture (H9/S17) was dissociated with 1 mg/ml collagenase IV (Invitrogen Corp.) and 0.05% trypsin/0.53 mM EDTA. The single-cell suspension was stained with CD34-phycoerythrin (BD Biosciences PharMingen; http://www.pharmingen.com). Propidium iodide was used to exclude dead cells, and live cells were analyzed for GFP and CD34 expression on a FACSCalibur equipped with CELLQUEST software (BD Biosciences).
Lentiviral Vector Preparation and Transductions
The derivation of the lentiviral vectors used in this study has been described elsewhere (A.R., T.S. Hawley, R.G.H., manuscript submitted). In brief, a 178-bp DNA fragment encompassing the cPPT and central termination sequence of HIV-1 was amplified by PCR from the pCMVΔR8.91 lentiviral packaging plasmid [10, 16] (provided by D. Trono, University of Geneva, Geneva, Switzerland). The PCR product was digested with NarI (provided by the forward primer) and ClaI (provided by the reverse primer) and subcloned into the SIN-EF1α-GFP vector  at a unique ClaI site upstream of the EF1α promoter, giving rise to SINF-EF1α-GFP. The interferon-β SAR-containing SINF-EF1α-GFP-SAR vector was constructed by inserting a 0.8-kb fragment of the SAR element from plasmid pCL  (provided by J. Bode, Gesellschaft für Biotechnologische Forschung mbH, Braunschweig, Germany), modified to contain 5′ XhoI and 3′ SalI sites, in reverse orientation into a unique XhoI site in the 3′ untranslated region of the GFP gene in SINF-EF1α-GFP. The 5′HS4-containing SINF-EF1α-GFP-HS vector was constructed from SINF-EF1α-GFP by blunt-end ligating a 1.2-kb fragment containing the chicken β-globin 5′HS4 insulator in direct orientation between nucleotides -418 and -18 relative to the U3/R boundary in the U3 region of 3′ LTR. The 5′HS4 insulator was obtained from plasmid pJC13-1, which contains four copies of the element  (provided by G. Felsenfeld, National Institute of Diabetes and Digestive and Kidney Diseases, National Institutes of Health, Bethesda, MD). The SINF-EF1α-GFP-SAR/HS vector contained SAR and 5′HS4 sequences in corresponding positions and orientations as described for the respective vectors carrying either element alone. It was generated by replacing an XhoI-XbaI 3′ LTR-containing fragment of SINF-EF1α-GFP with an analogous fragment obtained from an intermediate plasmid consisting of the SAR linked to the 3′ LTR harboring the 5′HS4 insulator within the ΔU3 region.
The vectors were packaged in 293T cells by cotransfection with the pMD.G plasmid encoding the VSV-G envelope glycoprotein and the pCMVΔR8.91 plasmid expressing Gag, Pol, Tat, and Rev [10, 16]. Three days after transfection, vector particles were harvested from the medium by ultracentrifugation and titered using HT1080 cells . Undifferentiated human ES cells were washed with PBS and dissociated with 0.05% trypsin/0.53 mM EDTA for 5 minutes. Dissociated cells and vector particles were incubated with 6 μg/ml polybrene (Sigma). Media was changed the next day.
Genomic DNA of the four GFP+ cell populations was extracted after cell sorting, and quantitative PCR was performed to detect GFP sequences to determine the vector copy number in each cell population. The vector copy number was determined by comparison with a standard curve of plasmid DNA containing the GFP gene. The quantitative PCR reactions were performed on an ABI PRISM 7700 Sequence Detection System (Applied Biosystems). The GFP primer and probe set used were: probe, FAM-CCG ACC ACA TGA AGC AGC ACG ACT T-TAMRA, forward primer, 5′-GCA GTG CTT CAG CCG CTA C-3′ and reverse primer, 5′-AGC CTT CGG GCA TGG C-3′. The primer and probe set against human 18S ribosomal DNA (Applied Biosystems) was used as the external positive control. The PCR thermal cycle was 50°C for 2 minutes, 95°C for 10 minutes, 40 cycles through 94°C for 15 seconds, and 60°C for 1 minute.
Data presented in Figure 1 and Table 1 are means ± standard error of three separate experiments carried out in duplicate. p values were calculated using a one paired, one-tailed Student's t-test, with a p value <0.05 considered to be significant.
Table Table 1.. The 5′HS4/SAR combination reduces chromosomal position effects on SIN lentiviral vectors in transduced ES cells
Transduced ES cells were analyzed by FACS for GFP expression at day 60. The CV values of three duplicate samples of each group were averaged, and a one-paired, one-tailed Student's t-test was performed by comparing the SINF-EF1α-GFP group to the other three groups.
*The results indicated that the CV values of the SINF-EF1α-GFP-SAR/HS group were significantly less than those of the SINF-EF1α-GFP group (p < 0.05).
Mean ± SE
69.3 ± 1.4
73.5 ± 3.6
76.6 ± 4.1
60.6 ± 3.2*
Results and Discussion
The lentiviral vectors used in this study are third-generation, self-inactivating (SIN), HIV-based constructs containing a 400-bp deletion in the U3 region (ΔU3) of the 3′ LTR, which results in the transcriptional inactivation of the 5′ LTR following reverse transcription and chromosomal integration (Fig.2) [16, 19] (A.R., T.S. Hawley, R.G.H., manuscript submitted). A central polypurine tract within the pol gene (cPPT), which facilitates nuclear translocation of the HIV preintegration complex and improves lentiviral transduction efficiency in several types of cells, was included in the vectors [20, 21]. The human elongation factor 1α (EF1α) promoter was chosen as the internal promoter to drive expression of the enhanced green fluorescent protein (GFP) transgene because transient transfection experiments demonstrated its effectiveness as a strong promoter in human ES cells (data not shown). This promoter contains an intron that is successfully transmitted by the lentiviral vector to target cells, resulting in enhanced transgene expression [16, 19]. The packaging construct employed does not contain the vif, vpr, vpu, or nef genes involved in the pathogenesis by wild-type HIV. The vectors were pseudotyped with the G glycoprotein of vesicular stomatitis virus (VSV-G), which facilitates concentration of vector particles to high titers and provides a broad target cell tropism .
The SINF-EF1α-SAR vector (F denotes central DNA flap sequences) contains a 0.8-kb fragment of the human interferon-β SAR between the GFP gene and the deleted 3′ LTR that is maintained in this position after integration of the vector into genomic DNA [17, 23]. SARs consist of adenine/thymine-rich DNA sequences that associate with the nuclear scaffold or nuclear matrix in vitro. SARs have been shown to increase transgene expression both in stably transfected cells and in transgenic mice, and are believed to collaborate with enhancers to generate an extended domain of accessible chromatin; they have also been shown to decrease de novo methylation of transgenes [14, 23]. The SINF-EF1α-GFP-HS (HS denotes the 5′HS4 insulator herein) vector has a 1.2-kb fragment of the chicken β-globin 5′ constitutive hypersensitive site 4 (5′ HS4) insulator inserted into the ΔU3 region of 3′ LTR that allows flanking copies of the element to surround the transgene cassette after vector integration [18, 24]. Insulators are DNA elements that have enhancer-blocking and chromatin-boundary activities that prevent chromosomal condensation . Recent studies have shown that the enhancer-blocking and barrier functions of the chicken 5′HS4 insulator are separable activities . The SINF-EF1α-GFP-SAR/HS vector contains both SAR and 5′HS4 sequences.
To examine efficiency of lentiviral transduction, 1 × 105 human ES cells were exposed at increasing multiplicity of infection (MOI) to SINF-EF1α-GFP vector particles. Transduction efficiency was evaluated by monitoring GFP expression using fluorescence-activated cell sorting (FACS) analysis. At 72 hours posttransduction, 20% of the ES cells expressed GFP when transduced at an MOI of 1, which increased to 87% at an MOI of 12 (Fig.1A).
To investigate long-term transgene expression, human ES cells were transduced with the SINF-EF1α-GFP, SINF-EF1α-GFP-SAR, SINF-EF1α-GFP-HS, or SINF-EF1α-GFP-SAR/HS vectors at an MOI of 1. Each experiment was repeated three times. At 72 hours posttransduction, GFP+ ES cells were sorted by FACS. Quantitative polymerase chain reaction (PCR) confirmed that these GFP-sorted human ES cells had a low average vector copy number (1-2 copies/cell). Transgene expression was examined by FACS analysis during prolonged culture without additional selection. There was a decline in the percentage of cells expressing GFP between 72 hours and 30 days (Fig.1B), but it is not clear how much of this was due to transient expression from nonintegrated vectors . By 60 days posttransduction, the percentage of cells expressing GFP did not differ significantly among the four vectors (by one paired, one-tailed Student's t-test).
Monitoring of the level of transgene expression during the 2-month period demonstrated that the mean GFP fluorescence intensity (MFI) of cells transduced with either SINF-EF1α-GFP-SAR or SINF-EF1α-GFP-SAR/HS was up to 100% higher than that of cells transduced with SINF-EF1α-GFP or SINF-EF1α-GFP-HS (Fig.1C). At day 60, the coefficient of variation (CV) of the fluorescence intensity values, which is a relative indicator of homogeneity of expression, was significantly less for SINF-EF1α-GFP-SAR/HS-transduced ES cells compared with SINF-EF1α-GFP-transduced cells (by one paired, one-tailed Student's t-test), suggesting that this vector has a somewhat reduced susceptibility to chromosomal position effects (Table 1).
Because transgene silencing might be most pronounced during differentiation, lentiviral vector expression was also examined following in vitro differentiation of human ES cells into hematopoietic cells. At 30 days posttransduction, transduced ES cells were induced to differentiate into CD34+ cells by coculturing on S17 mouse stromal cell feeders. We have previously demonstrated that CD34+ cells obtained by this procedure are enriched for hematopoietic precursors . FACS analyses indicated that about 1% of the transduced ES cells in each of the four vector groups differentiated on S17 cells was CD34+. This percentage is comparable to previous results with nontransduced human ES cells  and suggests that transduction and GFP expression did not dramatically alter the potential of human ES cells to differentiate into hematopoietic cells. GFP expression in CD34+ populations was analyzed, and a similar percentage of GFP-expressing cells was observed in undifferentiated and differentiated (CD34+) ES cells for any of the four vectors (Fig.3).
These results demonstrate that lentiviral vectors support efficient transduction and long-term transgene expression in human ES cells, confirming and significantly extending previous findings with this vector platform in mouse ES cells . The data also corroborate other recent studies with this series of lentiviral vectors, which demonstrated that inclusion of the interferon-β SAR improves transgene expression levels (A.R., T.S. Hawley, R.G.H., manuscript submitted). Notably, differentiation of human ES cells into CD34+ cells did not increase the rate of silencing as has been observed for MLV-based oncoretroviral vectors during differentiation of mouse ES cells . This could mean that lentiviral vectors are relatively resistant to silencing during human ES cell hematopoietic differentiation, or that favorable sites of integration for expression had already been selected during FACS and culture of ES cells prior to CD34+ cell differentiation. Because high levels of sustained expression have been a limiting factor for gene therapy involving the hematopoietic system, these observations suggest the intriguing possibility that optimizing these vectors for use in human ES cells may result in improved performance in human hematopoietic stem cells. It will be important to test whether these vectors are also resistant to silencing during differentiation along other somatic cell lineages.
Genetic manipulation of human ES cells using liposome transfection reagents has previously been described . In our hands, these reagents typically give stable (drug selectable) transfection frequencies of approximately 1/105 human ES cells, and these rates are high enough for many routine applications. However, the high stable transduction efficiencies that lentiviral vectors allow are essential for certain types of experiments. First, if the expression of a particular gene causes apoptosis or differentiation, selecting stably transduced ES cell clones that overexpress the target gene will fail. If the ES cells can be transduced at high efficiency with such a gene, the function of the gene can be ascertained immediately after transduction without having to select stable ES cell clones. Second, if the goal is to express a gene from a tissue-specific promoter during ES cell differentiation, it is usually necessary to preselect prior to differentiation for stable transfection with a second promoter controlling a selectable marker. While this approach works in many cases, promoter interference can preclude its use in specific instances. An alternative is to transduce a high percentage of the cells without drug selection, and select individual ES cell clones by PCR. Third and most important, high stable transfection efficiencies are essential for expression cloning, a method that can be used to screen for phenotypic effects of individual cDNAs in a complex cDNA library.
Thus, the efficiency with which lentiviral vectors can transduce human ES cells and sustain expression suggests that these vectors can be used to support the identification of key developmental control genes by expression cloning. Although expression cloning has been used in a variety of contexts in developmental biology, to our knowledge the approach has never been used with ES cells, probably because of the inadequacies of previous vectors. For example, MLV-based oncoretroviral vectors have been used successfully for expression cloning , but the vectors that were employed are efficiently silenced in ES cells. A single 6-well plate of human ES cells contains 0.6-1.2 × 107 cells, so given the high transduction efficiencies of lentiviral vectors, very complex cDNA libraries can potentially be screened. If a developmental event is controlled by a specific gene at the transcriptional level, it should be possible to identify that gene through expression cloning by designing suitable phenotypic selection strategies in ES cells and their derivatives. Selection strategies could include, for example, genetically modifying the ES cells with selectable markers expressed in a specific cell lineage, or simply using cell surface antigens and FACS. Expression cloning using lentiviral vectors and human ES cells should provide a powerful new functional approach for understanding the human genome.
This work was supported by National Institutes of Health grants R24 RR16209 (to J.A.T.), and R01 HL65519 and R01 HL66305 (to R.G.H.). We thank Kathleen Schell for cell sorting; Dr. Murray Clayton and Dr. Jia-qiang He for assistance with statistics; Christine A. Daigh, Lynn Schmidt, and Dr. Jamie Sperger for critically reading the manuscript, and Robert A. Becker for preparing the figures.