|Home | About | Journals | Submit | Contact Us | Français|
Conceived and designed the experiments: LS DS. Performed the experiments: LS DS. Analyzed the data: LS DS. Contributed reagents/materials/analysis tools: LS DS. Wrote the paper: LS DS.
Read-through transcription from the adjacent E1a gene region is required for wild-type (wt) activity of the downstream adenovirus E1b promoter early after infection (read-through activation). However, whether a cellular chromosomal template can support read-through activation is not known. To address this issue, read-through activation was evaluated in the context of stably expressed templates in transfected cells. Inhibition of read-through transcription by insertion of a transcription termination sequence between the E1a and E1b promoters reduced downstream gene expression from stably integrated templates. The results indicate that the mechanism of read-through activation does not depend on the structure of early adenovirus nucleoprotein complexes, a structure that is likely to be different from that of cellular chromatin. Accordingly, this regulatory interaction could participate in the coordinated control of the expression of closely linked cellular genes.
In higher eukaryotic genomes, functionally or developmentally related transcription units are often arranged in groups –. Sometimes, such gene arrangements result in cis-acting transcriptional interactions between the genes in a cluster –. One well recognized cis-acting transcriptional interaction is transcriptional interference, the suppressive influence of one active transcriptional unit on another linked unit. Transcriptional interference has been described in a variety of experimental systems –.
Transcription from an upstream promoter also can activate a downstream promoter. The E1a and E1b genes of adenovirus 5 are tightly linked. Part of E1a exon 2, including amino acid coding sequences and the 3′ untranslated region, overlaps the E1b promoter region – (Figure 1A). Primary transcripts initiated from the E1a promoter invade the E1b promoter and coding region, and these read-through transcripts are processed to produce E1a mRNA , . Only transcripts that originate from the E1b promoter are precursors of E1b mRNA , . Artificial termination of read-through transcription from the E1a promoter by insertion of ectopic transcription termination sequences (GGT: globin gene termination sequence) dramatically reduces early E1b gene expression in cis –. Since point mutations that inactivate the transcription termination function of GGT restore downstream promoter activity, the ability to block read-through transcription is the only property of GGT that is required for inhibition , . Therefore, read-through transcription is required for wild-type (wt) activity of the E1b promoter early after infection. There is evidence that the mechanism of the interaction depends upon a cis-dominant property of the early viral template  but the effect is “local” rather than global .
The mechanism of read-through activation could depend on the structure of early adenovirus nucleoprotein, which is likely to be different from that of cellular chromatin. Alternatively, read-through activation might not require early viral chromosome structure, allowing adjacent genes to take advantage of this interaction for coordinated expression in the context of cellular chromatin. To explore this possibility, read-through activation from reporter constructs that retained the basic genetic organization of the E1a-E1b region from the viral genome was evaluated in the context of templates stably integrated into the cellular genome.
Read-through activation early after infection is readily observed with recombinant adenoviruses that have the E1b coding region replaced by the luc reporter gene . In these viruses, insertion of the mouse β-globin gene transcription termination sequence GGT between E1a and E1b strongly reduces early expression of the downstream E1b-luc gene, whereas inactivation of the termination function of the inserted sequence restores E1b promoter function.
To determine whether read-through activation affected gene expression in chromosomal DNA copies stably integrated in the cellular genome, we constructed plasmids with regulatory sequence arrangements identical to those in the virus but with coding region gene replacements to facilitate selection and analysis of cell lines. The E1a coding region was replaced with a selectable marker gene (neo, G418 resistance) and the gfp or luc reporter genes were substituted for the E1b coding region. This strategy allowed read-through activation in G418-resistant cell lines to be scored readily by evaluating fluorescence intensity or luciferase production. To evaluate the read-through requirement for reporter gene expression, plasmids contained GGT (pNeoGGTE1bGFP and pNeoGGTE1bLuc), GGT inactivated by a double set of point mutations (DPM, pNeoDPME1bGFP and pNeoDPME1bLuc), or no insertion (pNeoE1bGFP and pNeoE1bLuc) between the E1a-neo and E1b-reporter genes (Fig. 1B).
We considered targeting insertions to a specific site for these experiments. However, whether any particular site contained cis-dominant elements or properties that might affect read-through activation could not be predicted. On the other hand, it was likely that the potential for cis-dominant effects of the integration site to confound the analysis would be revealed by comparing the results from either pooled or individual cell clones with untargeted integration sites. Accordingly, untargeted clones were isolated and both pooled and individual clones were analyzed. Also, it was possible that cis-acting effects of a particular selection or reporter gene would interfere with read-through activation. To reduce the possibility that such interference would compromise the analysis, the plan was to use two selection genes, as well as the two reporter genes. However, attempts to isolate puromycin-resistant HeLa cells from E1a promoter-driven constructs analogous to the plasmids shown in Fig. 1B were unsuccessful (data not shown).
To obtain integrated sequence arrangements in the untargeted clones as similar as possible to that of the E1 gene cluster in the adenovirus genome, we sought conditions for DNA introduction predicted to favor low copy number integrations. For electroporated DNA, parameters can be adjusted to produce integration of from one to about 20 copies of plasmid DNA –, as opposed to a large amount of genomic DNA in each cell . Similar information is not available for lipofection. For three different experiments, companion G418-resistant cell lines expressing high (+) or undetectable (−) levels of GFP after introduction of the pNeoGGTE1bGFP plasmid were analyzed for the pattern of integrated plasmid DNA. The results (Fig. 2) show that we were successful in obtaining cells lines with low copy number integrations. None of the cell lines had more than a few copies of integrated plasmid. There was no relationship between GFP expression and a particular integration pattern. GFP-expressing cell lines had both single copy (B+, C+) and tandem copy (A+) insertions. Likewise, GFP-negative lines were single copy (B−) or multi-copy (A−, C−). Cell line B- may not have retained a viable copy of the GFP gene. Also, at least in this small sample, there seemed to be no advantage to be gained by selecting clones of a particular integration class to standardize the subsequent analysis.
In the first set of experiments, GFP expression was evaluated quantitatively either by visual scoring of individual G418-resistant colonies (Table 1) or flow cytometry of pooled cell cultures (Table 2, representative histograms of GFP-expressing cells in the cultures are shown in Fig. 3). For the latter, G418 was omitted during the plating to avoid selective amplification of any of the original clones and to maintain the GFP profile of the population. By either method, GFP expression was inhibited by insertion of GGT and expression was restored by inactivation of the transcription termination sequence (DPM, Tables 1 and and2,2, Fig. 3). Similar results were obtained when RNA was assayed in the pooled cultures by hybridization and nuclease protection (Fig. 4). This method allowed correctly initiated transcripts to be measured, a particularly important consideration for E1b-promoted RNA since read-through transcripts initiated from the E1a promoter can contain E1b sequences [e.g., [(17)]]. Standardized to the amounts of E1a-neo RNA, E1b-gfp transcription from the integrated DNA was inhibited by GGT and partially restored when the terminator was inactivated (Fig 4A, “Neo” and “GFP” panels, quantification shown in Fig. 4B). As observed previously in early virus infections, GGT reduced read-through transcription substantially and inactivation of the termination function partially restored levels of read-through transcription (Fig. 4A, “RT” panel). These results show that termination of E1a-neo transcription interfered in cis with expression of the downstream E1b-gfp gene from integrated DNA copies.
Individual G418-resistant clones produced by transfection with NeoLuc plasmids were screened for luciferase activity and luciferase-positive clones derived from the different plasmids were analyzed quantitatively for RNA production by hybridization protection (Fig. 5). The cell lines differed markedly in the transcription activity of the NeoLuc cassette. E1a-neo and E1b-luc RNA levels (Table 3) varied more than thirty-fold and the latter correlated well with the results of luciferase assays from the same cell line (data not shown). After normalization to E1a-neo RNA, E1b-Luc RNA expression in the cell lines containing an inactivated termination sequence (DPM) was increased over the levels obtained from inserted sequences with a functional terminator (GGT) (Table 3). As with the NeoGFP cassettes, the termination sequence was effective in reducing read-through transcription to near background levels in all of the cell lines obtained by insertion of a NeoGGTE1bLuc expression unit (Table 3). Note that some cell lines produced levels of read-through RNA that were below the threshold that could be measured (nd, Table 3). These results support the conclusion obtained with the NeoGFP cell lines, that integrated DNA can support read-through activation.
Here we provide evidence that cellular chromatin supports read-through activation of a closely linked gene. The activation of downstream gene expression was observed at both population and individual clone levels in non-targeted gene insertions. Read-through transcription enhances expression of a downstream HIV-1 proviral genome inserted into an intron of the dihydrofolate reductase (DHFR) gene when the DHFR promoter and HIV promoter are in the same orientation . Our results support and extend this finding by showing that activation by read-through transcription in the cellular genomic context did not require either a specific site of integration or the retrovirus proviral elements. Accordingly, the read-through activation mechanism should be available to coordinate gene expression of closely linked genes in both viral and cellular genomes. Non-coding transcription also has been implicated in transcriptional activation of at least one downstream cellular gene , although a trans-acting function for the non-coding transcript was not ruled out completely.
Activation in cis of downstream gene expression in the cells lines by read-through transcription was modest, about two-fold in our experimental system, and less than the magnitude of activation we observed in early virus infections , . The modest effect of read-through on activation of the downstream reporter in cell lines could be related to the use of the Neo gene for selection. Silencing of linked promoters by neo was reported previously in assays of activity from transiently transfected templates and integrated sequences . We also observed sharply reduced E1a-neo and E1b-mediated transcription in adenovirus strains with Neo gene replacements for E1a (D. Spector, unpublished data). Conceivably, the silencing elements in the Neo coding region, or other DNA sequences in that region, could directly impact the activation mechanism as well. If so, identification of the relevant sequences in the Neo coding region could provide further insight into the mechanism of activation.
Gene expression from integrated sequences also is undoubtedly affected by the local chromatin environment and nearby regulatory elements. In fact, selection for Neo expression probably favors the survival of cell lines with integration sites in genomic locations that are transcriptionally active. Such a bias might increase the probability of integration in the vicinity of a strong cellular enhancer; strong enhancers relieve the requirement for activation by read-through transcription in adenovirus . If the same is true for the cellular copies integrated near strong enhancers, then GGT-containing clones should have higher levels of E1b-mediated expression than expected. This circumstance would also result in underestimation of the potential benefit from the read-through mechanism.
We note that, as in viruses, GGT blocked read-through transcription in integrated sequences and that mutation of the hexanucleotide recognition sequence for polyadenylation restored read-through. These results highlight the versatility of the β-globin element as an effective transcription termination sequence.
We analyzed read-through from randomly integrated expression cassettes because we could not predict how any particular configuration of the targeted integration site would affect the interaction. The results here provide some guidance in that respect. It would seem desirable to minimize potential effects of nearby control elements, either by utilizing an “inert” site or providing insulator sequences ,  as boundaries for the expression cassette. Also, careful consideration must be given to choice of selectable marker and reporter gene sequences to ensure that they are inert as well with respect to the molecular mechanism under investigation.
Plasmids with neomycin (Neo) resistance gene (neo) and the green fluorescent protein (GFP) reporter gene (gfp) were constructed from two sources. The Neo and GFP genes were derived from the donor plasmid, pEGFP-N3 (Clontech). The genes were inserted into “term” vector plasmids, which contain E1a enhancer-promoter regulatory sequences, a wt or mutated GGT, respectively, and the E1b promoter and gene.
The backbones of the “term” vectors were plasmids pm563, which has a mutation at the 5′ end of the E1a coding region that introduces an NcoI site , , and p563/112 . The former contains adenovirus sequences from E1a and the 5′ portion of E1b, whereas the latter includes most of the E1b gene and has an E1b dl112 allele . pm563Δ lacks E1a sequences between positions 563 and 1338, which were removed by excising the DNA between the unique NcoI and XbaI sites in the plasmid and adding an XbaI linker to retain the XbaI site at the junction. To make the first donor “term” plasmid, pΔterm, the MscI-BglII fragment of pβmajG [obtained from Erik Falck-Pedersen ], containing the mouse βmaj gene transcription termination sequence, was inserted into the XbaI site of pm563Δ.
The second donor “term” plasmid, pΔtermdpm, was constructed in multiple steps. First, pterm and pterm112 were constructed by inserting the MscI-BglII fragment of pβmajG into the NcoI site of pm563 or p563/112, respectively. Next, pΔterm112 was produced by exchanging the HpaI fragment from pΔterm that includes the E1a deletion for that of pterm112. Then, the BstXI- XbaI fragment in pΔterm112 that contains the polyadenylation sites in the transcription termination sequence was exchanged for the corresponding fragment from pD''EF (obtained from Erik Falck-Pedersen ), which has those sites inactivated by mutation, to produce pΔtermdpm112. The deleted E1a region was restored by exchange of the HpaI fragment from pterm to produce plasmid ptermdpm112. Finally, the EcoRI-XbaI fragment from ptermdpm112 that contains the mutated polyadenylation sites was exchanged into pΔterm to produce pΔtermdpm.
To produce the plasmids containing neo and gfp, a fragment containing the Neo coding region was excised from pEGFP-N3 by AvrII digestion and inserted into the BseR1 site of pΔterm or pΔtermdpm. The resulting plasmids, with neo in place of the E1a coding region, were designated pNeoGGT and pNeoDPM. A SmaI-SspI fragment containing the GFP coding region, its polyadenylation signals, and polyadenylation site, was excised from pEGFP-N3 and substituted for an EcoNI-HindIII fragment (the E1b gene region) of the NeoGGT and NeoDPM plasmids. The resulting plasmids, pNeoGGTE1bGFP and pNeoDPME1bGFP, contained neo driven by the E1a enhancer-promoter (E1a-neo), a wt or mutated GGT terminator, and gfp under the control of the E1b promoter (E1b-gfp).
A third plasmid, pNeoE1bGFP, without termination sequences between the two genes was constructed by excising the CT termination sequence , as a BstXI-HindIII fragment, from pNeoCTE1bGFP and self-ligating the vector. To construct pNeoCTE1bGFP, a BstXI-SalI fragment that contains CT and the E1b promoter was excised from pACCTE1bLuc  and substituted for a BstXI-BamHI fragment, containing GGT and the E1b promoter, of pNeoGGTE1bGFP.
The expression cassettes from pNeoE1bGFP, pNeoGGTE1bGFP, or pNeoDPME1bGFP were excised as EcoRI-NdeI fragments and substituted for the EcoRI-FseI fragment in pAC343CTE1bLuc , which includes the E1a and E1b-luciferase gene (luc) regions, to produce pACNeoE1bGFP, pACNeoGGTE1bGFP, or pACNeoDPME1bGFP, respectively. Plasmids with neo in place of E1a and luc in place of E1b were constructed by exchanging a PshAI-EcoRI fragment of p343E1bLuc  containing E1b-luc for a similar fragment containing E1b-gfp in pACNeoE1bGFP, pACNeoGGTE1bGFP, or pACNeoDPME1bGFP. The resulting plasmids were designated pNeoE1bLuc, pNeoGGTE1bLuc, or pNeoDPME1bLuc, respectively.
Both lipofection and electroporation were used for DNA uptake and isolation of stable neomycin-resistant cell lines with copies of plasmid DNA randomly integrated into the cellular genome. For lipofection, HeLa cells were subcultured into 6-well plates one day before transfection at 70–90% of confluence. 3 µl of Fugene 6 reagent and 1 µg of plasmid DNA were diluted to a volume of 100 µl with Opti-MEM medium and added to 1 well. After 48 hrs, cells were subcultured 1 to 30 into 150 mm dishes and placed under G418 selection (400 µg/ml) for two weeks until resistant cell colonies appeared.
For electroporation, HeLa cells were removed from monolayer surfaces and suspended in electroporation buffer (1× HeBSS, pH 7.10) at a concentration of 1×107 cells/ml. 1 ml of cell suspension was transferred to a sterile electroporation cuvette. Linearized plasmid DNA (1, 2, or 5 µg) was added to the cuvette and the suspension was mixed by inversion. Electroporation was performed at a setting of 230V/960 µF in the Gene Pulser XcellTM (Bio-Rad Laboratories, Inc.). The suspension was removed and sequential rinses were performed with 0.8 ml aliquots of medium to ensure complete removal of the cells. The rinses were combined with the electroporated cells and transferred to a T75 flask containing 40 ml of medium. After allowing 48 hrs for cell recovery, the cells were inoculated into 150 mm dishes at 6×105 cells/dish, corresponding to three to six dishes for each electroporated cell sample and construct. Cells were placed under G418 selection (400 µg/ml) for two weeks until resistant colonies appeared. The E1b-luc-containing colonies were trypsinized and subcultured into 12-well plates (one colony per well). After reaching confluence, cells from a single well were expanded for maintenance under neomycin selection and screened for luciferase activity. About 70% of the neo-resistant colonies produced measurable luciferase.
Total RNA was prepared from cell lines with Trizol (Invitrogen Corp.) using the protocol provided by the supplier. Specific transcripts were quantified by hybridization and protection from nuclease S1 digestion as described previously . For hybridization, double-stranded probes labeled at a single 5′-end with T4 polynucleotide kinase and [gamma-32P] ATP were prepared as described previously . To detect E1a-neo transcription, an 1126-bp probe was prepared as a NcoI (labeled site)-EcoRI fragment from pNeoGGTE1bGFP or pNeoE1bLuc. To detect E1b-luc RNA, a 1576-bp probe was prepared as a ScaI (labeled site)-SacII fragment of pAC343E1bLuc . To detect E1b-gfp transcription, a 536-bp probe was isolated as a BssSI (labeled site)–PshAI fragment from pNeoGGTE1bGFP. To detect read-through transcription from cell lines with the luc gene, a 1215-bp probe was prepared as a HpaI (labeled site)-SacII fragment of pAC343E1bLuc. To detect read-through transcription from cell lines with E1b-gfp genes, a 1410-bp probe was prepared as an HpaI-FspI fragment from pNeoCTE1bGFP or pNeoCTDPME1bGFP. The latter was constructed by isolating a BstXI-SalI fragment from pACCTdpmE1bLuc  and replacing the BstXI-BamHI fragment of pNeoGGTE1bGFP.
Band intensities were quantified at the Penn State Hershey Core Facility by scanning densitometry of ×-ray film (Kodak ×AR5) in a GS-800 Calibrated Densitometer (Bio-Rad), or by using a Molecular Dynamics phosphorimaging screen that was analyzed with Quantity One software and an FX scanner (Bio-Rad Laboratories, Inc.). E1b-dependent transcription was normalized for template copy number and gel loading as described previously .
Luciferase assays were performed on cell lines harvested at about 80% confluence. Cell lysates were prepared and the luciferase activity was assayed using the Luciferase Assay System with Reporter Lysis Buffer (Promega Corp.) according to the manufacturer's protocol. Light emission was quantified on an FB12 Luminometer (Zylux Corp.). All values were normalized to extract protein concentrations (DC protein assay; Bio-Rad Corp.).
Flow cytometry was performed at the Penn State Hershey Core Facility. For determination of GFP expression, G418-resistant colonies generated as described above were trypsinized, pooled, and the cells were replated for expansion without G418 selection. Cell suspensions were prepared subsequently at a concentration of about 1×106 cells/ml and about 10,000 cells were analyzed in each flow cytometry run. Cells with a fluorescence intensity parameter FL1-H ≥101.1 (determined from a sample that expressed no GFP) were scored. The percentage of green cells in the population and the fluorescence intensity per cell were recorded for each sample. Values from replicates were averaged.
To detect plasmid DNA sequences integrated in the genomes of stable cell lines, G418-resistant colonies obtained as described above were trypsinized and subcultured into 6-well plates (1 colony per well). After reaching confluence, cells from a single well were expanded for maintenance under neomycin selection. Cellular DNA was extracted and digested (10 µg/sample) with SphI, and blot hybridization was performed as described . The probe was an EcoRI-NdeI fragment that contains the full expression cassette in plasmid pNeoE1bGFP. Reconstructions were performed with salmon sperm DNA but calculations were based on copies per quasi-tetraploid human DNA content of HeLa cells.
We thank Richard Courtney and members of the Department of Microbiology and Immunology for their support. Anonymous reviews provided useful comments.
Flow cytometry, DNA sequencing, and phosphorimaging instrumentation were provided by the Penn State, Hershey, core facility.
Competing Interests: The authors have declared that no competing interests exist.
Funding: The initial stages of this work were funded by Public Health Service grant GM058214 from the National Institute of General Medical Sciences. There is no current external funding source for this study. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.