Transcriptional regulation provides an ideal target for therapeutic intervention. As such, tools for studying transcriptional modulators of disease genes will help to facilitate the development of novel therapeutics 1. Cell lines have been used to study the expression of specific genes involved in disease development or at signal transduction checkpoints, and are currently a front-line approach for early-stage drug discovery. A number of indirect techniques are available to assess gene transcription in cells including ELISA and gene arrays or quantitative PCR for measuring the gene transcript levels. However, these methods are time consuming, resource intensive and/or do not directly assess the transcriptional activity of an endogenous promoter. Moreover, they are not amenable to high-throughput screening (HTS) for efficient detection of drug-induced changes in disease gene expression.

Cell-based gene reporter assay systems were developed as an alternative system amenable to HTS over 10 years ago, and have been widely used to study transcription and gene regulation. Specifically, linking easily detectable reporter genes – such as luciferase, β-galactosidase or green fluorescent protein – to defined gene promoters and regulatory elements has resulted in the production of numerous reporter vectors. Transient transfection of such reporter vectors into cultured cells and quantitative analysis of the reporter gene product is a fast and efficient way to study disease gene expression. Moreover, the establishment of cell lines containing random stable integrants has made possible the development of cell-based reporter assays 2, which have now been successfully scaled-up for HTS following advances in robotics and fluorescence/luminescence plate-reader technologies 34. Recently, a novel reporter system was developed in which Flp recombinase is used to produce flippase recognition target (FRT) single site-specific integration of a reporter gene construct at a transcriptionally-active genomic locus in cultured cells 5. This approach has several advantages over randomly integrated reporter constructs including single copy construct integration and a single chromatin context within which the effects of promoter mutations or single nucleotide polymorphisms (SNPs) on gene expression can be studied 5. Moreover, this reporter system has been used to screen small molecules for inhibition of the pro-inflammatory cytokine, tumor necrosis factor (TNF) 6. Although randomly integrated and FRT single site-specific reporters are presumed to reflect endogenous regulation of the disease gene, this is a questionable assumption given the unknown epigenetic influences of chromatin structure on gene transcription along with missing genetic elements that regulate gene expression at the endogenous locus. To this end, optimal systems would utilize gene-targeted reporters controlled by endogenous regulatory sequences and governed by an inherited epigenetic program unique to a given disease gene locus. Although gene targeting in mouse embryonic stem cells makes it possible to precisely integrate exogenous DNA sequence into a predetermined 'target' gene locus 7, such systems have been much less effective in somatic cells. An alternative approach, utilizing single-stranded recombinant adeno-associated virus (rAAV) to promote homologous recombination between the targeting construct and the chromosome 891011 has been widely applied to genetically modify endogenous genes by insertion, deletion/replacement, and point mutation 11121314. The efficiency of gene targeting using single-stranded rAAV vectors is also much higher than that observed with adenovirus- or retrovirus-based vector systems 13. Self-complementary rAAV (scAAV) vectors have been shown to promote more efficient viral transduction than single-stranded rAAV vectors both in vitro and in vivo 15. However, these double-stranded vectors do not appear to contribute to the gene targeting reaction 1316.

The TNF-α gene maps to chromosome 6p21.3, contains four exons, and spans approximately 3 kb of DNA in human cells 1718. TNF-α gene expression is cell type-specific and induced by a wide variety of stimuli such as phorbol 12-myristate 13-acetate (PMA) and lipopolysaccharide 192021. The TNF-α protein is a multifunctional cytokine, and is involved in the regulation of a broad spectrum of biological processes 222324. The TNF-α gene appears to be silenced in HeLa cells, as evidenced by undetectable levels of mRNA by northern blot and protein by ELISA 25. In the present study, we sought to engineer a HeLa cell line containing a targeted luciferase reporter in exon 1 of the TNF-α gene. We also sought to compare the patterns of Renilla luciferase (R-Luc) induction with endogenous TNF-α mRNA transcription between targeted and non-targeted cell lines in response to drug treatment. The production of a TNF-α gene-targeted reporter cell line will provide a sensitive and more predictive analytical tool for identifying molecules that modulate TNF-α gene transcription.

An rAAV targeting vector (AV.TNF-RL.targ) was generated to facilitate fusion of the Renilla luciferase (R-Luc) reporter gene to the TNF-α gene locus in HeLa cells (Figure 1a). The vector harbors a 2.1 kb genomic DNA fragment from the TNF-α locus, which was split into left and right homologous arms by the insertion of a R-Luc cDNA and loxP sites that flank a phosphoglycerate kinase (PGK) promoter-driven Zeocin expression cassette. The insertion site in exon 1 is immediately downstream of the TNF-α start codon, fusing the R-Luc gene in-frame to the TNF-α transcript. Since the left homologous arm of the targeting vector encodes the TNF-α core promoter and contains other regulatory elements necessary for initiation of transcription, we are able to compare reporter expression profiles between the targeted and non-targeted cell lines, the latter of which are derived from the random integration of AV.TNF-RL.targ in HeLa cells. A Zeocin-resistant gene serves as a selectable marker for clonal expansion of cells in which the rAAV genome has been stably integrated. Enrichment of stably integrated cells is required for this type of insertional gene targeting.

HeLa cells were infected with AV.TNF-RL.targ (Figure 1a) and re-plated for clonal expansion under Zeocin selection. Zeocin-resistant colonies were picked and transferred to replicate 96-well plates. Cells in replica plates were lysed for PCR screening with two sets of primers, which hybridize to sequences outside the right and left targeting arms and inside the exogenous insert. Clone #28 was identified as a positive targeted clone, from 192 clones screened, and its left-side PCR product was cloned into the pBlunt4PCR vector (Invitrogen) for sequence confirmation. Sequencing results revealed the presence of both the non-virus-derived flanking sequences and the expected in-frame fusion of the R-Luc cDNA in the TNF-α gene (data not shown). This positive clone (designated as Tg#28zeo^R) was expanded and the genomic DNA was analyzed by digestion with several restriction enzymes (Figure 1b; lanes 2, 4 and 6). Genomic DNA from parental HeLa cells was used for comparison (Figure 1b; lanes 1, 3 and 5). The Southern blot was probed with TNF-α left arm-homologous sequences. The additional bands observed in the digested Tg#28zeo^R samples (Figure 1b; lanes 2, 4 and 6) are indicative of targeted insertion of the R-Luc cDNA at the TNF-α gene locus. No additional random vector integration was observed (Figure 1b).

The exogenous PGK promoter and transcription of the Zeocin gene could affect the transcriptional activity of the targeted TNF-α gene. To eliminate any possible artificial induction in R-Luc activity, the selection cassette was removed from the targeted intermediate, Tg#28zeo^R (Figure 1c). Flanked by a pair of LoxP sites, the PGK-Zeocin cassette can be easily excised from the targeting AAV genome. Cre recombinase-mediated excision was used to remove this selection cassette from the targeted Tg#28zeo^R line and also from the non-targeted cell lines that harbor random integrations of the targeting virus. A recombinant adenoviral vector, Ad.Cre, was used to deliver Cre recombinase to the cells. Southern blot analysis with probes for TNF-α and PGK/Zeo demonstrated that Ad.Cre infection resulted in loss of the selection cassette from the targeted intermediate, producing the final TNF-α reporter cell line, Tg#28zeo^- (Figure 1c and 1d).

Individual clones expanded from a single cell were isolated from the Zeocin-sensitive cell pool by limited dilution. Five independent lines were randomly selected and basal levels of R-Luc expression among these was compared. No apparent differences were observed between individual Tg#28zeo^- lines, and the expression levels were very similar to that in the original cell pool (Figure 1e). However, basal R-Luc activity in the targeted intermediate was more than 300-fold higher than in the clones lacking the Zeocin selection cassette (Figure 1e). Thus, as predicted, this selection marker enhanced transcription from the TNF-α gene locus, arguing that R-Luc activity in Tg#28zeo^- cells should more closely reflect endogenous TNF-α gene regulation than reporter activity in Tg#28zeo^R cells.

Current TNF-α reporter vectors contain only about 1.0 kb of core promoter located upstream of the TNF-α gene (Invitrogen Corp., Carlsbad, CA; Panomics, Inc., Fremont, CA). In addition, these plasmid-based TNF-α/reporter constructs are randomly inserted into the host cell genome following transfection. In theory, the fidelity of TNF-α gene expression in these randomly integrated reporter cell lines may be influenced by missing regulatory sequences (enhancer and/or silencer) not part of the 1.0 kb core promoter of the TNF-α gene 202627. Indeed, recent studies have demonstrated that the regulation of TNF-α expression involves distal enhancers located over a 12 kb region, and that these enhancers interact to form a novel double-loop chromatin configuration. This structure circularizes the TNF-α gene and thereby facilitates transcription 282930. Moreover, epigenetic regulation is important for the control of TNF-α transcription 31 and key epigenetic modifications at the TNF-α locus may be missing from the regulatory elements governing reporter expression in the randomly integrated reporter cell lines. In addition to the lack of intact endogenous regulatory elements and epigenetic modifications associated with the self-contained TNF-α core promoter/reporter gene, expression of the randomly integrated reporters is very likely to be influenced by both genetic and epigenetic features associated with the insertion site, in a very unpredictable manner. Thus, we hypothesized that the TNF-α gene would be an ideal platform to test whether targeted reporter expression more closely reflects endogenous gene expression patterns than randomly integrated reporters.

To test this hypothesis, we isolated 18 non-targeted TNF-α/R-Luc reporter clones in which the PGK-Zeocin cassette had been excised by Ad.Cre infection. We then compared basal R-Luc activity in the targeted and non-targeted reporter lines. Activity varied widely among the non-targeted lines relative to that in the targeted line. Four non-targeted (nTg) lines representing the range of basal R-Luc activity were selected for additional comparison to the Tg#28zeo^- targeted clone (Tg) (Figure 2a). TNF-α mRNA was purified from the reporter lines and quantified by TaqMan PCR (Figure 2a). Basal TNF-α mRNA levels were similar in all of the nTg lines, whereas the level in the Tg line was somewhat lower. Reduced expression in the Tg line was most likely due to the disruption of one TNF-α allele as a consequence of the in-frame insertion of the R-Luc cDNA. Tg and nTg cell lines were then treated with known inducers of TNF-α expression. Drugs representing different activation pathways were used: the protein kinase C activator phorbol 12-myristate 13-acetate (PMA) 3233; the DNA topoisomerase II inhibitor doxorubicin (DOX) 3435; the histone deacetylase inhibitor trichostatin A (TSA) 3136; and the DNA methylation inhibitor 5-aza-2'-deoxycitidine (Aza-dC) 31. The extent of drug-induced reporter activity varied dramatically among the nTg cell lines tested – ranging from little (nTg2) or no induction (nTg1 and nTg3) to an induction profile similar to that in the Tg cell line (nTg4) (Figure 2b). However, even where such similarities existed, differences between the Tg and nTg4 lines were apparent. Drug-induced changes in TNF-α and R-Luc mRNA expression in the Tg line were then quantitated by TaqMan PCR and compared with R-Luc activity in the Tg line (Figure 2c). The induced patterns of both the TNF-α and R-Luc mRNAs closely reflected the patterns of R-Luc protein activity following drug treatment, indicating that reporter expression accurately reflects endogenous TNF-α gene expression in the Tg clone.

Throughout the approximately one-year study period, the basal level of R-Luc activity remained low and the levels of PMA- and TSA-induced reporter expression remained constant in the targeted cell line (data not shown). However, we did not conduct a systematic analysis of drug-induced reporter activity at regular intervals during this time. Nevertheless, while the long-term affects from integration of reporter gene sequences are unknown, our data do not indicate a progressive transcriptional silencing of the R-Luc gene.

We next probed potential regulatory differences between Tg and the nTg cell lines further by investigating their responsiveness to 5,6-Dimethylxanthenone-4-acetic acid (DMXAA) 37. This is an agent that induces vascular permeability and tumor cell death in human solid tumors by activating TNF-α transcription 3839 and is currently in Phase II clinical trials 40. At a fixed drug concentration, DMXAA-induced R-Luc activity was observed in the Tg cell line but not in any of the tested nTg cell lines (Figure 3a). In DMXAA dose-response studies, R-Luc activity was induced in the Tg line by as much as ~10-fold, while induction in the nTg4 line (which was the nTg line that most closely mirrored Tg line responses by other drugs, Figure 2b) was insignificant (Figure 3b). This differential drug-based induction was not due to cell line-dependent differences in cellular toxicity (Figure 3c). Moreover, little to no difference was observed when comparing the up-regulation of TNF-α mRNA following DMXAA treatment in the Tg and nTg4 lines (Figure 3d). These data suggest that the 1.0 kb TNF-α core promoter region does not encode the DMXAA response element(s).

Anthracycline antibiotics are also known activators of TNF-α promoter transcription. Dose-response studies with four closely related anthracycline antibiotics demonstrated a pronounced up-regulation of R-Luc activity in the Tg cell line at a drug concentration of 1 μM (Figure 4a). Anthracycline exposure did not appear to significantly reduce cell viability in the Tg line at this drug concentration (Figure 4b). Therefore, both Tg and nTg4 cell lines were treated with anthracyclines at 1 μM drug and assayed for R-Luc activity. Differential R-Luc activity was evident between these cell lines, most notably following Idarubicin treatment (Figure 4c). Indeed, Idarubicin induced R-Luc activity ~300-fold in the Tg line, but only ~50-fold in the nTg4 line – representing a 6-fold difference in induction between the Tg and the nTg4 cell lines. Differential induction between Tg and nTg4 cell lines was also observed following treatment with daunorubicin, doxorubicin and epirubicin. Again, these differences were not due to differences in anthracycline-induced cell death in the Tg and nTg4 cell lines (Figure 4d). Rather, we attribute the differences in R-Luc reporter activity to unique genetic and/or epigenetic features of the endogenous TNF-α gene locus. We also conclude that targeted reporter cell lines may be superior tools for screening drugs that modulate the transcriptional activity of target genes. The use of such cell lines with biophotonic imaging as presented here (Figure 4a and 4b) may be extremely useful for multi-parameter HTS to identify novel therapeutics.

We conclude that gene-targeted reporter cell lines may more accurately index endogenous gene expression to facilitate predictive cell-based screening for drug discovery.

ZY contributed to the design of the study, the collection, analysis and interpretation of data and preparation of the manuscript. DL participated in the collection and analysis of data. JFE and GHL contributed to the conception and design of the study, analysis and interpretation of data and manuscript preparation. All authors read and approved the final manuscript.