Introduction

With completion of the human (and other) genome projects definition of gene functions using functional genomic technologies (such as gene knock-outs or transgenesis) has become even more important. As the relevance of single-gene approaches is often limited, novel techniques aimed at elucidating the functions of gene networks by either ectopic expression or targeted downregulation of several genes are urgently needed. Lentiviral (LV) vectors are useful tools for analyzing gene functions in different tissues by either over-expression of a gene-of-interest or its downregulation using short hairpin RNA technology.1, 2 To facilitate simultaneous analysis of multiple genes we have recently introduced lentiviral gene ontology (LeGO) vectors.3 This new tool comprises a set of LV vectors containing different expression cassettes (for transgenes and/or short hairpin RNAs of interest) in conjunction with individual fluorescent marker genes.3, 4, 5, 6

Fluorescent proteins (FPs) are well suited for marking and selecting gene-modified cells. However, selection of cells based on FP expression requires fluorescence-activated cell sorting (FACS), a laborious and expensive procedure not readily available for many scientists. Furthermore, FACS may be too stressful for some less robust cells and bears a certain contamination risk. In our initial work, we have shown that the use of drug-selectable fusion genes, consisting of eGFP or dTomato and the blasticidin resistance (BSD) may provide a strategy to overcome these FACS-associated problems.3

In keeping with the ‘building blocks’ principle we have now established and tested a variety of LeGO vectors encoding different drug-selectable FPs. To this end, we have created 10 fusion proteins representing combinations of one of the five different FPs (Cerulean (Cer), Venus (V), eGFP (G), dTomato (T) and Cherry (C)) and one of the five different resistance genes frequently used in cell biology (neomycin (Neo), hygromycin (Hygro), puromycin (Puro), blasticidin (BSD) and zeocin (Zeo) resistance). We also present data on the performance of new, codon-optimized variants of Neo, Hygro and Puro genes. Finally, we show that LeGO vectors encoding these novel drug-selectable FPs facilitate fast and efficient drug-mediated selection of cells, including neural stem cells (NSCs), concurrently or sequentially transduced with the different constructs.

Results

Novel blasticidin-selectable fluorescent protein genes are functional

We earlier reported3 the generation of eGFP/BSD and dTomato/BSD fusion constructs. To generate novel drug-selectable FPs representing easily distinguishable colors, we extended this platform by constructing gene fusions of BSD with Cerulean, Venus or mCherry (Figure 1).

Figure 1
figure 1

Schematic representation of new basic lentiviral gene ontology (LeGO) constructs encoding novel drug-selectable fluorescent proteins (FPs). (a) New vectors are based on LeGO–G/BSD, a third-generation lentiviral (LV) vector encoding an short hairpin RNA (shRNA) expression cassette in addition to a drug-selectable eGFP/BSD fusion gene.3 (b) Alternative blasticidin selectable FPs were generated in LeGO–G/BSD by replacing eGFP with any of the four FPs indicated (that is, Cerulean, Venus, dTomato and mCherry). (c) In a next step, the BSD gene in LeGO–G/BSD was substituted by each of the four alternative antibiotic-resistance genes indicated (that is, Zeo, Puro, Neo or Hygro, see main text). Abbreviations: ΔLTR, self-inactivating long-terminal repeat; Ψ, LV packaging signal; RRE, rev-responsive element; cPPT, central poly-purine tract; U6, murine U6 pol-III promoter (for shRNA); loxP, recognition sites of Cre recombinase; SFFV, spleen focus-forming virus U3 promoter; wPRE, Woodchuck hepatitis virus posttranscriptional regulatory element. (d) For proof of principle, the indicated combinations of FPs and antibiotic-resistance genes were generated and functionally tested in LeGO vectors.

Full size image

To analyze functional integrity of the different FP/BSD fusion proteins, myeloid K562 cells were transduced with the novel LeGO vectors LeGO–Cer/BSD, LeGO–V/BSD and LeGO–C/BSD. All constructs were readily expressed, but expression levels of fusion proteins were reduced by >50% (based on mean fluorescence intensities) when compared with the respective FPs alone (Figure 2a). Titers for all FP/resistance fusion and FP constructs were in the range of 9–19 × 106 ml–1 with no significant differences (data not shown).

Figure 2
figure 2

Fusion proteins of blasticidin-S deaminase (BSD) with different fluorescent proteins (FPs) are readily detectable by fluorescence-activated cell sorting (FACS) and facilitate efficient selection with blasticidin. (a) K562 cells transduced with lentiviral gene ontology (LeGO) vectors encoding Cerulean, Venus or mCherry are identified as clearly separated populations by FACS analysis. (b) Fusion of the different FP genes with the BSD gene results in a decrease in protein expression of approximately 60% as established based on mean fluorescence intensities, but fusion proteins are still readily detectable even in cells with only one vector copy.7, 8 (c) Importantly, application of 10 μg ml–1 blasticidin for 8 days results in almost pure populations of modified cells. Percentages of transduced cells and mean fluorescence intensities as measured by FACS are indicated in each plot. All FACS data were obtained 8 days after transduction.

Full size image

To assess functional activity of the BSD gene, K562 cells were transduced with each of the three constructs—LeGO–Cer/BSD, LeGO–V/BSD and LeGO–C/BSD—at low multiplicities of infection to avoid multi-copy insertions.7, 8 Subsequent treatment with 10 μg ml–1 blasticidin resulted in almost pure populations of transduced cells after 8 days (Figure 2c). Similar results were obtained for a large variety of other cell lines representing various tissues. As observed for eGFP/BSD and dTomato/BSD (see, Weber et al.3), fusion to the BSD-resistance gene prevented nuclear localization of all fusion FPs tested (not shown).

Antibiotic-mediated selection of cells transduced with bicistronic lentiviral gene ontology vectors

Next we analyzed whether in a bicistronic vector configuration (transgene of interest–internal ribosome entry site–drug-selectable FP) expression of the drug-selectable FP would still be sufficient to facilitate efficient selection of transduced cells. To address this question, we constructed bicistronic LeGO vectors co-expressing the transcription factors Olig2 with tdTomato/BSD and Nkx2.2 with eGFP/BSD, respectively. Olig2 and Nkx2.2 have important roles in the specification and maturation of the oligodendrocyte lineage, respectively.9 As shown in Figure 3, selection of transduced NIH3T3 cells with blasticidin resulted in almost pure populations of cells co-expressing Olig2 and tdTomato or Nkx2.2 and eGFP. In another experimental setting, LeGO vectors were constructed to co-express various drug-selectable FPs with different variants of the Herpes simplex virus thymidine kinase. Again, co-expression was highly efficient allowing both positive and negative selection of transduced cells.10

Figure 3
figure 3

Expression of the transcription factors Olig2 or Nkx2.2 (immunocytochemistry) in antibiotic-selected NIH3T3 transduced with bicistronic lentiviral gene ontology (LeGO) vectors. NIH3T3 cells were transduced with bicistronic LeGO vectors encoding the basic helix-loop-helix transcription factor Olig2 and a tdTomato/BSD fusion protein (a–c) or the homeodomain transcription factor Nkx2.2 and a eGFP/BSD fusion protein (d–f). After selection with blasticidin, the vast majority of cells in these cultures did co-express tdTomato (a) and Olig2 (b) or eGFP (d) and Nkx2.2 (e). Cultures were additionally stained with DAPI (c, f) to visualize the nuclei of all cells.

Full size image

Novel drug-selectable fluorescent proteins facilitate marking and selection of cells

To facilitate multi-gene analyses in single cells (for example, to analyse gene networks) additional bifunctional genes are needed to allow simultaneous drug-based selection of cells concurrently transduced with different LeGO vectors. Consequently, we next tested fusion constructs of FPs with four other selection markers—Neo, Hygro, Puro and Zeo. For proof of principle, these four resistance genes were fused to the eGFP gene and cloned into the basic LeGO-G vector thus creating LeGO–G/Neo, LeGO–G/Hygro, LeGO–G/Puro and LeGO–G/Zeo (Figure 1). Taken together, any FP and any antibiotic-resistance gene were tested in the context of at least one fusion gene (Figure 1d). On the basis of the modular design of the LeGO vectors further combinations could easily be constructed.

The novel vectors encoding eGFP in fusion with Neo, Hygro, Puro or Zeo as well as appropriate control vectors (LeGO–G, LeGO–G/BSD) were used to transduce K562 cells at one vector copy per transduced cell.7, 8 As depicted in Figure 4a, eGFP expression was readily detectable in transduced cells for all fusion constructs. However, whereas eGFP/Zeo and eGFP/Neo showed high expression levels (when compared with eGFP/BSD), transgene levels were significantly reduced for eGFP/Hygro and eGFP/Puro (Figure 4a). Similar results were obtained for 293T cells (not shown).

Figure 4
figure 4

Fluorescence-activated cell sorting (FACS) detection and functionality of novel fluorescent reporters conferring resistance to various antibiotics. (a) Fusion of eGFP to any of the five different antibiotic-resistance genes BSD, Zeo, Neo, Hygro or Puro results in decreased eGFP expression levels reflected by lower mean fluorescence intensities (representative analysis of K562 cells). Importantly, all fusion proteins remained unambiguously detectable even at low gene transfer rates. Moreover, for all fusion proteins tested, selection with the respective antibiotics resulted in almost pure populations of gene-modified cells after just 8 days (lower panel of FACS dot plots). Percentages of transduced cells and mean fluorescence intensities as measured by FACS are indicated in each plot. (b) Three of the antibiotic-resistance genes were codon-optimized to improve protein translation. As indicated by mean fluorescence intensities (MFI), codon-optimization (filled green areas) resulted in clearly improved expression levels of eGFP/Neo-opt and eGFP/Hygro-opt genes as compared with their parental genes (open areas). No effect on expression was observed for the codon-optimized Puro-opt gene. FACS data were acquired 10 days after transduction (performed in parallel for all vectors) in the absence of any selection. A full color version of this figure is available at the Gene Therapy journal online.

Full size image

Irrespective of their expression levels (as assessed by FACS) selection of cells expressing the various FP/resistance fusion proteins with their respective antibiotics was highly efficient. In fact, K562 cells transduced at low levels (⩽15%) with either of the four novel constructs (or LeGO–G/BSD as a control) were selected to almost pure populations after a culture period of only 8 days in the presence of the appropriate drug (Figure 4a).

It is noteworthy that as eGFP/BSD (see Weber et al.3), the eGFP/Hygro fusion protein showed no nuclear localization. In comparison, eGFP/Zeo was predominantly localized in the nucleus, whereas eGFP/Neo and eGFP/Puro, similar to eGFP alone, were evenly distributed throughout the entire cell (not shown).

Codon-optimized drug-selectable markers

In addition to low expression levels of some drug-selectable FPs (Figure 4a), we also observed slightly reduced titers (app. factor 2) of LeGO vectors encoding FP/resistance fusion genes. Both, low expression levels and low titers might be related to the length of the respective fusion genes, their primary coding sequences and/or the presence of transcriptional silencers.11 We therefore in silico analyzed sequences of all resistance markers with regard to human codon usage and the presence of potential aberrant splice sites and/or transcriptional silencers. On the basis of this analysis, we next synthesized codon-optimized versions of the three weakly expressed resistance genes Hygro, Neo and Puro, and used these variants to generate LeGO–G/Neo-opt, LeGO–G/Hygro-opt and LeGO–G/Puro-opt. Expression levels of original and codon-optimized LeGO constructs were compared in K562 cells containing only one vector copy per transduced cell. As shown in Figure 4b, codon-optimization resulted in increased expression levels of eGFP/Neo-opt and eGFP/Hygro-opt, but not eGFP/Puro-opt. Treatment of cells transduced with eGFP/Neo-opt, eGFP/Hygro-opt or eGFP/Puro-opt with the respective antibiotics proved normal functionality of all codon-optimized resistance genes (Figure 5, and not shown).

Figure 5
figure 5

Efficient selection of triple-transduced cells by simultaneous application of three antibiotics. (a) NIH3T3 cells were simultaneously transduced with the three vectors lentiviral gene ontology (LeGO)–Cer/BSD, LeGO–G/Neo-opt and LeGO–C/Zeo. Gene transfer rates were adjusted to approximately 64% for each individual vector, resulting in approximately 35% of triple-positive cells (represented by three-coloured circles). After transduction cells were selected for 9 days using the three respective antibiotics. (b) Fluorescence-activated cell sorting (FACS) analysis in the presence (filled areas) versus absence (open areas) of the three antibiotics was performed after selection. As evident, almost pure populations of cells expressing all three marker genes were obtained.

Full size image

Antibiotic-mediated selection of triple-transduced cells

We next tested the efficacy of our approach for simultaneously transducing cells with three LeGO vectors each encoding a different FP/resistance fusion gene and subsequently selecting cells, which co-express all three different transgenes. To this aim, NIH3T3 fibroblasts were transduced with the novel vectors LeGO–Cer/BSD, LeGO–G/Neo-opt and LeGO–C/Zeo with gene transfer rates of approximately 64% for each construct (Figure 5). This relatively high gene transfer rate ensures the generation of sufficient numbers of triple-positive cells (in this experiment approximately 35%). At the same time, the vast majority (approximately 90%) of transduced cells could statistically be expected to contain only 1–2 vector copies of each individual vector.8 Transduced cells were subsequently submitted to triple-antibiotic selection for 9 days. Almost pure populations of triple-transduced cells were obtained (Figure 5b) showing the principal usefulness of the novel drug-selectable FPs for multi-gene analysis experiments. Similar data were obtained for K562 and 293T cells.

Efficient selection of double-transduced mesenchymal and neural stem cells

We finally assessed performance of the novel LeGO vectors in different types of primary stem cells. First, we transduced mesenchymal stem cells (MSC, see, for example, Lange et al.12) simultaneously with two vectors (LeGO–G/Neo-opt and LeGO–C/Zeo). In accordance with our aim to ensure single-copy gene transfer for each individual vector, we obtained transduction efficiencies of 5 and 15% for LeGO–G/Neo-opt and LeGO–C/Zeo, respectively.7, 8 As expected, the number of double-positive cells was below 2% (Figure 6a, gate P2). Transduced cells cultivated in the absence of selective pressure (‘control’) showed a slight decrease in the percentage of double-positive cells (Figure 6b). In striking contrast, double selection with both G418 and zeocin for 10 days resulted in an almost pure population (>95%) of double-positive cells (Figure 6c).

Figure 6
figure 6

Efficient simultaneous selection of mesenchymal stem/stromal cells co-transduced with two different lentiviral gene ontology (LeGO) vectors. Human mesenchymal stem cells (MSCs) were co-transduced with LeGO–G/Neo-opt and LeGO–C/Zeo and thereafter selected by simultaneous application of 800 μg ml–1 G418 and 25 μg ml–1 zeocin for 10 days. Control-transduced MSC cultured in parallel with identical amounts of antibiotics were dead by day 4 of treatment. (a) Transduction efficiencies of 5 and 15% for LeGO–G/Neo-opt and LeGO–C/Zeo, respectively, resulted in approximately 1.5% double-positive cells (gate P2) as established by fluorescence-activated cell sorting (FACS). (b) Cultivation of cells in the absence of antibiotics resulted in a slight decrease in the percentage of double-positive cells after 2 weeks. (c) On the contrary, double selection with both G418 and zeocin resulted in an almost pure population (96.6%) of cells expressing both eGFP and mCherry. A full color or version of this figure is available at the Gene Therapy journal online.

Full size image

For efficient co-expression of transgenes in adherently propagated NSCs (see, for example, Conti et al.13) from the cerebral cortex of mouse embryos, we generated novel LeGO vectors encoding tdTomato/BSD or eGFP/Neo-opt under control of the chicken beta-actin (CAG) promoter (LeGO–CAG–T2/BSD and LeGO–CAG–G2/Neo-opt; Figure 7a).14 The use of tdTomato instead of dTomato is associated with stronger fluorescence signals, but reduced vector titers.

Figure 7
figure 7

Stable transgene expression in antibiotic-selected neural stem cells (NSCs) and their differentiated progeny after transduction with two different lentiviral gene ontology (LeGO) vectors. (a) Cultures of adherently propagated NSCs (refs 3, 13) were first transduced with LeGO–CAG–T2/BSD and modified cells were selected with blasticidin. NSCs were subsequently transduced with LeGO–CAG–G2/Neo-opt and further cultivated in the presence of blasticidin and G418. (b) Antibiotic-selected cultures consisted of cells co-expressing eGFP (a) and tdTomato (b; c represents an overlay of a and b). Astrocytic (d–f) or neuronal (g–i) differentiation of selected NSCs resulted in cultures containing glial fibrillary acidic protein (GFAP)-immunoreactive astrocytes (f) or β-tubulin III-immunoreactive neurons (i; some neurons are marked with arrows in g–i) co-expressing eGFP and tdTomato. Bar in i (for a–i): 50 μm.

Full size image

For both vectors, initial transduction efficiencies were low. However, pure populations of tdTomato/eGFP-positive NSCs could easily be derived by antibiotic selection (Figure 7b). Furthermore, subsequent astrocytic or neuronal differentiation of engineered NSCs resulted in cultures containing tdTomato/eGFP-positive astrocytes or nerve cells, respectively (Figure 7b). Together, these data further show the usefulness of LeGO vectors containing drug-selectable FPs for gene analysis in comparatively demanding cell populations.

Discussion

After the ‘building blocks’ principle, we have developed a series of novel LeGO vectors equipped with drug-selectable FPs. LeGO vectors combine the advantages of LV vectors, such as efficient transduction of various target cells,1, 2 with a high degree of flexibility, which is based on their modular design.3 In fact, different LeGO vectors for cell marking, ectopic gene expression as well as short hairpin RNA-mediated gene knock down are available with multiple different fluorescent markers, which makes them ideally suited for multi-gene analysis in single cells.

Despite completion of the human genome project, in many cell types surprisingly little is known regarding the functions of individual genes and particular of gene networks. Even in hematopoietic stem cells, ‘the most thoroughly characterized type of adult stem cells, … the signalling mechanisms regulating fate events, such as self-renewal and differentiation, have remained elusive’.15 To address this problem, a vector tool dedicated to simultaneous multi-gene analysis should be of special interest. It is noteworthy that standard LeGO vectors could be expected to be particularly suited for gene analysis in hematopoietic cells because they contain an spleen focus-forming virus promoter ensuring robust expression of transgenes in this cell type.3 Indeed, recently LeGO vectors have successfully been used in both murine and human HSC for short hairpin RNA-mediated transcriptional silencing.4, 16

In this study, we have shown that LeGO vectors facilitate efficient gene expression in (human) MSCs, which represent a highly relevant target.12, 17 In addition, the novel LeGO vectors containing the CAG promoter14 have widened the field of potential applications. As shown in this work, they facilitate efficient transgene expression in NSCs and their differentiated progeny. Potential applications of these novel LeGO vectors include the elucidation of pathways involved in the normal differentiation or in malignant transformation of NSCs.18

Another important application of integrating vectors is gene marking, which is considered as one of the most successful applications of cell and gene therapy.19 Again, standard LeGO vectors should be particularly useful for analyzing hematopoiesis. Indeed, unambiguous genetic marking may become very important in the context of single-cell imaging studies20 and for the analysis of novel ex vivo models of definitive hematopoiesis.21 The set of already available LeGO vectors, together with the possibility to rapidly include additional markers (based on their modular structure) make them ideal tools for simultaneous marking of different cell populations. This will, for instance, allow studying cell–cell interactions or the fate of different cell types or even clones within given organ structures.

Introduction of multiple novel drug-selectable FPs represents a significant expansion of the previous set of LeGO vectors and opens the way for additional applications. In fact, these markers combine the established advantages of fluorescent markers (that is, easy detection by FACS and fluorescence microscopy) with those of drug-resistance genes, namely high-efficiency selection in closed systems without the need of special equipment. This study shows the functionality of a set of drug-selectable FPs in various cell populations, including MSCs as well as NSCs and their differentiated progeny. It is noteworthy that these drug-selectable FPs could easily be included in any eukaryotic vector system and should therefore be of great interest for many researchers working on gene analysis. We suppose that on-line efficacy control (for example, by FACS or fluorescence microscopy) of antibiotic selection would be an interesting application of the novel fluorescent antibiotic-resistance proteins, which might help optimizing selection protocols. It might also be noted that the strategy for generating drug-selectable FPs suggested here could easily be adapted to other fluorescent markers by introducing the BsrGI site into the 3′-end of the marker gene.

Importantly, we have also found that all of the drug-selectable fusion proteins show decreased fluorescence intensities as compared with the corresponding fluorescent markers. This needs to be taken into account, when those markers are to be used in the context of more complex expression vectors (for example, in bi- or multicistronic settings) by choosing optimal marker gene combinations. Also detection of orange and red FPs could be improved using adapted cytometer settings, particularly a green laser for optimal excitation. Otherwise measurements often result in an underestimation of expression levels. As we have shown here, even quite low expression levels of drug-selectable FPs (as estimated by FACS) after single-copy gene transfer allowed efficient selection of cells transduced with either mono- or bicistronic vectors (expressing different types of genes-of-interest).

Finally, we have introduced three novel codon-optimized antibiotic-resistance genes. Although all three were functional, two of them, Neo-opt and Hygro-opt showed clearly improved expression characteristics as compared with their non-modified ‘ancestors’. We suggest that replacement of ‘old-mannered’ resistance genes by the novel variants may significantly improve performance of various vectors systems by avoiding deleterious effects of the gene sequence on transgene expression and/or vector titer.11

In conclusion, we suppose that each of the different novel tools presented here (novel LeGO vectors, drug-selectable FPs and codon-optimized antibiotic-resistance genes) may be of great value for researchers working in the fields of cell marking and/or functional gene analysis.

Materials and methods

Cloning procedures and sequences

Details on cloning procedures of LeGO-vectors and sources of FPs were provided earlier.3 Complete vector sequences and maps are available at www.LentiGO-Vectors.de.

The antibiotic-resistance genes were derived from the following plasmids:

BSD::

Blasticidin-S deaminase (bsd), pcDNA6.2/EmGFP-Bsd/V5-DEST, Invitrogen (Karlsruhe, Germany).

Zeo::

Stretoalloteichus hindustanus bleomycin gene (Sh ble), pcDNA4/HisMax-C, Invitrogen.

Puro::

Puromycin N-acetyl-transferase (pac), pLKO.1–puro, Sigma (Munich, Germany).

Hygro::

Hygromycin phosphotransferase (hph), pVITRO4-mcs, Invivogen (San Diego, CA).

Neo::

Aminoglycoside 3′-phosphotransferase (APH), MESV-X-neo (see, for example, Taoudi et al.22), kindly provided by Carol Stocking (HPI, Hamburg, Germany).

Antibiotic-resistance genes were amplified by PCR to introduce BsrGI sites at both ends and subsequently cloned in frame into the unique BsrGI site of the FP complementary DNAs within the LeGO-vectors. The primers shown in Table 1 were used.

Table 1 Primers used for cloning of antibiotic-resistance genes
Full size table

Codon-optimized variants of Puro, Hygro and Neo complementary DNAs without alterations of the amino acid sequence were designed and synthesized by GeneArt (Regensburg, Germany).

The CAG promoter was used as a SalI/XhoI fragment, derived from pPyCAGiP-HcRed (see Chambers et al. 23), a kind gift of Frank Edenhofer (Life & Brain, University Bonn, Germany).

Cell culture and virus production

Cell culture was performed according to earlier published protocols.3, 13 Detailed protocols on LV vector production and titration are available at www.LentiGO-Vectors.de.

Transduction and selection of K562 cells using blasticidin

In parallel settings, K562 cells (ATCC: CCL-243) were transduced with one of six different LeGO-vectors, expressing the FPs Cerulean, Venus or mCherry alone (LeGO–Cer, LeGO–V, LeGO–C) or in fusion with the blasticidin-resistance protein (LeGO–Cer/BSD, LeGO–V/BSD, LeGO–C/BSD). Selection was carried out using 10 μg ml–1 blasticidin (Carl Roth, Karlsruhe, Germany), starting with 1.5 × 104 cells in 1 ml medium per well of a 24-well plate. During selection cells were kept at low density and medium was exchanged every 1 to 2 days. Control cells expressing the FP without a fused-resistance protein died within 1 week. After 8 days of selection cells were analyzed by FACS.

Transduction and selection of NIH3T3 cells with bicistronic lentiviral gene ontology vectors

Bicistronic LeGO vectors3 use the encephalomyocarditis virus internal ribosome entry site for efficient co-expression of the transgene-of-interest and a drug-selectable FP. Using standard cloning approaches, we generated LeGO vectors encoding the basic helix-loop-helix transcription factor Olig2 or the homeodomain transcription factor Nkx2.2 and a tdTomato/BSD or eGFP/BSD fusion protein behind the internal ribosome entry site, respectively. Vectors were used to transduce NIH3T3 fibroblasts, and positive cells were selected by cultivation in a medium containing 8 μg ml–1 blasticidin. Expression of the transcription factors in blasticidin-selected cultures was analyzed by immunocytochemistry after a culture period of 3 weeks. Cultures were fixed in phosphate-buffered saline containing 4% paraformaldehyde and blocked in phosphate-buffered saline containing 0.1% bovine serum albumine and 0.1% Triton X-100. Subsequently, cultures were incubated with antibodies to Olig2 (R&D Systems, Wiesbaden-Nordenstadt, Germany) or Nkx2.2 (Developmental Studies Hybridoma Bank, University of Iowa, Iowa City, IA, USA). Primary antibodies were detected with fluorochrome-conjugated secondary antibodies (Dianova, Hamburg, Germany), and cell nuclei were visualized by staining the cultures with Hoechst 33258 (Sigma, Taufkirchen, Germany).

Transduction and selection of K562 cells with five different antibiotics

Totally six different LeGO-vectors either expressing eGFP alone (LeGO–G) or fusions of eGFP with the antibiotic-resistance proteins BSD, Zeo, Puro, Hygro and Neo (LeGO–G/BSD, LeGO–G/Zeo, LeGO–G/Puro, LeGO–G/Hygro and LeGO–G/Neo) were used to transduce K562 cells.

For the 10 different selection settings (control vector LeGO–G with the various antibiotics and the five novel LeGO vectors with the respective antibiotics, see Table 2), 3 × 104 cells were plated per well of a 24-well plate in 1 ml medium. Concentrations of different antibiotics are listed in Table 2. Medium was exchanged every 1 to 2 days and cells were kept at low density. Control cells expressing the FP without a fused-resistance protein died within 1 week. After 8 days of selection, selected cells were analyzed by FACS.

Table 2 Experimental groups to verify functional integrity of novel drug-selectable fluorescent proteins
Full size table

Triple selection of triple-transduced NIH3T3 cells

NIH3T3 cells (ATCC: CRL-1658) were simultaneously transduced with three LeGO-vectors, expressing fusions of different FPs and different resistance proteins (LeGO–Cer/BSD, LeGO–G/Neo-opt and LeGO–C/Zeo). We aimed for relatively high transduction rates to ensure transduction of a reasonable proportion of cells by all three vectors. For this purpose, 5 × 104 cells were resuspended in 500 μl medium per well in a 24-well plate in the presence of 8 μg ml–1 polybrene. For transduction, unconcentrated supernatant containing LV particles (9.1 μl LeGO–Cer/BSD, 6.0 μl LeGO–G/Neo-opt and 8.0 μl LeGO–C/Zeo) were added and the 24-well plate was centrifuged for 1 h at 1000 × g and 24 °C.

Cells simultaneously transduced with three LeGO-vectors encoding the different FP genes not fused to the respective resistance genes (LeGO–Cer, LeGO–G and LeGO–C) served as a control.

For selection 105 cells were resuspended in 1 ml medium per well of a 6-well plate. The three antibiotics were simultaneously applied using the following concentrations: 40 μg ml–1 blasticidin, 1600 μg ml–1 G418 and 25 μg ml–1 zeocin. Medium was exchanged every 1 to 2 days. Cells were kept at low density and were re-plated every 3 days to avoid colony formation. Control cells expressing the FPs without a fused-resistance protein died within 3 days. After 9 days of selection, selected cells were analyzed by FACS.

Transduction and selection of human mesenchymal stem/stromal cells

We used adherently growing human MSC/stromal cell cultures isolated and expanded in the absence of fetal calf serum as described.12 Phenotypic and functional identity of MSC had been confirmed using established assays.12 MSCs were simultaneously transduced by spinoculation for 1 h at 1000 × g in parallel with the two vectors LeGO–G/Neo-opt and LeGO–C/Zeo using multiplicities of infection of 2.5 for each vector. After 2 days, microscopic analysis showed a too low transduction rate and another transduction using a 10 times higher multiplicities of infection for each vector were carried out. Selection of double-positive cells by simultaneous application of 800 μg ml–1 G418 and 25 μg ml–1 zeocin was carried out for 10 days. FACS analyses for eGFP and mCherry expression were performed at the beginning and 3 days after completion of selection. Control cells transduced with two corresponding vectors without resistance genes, LeGO–G and LeGO–C, were cultured in parallel in the presence of the same concentrations of antibiotics. Control cells were dead by day 4 of treatment.

Transduction and selection of neural stem cells

Adherently growing NSC cultures from the cerebral cortex of mouse embryos were derived as described elsewhere.13 In brief, we first prepared neurosphere cultures from the cerebral cortex of 14 days old mouse embryos according to established protocols (see, for example, Ader et al.24). After 2–3 passages in Dulbecco's modied Eagle's medium/F12 medium containing 1% N2 (Invitrogen), 10 ng ml–1 epidermal growth factor and 10 ng ml–1 fibroblast growth factor-2 (both from TEBU, Offenbach, Germany), neurospheres were enzymatically dissociated with Accutase (PAA Laboratories, Coelbe, Germany). Cells were plated into gelatine-coated tissue culture flasks and further cultivated in NS-A medium (Euroclone, Pero, Italy) containing 10 ng ml–1 epidermal growth factor, 10 ng ml–1 fibroblast growth factor-2 and 1% modified N2.25

For transductions, NSCs were plated into 6-well plates coated with poly-L-ornithine and 1% Matrigel (Becton Dickinson, Heidelberg, Germany), and spinoculated with LeGO–CAG–T2/BSD in the presence of 8 μg ml–1 polybrene.3 Positive cells were selected by cultivation in NS-A medium containing 1% modified N2, 10 ng ml–1 epidermal growth factor, 10 ng ml–1 fibroblast growth factor-2, 1% B27 (Invitrogen) and 4 μg ml–1 blasticidin. Cultures were again expanded, then transduced with LeGO–CAG–G2/Neo-opt and subsequently maintained in expansion medium containing 4 μg ml–1 blasticidin and 200 μg ml–1 G418 to select for cells co-expressing eGFP and tdTomato. To differentiate eGFP/tdTomato-positive NSCs into astrocytes, cells were maintained for 5 days in NS-A medium containing 2% B27 and 1% fetal calf serum and lacking mitogens and N2. For neuronal differentiation, cells were first cultivated in NS-A containing 1% N2, 2% B27 and 5 ng ml–1 fibroblast growth factor-2 for 5 days, followed by a further cultivation period of 5 days in a 1:1 mixture of NS-A and Neurobasal (Invitrogen) medium containing 0.25% N2 and 2% B27. Non-differentiated and differentiated cultures were fixed with 4% paraformaldehyde in phosphate-buffered saline. Differentiated cultures were incubated with rabbit polyclonal glial fibrillary acidic protein (1:500; Dako, Glostrup, Denmark) or rabbit polyclonal β-tubulin III antibodies (1:200; Convance, Berkeley, CA, USA) to label astrocytes or neurons, respectively. Primary antibodies were detected using Cy5-conjugated secondary antibodies (1:200; Jackson ImmunoResearch, West Grove, PA, USA). Cultures were incubated with 4′,6-diamidino-2-phenylindole dihydrochloride hydrate (DAPI) (Sigma) to label the nuclei, and analyzed using a Olympus FluoView 1000 (Olympus, Hamburg, Germany) confocal laser scanning microscope.

Flow cytometry

Flow cytometry was performed on FACSCantoII (Becton Dickinson) using the 407 nm and 488 nm lasers. Data were analyzed using FACSDiva software (Becton Dickinson) and processed using the figure-improvement toolbox Diva-Fit.26