Introduction

Recombinant adenoviruses are attractive vehicles for in vitro and in vivo gene transfer into a wide range of cell types. First generation, E1-deleted, adenovirus vectors as well as E2- and/or E4-deleted vectors contain within their genome promoters for the early regions E1A, E2, E3 and E4, for two delayed early units (IX and IVα2), and for one late unit (MLP). Helper-dependent, ‘gutless’ vectors12 are deleted for viral sequences except for those required for viral replication and packaging, which include the ITRs and part of the E1A promoter. Adenoviral promoters contain multiple recognition sites for transcription factors often representing cis-acting DNA sequences that increase transcription in a manner that is independent of their orientation and distance relative to the RNA start site. The latter properties define these sequences as enhancers. The presence of viral enhancers and upstream regulatory elements in the vector genome implies a possible interference with heterologous promoters used to drive transgene expression. This interference may affect the activity, tissue specificity and/or inducibility of these heterologous promoters. However, tissue-specific or regulatable gene expression is often a critical prerequisite for studying gene functions in vitro and in vivo, as well as for gene therapy approaches. For a number of genetic disorders, the transgene expression must be restricted to the target tissue or must be at a specific level requiring endogenous or inducible promoters, whose activity can be fine-tuned.

Among the viral sequences that are potentially able to interfere with the heterologous transgene promoters are the ITRs, the E1A enhancer, the E2, E4 and pIX promoters (Figure 1c). All recombinant adenovirus vectors produced so far contain the adenoviral (Ad5) sequences from nucleotide (nt) 1 to nt 341 including the Ad ITR and the E1A enhancer. The 102 bp long Ad ITR contains binding sites for a number of transcription factors (SP1, ATF, NFIII/OTF-1, NFI/CTF)3 which have been shown to regulate transcription of cellular and viral genes. The deletion of the corresponding transcription factor binding sites impairs viral replication. In its normal location, the ITR acts as an enhancer for transcription from the E4 promoter4 and stimulates transcription (in a location-dependent manner) from the E1A promoter.35 The region between the E1A enhancer and the ITR has also been shown to exhibit enhancer activity.3 The E1A enhancer contains two functionally distinct domains, one is specific for E1A (nt 200–300) and the other transactivates all early units (nt 250–280).678 The upstream half of the E1A enhancer appears to have a specific conformation critical for its transactivation9 and represents the binding site for a cellular protein termed EF-1A. The E1A enhancer overlaps physically with cis-acting sequences required for packaging of viral genomes (nt 194–358)10 and can therefore not be removed from Ad vectors (Figure 1b). The E2 and the E4 promoters located at m.u. 77.5 and 99.1 also contain sequences, which have the qualitative properties of enhancer elements.1112 These sequences can function in either orientation over a distance of several kilobases in an E1A-independent or -dependent manner.13 Since the E2 and E4 promoters require E1A for transactivation, the corresponding enhancers may not critically affect heterologous promoters within E1-deleted vectors. On the other hand, a number of recent studies demonstrated that E2 and E4 proteins are expressed in cells transduced with first-generation adenovirus vectors indicating that these enhancers are active in the absence of E1A.1415 It was demonstrated earlier that cellular transcription factors can functionally substitute E1A in its transactivator function, which makes the E2 and the E4 promoter potential candidates for interference.151617 The pIX promoter is orientated rightwards with respect to the viral genome and is located directly adjacent to the E1 insertion site for transgene cassettes. Besides enhancers that act distance and orientation independent, viral sequences contain transcription factor binding sites that represent upstream regulatory elements (UREs).18 UREs can activate transcription from a core promoter when located within a certain distance from the transcription start site.

Figure 1
figure 1

Structure of constructs and recombinant adenoviruses. (a) Transgene expression cassette. The heavy-metal responsive promoter contains a core promoter and five MREs at positions −54 to −150.29 The core promoter resembles a TATA box followed by a tet-operator sequence54 and the transcription start site of the thymidine kinase gene.29 The lengths of 5′ and 3′ non-translated sequences present in the hAAT fragment are indicated. The 12 nt long 5′NTR contains the consensus sequence for an initiator element (YA+1YTCYYY).39 The schematic presentation of the transgene expression cassette used in all further constructs is shown at the bottom. bPA: bovine growth hormone gene polyadenylation signal. (b) Plasmids containing adenoviral sequences. The transgene expression cassette was inserted into a derivative of pΔE1sp1A with Ad sequences 22–341 and 3524–5790. These regions contain the E1A enhancer (E1AE), which overlaps with the Ad packaging signal (Ψ), and the pIX units including the pIX promoter directly adjacent to the insertion site. (c) First-generation adenoviruses. The vectors were obtained by recombination in 293 cells between pΔE1sp1A derivatives containing the transgene cassette and pJM17. The adenoviral genome is deleted for the E1 genes and contains deletion/substitution within the E3 region, which do not affect the E3 promoter.55 The transgene cassette was inserted into the E1 region. The transcription start sites of promoters that can potentially interfere with the transgene expression unit are indicated.

Full size image

Insulators are DNA elements that protect an integrated reporter gene from chromosomal position effects or block enhancer activated transcription from a downstream promoter (for review see Refs 19 and 20). Chromatin boundary elements with insulator activity have been found for Drosophila melanogaster loci (Gypsy, suppressor of Hairy wing, scs, scs′, Fab-7), for the chicken β-globin locus, and for the T cell receptor locus. A number of models have been proposed for the mode of action of insulators which fall into two major categories: steric models and tracking models.2021 Steric models assume that insulators separate enhancer and promoter into two inaccessible to each other domains. Tracking models postulate that insulators block some activating signal that travels along the DNA from enhancer to promoter. An unambiguous distinction between these models seems to be difficult due to lack of sufficient data. Furthermore, a recent study demonstrated that the activity of the HS-4 insulator could be affected by flanking chromatin sequences suggesting that the function of HS-4 elements as insulators is more complex.22

The DNase hypersensitivity region 4 (HS-4) of the chicken β-globin locus resembles a CpG island-like sequence and functions as an insulator.23 This region is located within a 1.2 kb DNA fragment, which was used in most insulator studies. Two copies of the 1.2 kb fragment flanking a human γ-globin promoter shielded against transactivation by an upstream enhancer from the mouse β-globin locus. The HS-4 domain is DNase-I sensitive in all cell types indicating association with ubiquitous factors and arguing against tissue specificity of the given insulator.23 One of these factors, a zinc-finger DNA-binding protein that is highly conserved in vertebrates was recently identified.24 So far, insulators have been successfully applied to shield integrated transgenes from position effects or enhancers in a chromosomal context.202425 Very recently, the HS-4 insulator has been employed in retrovirus vectors to protect transgene expression from negative position effects.2627

The goals of this study were to evaluate the potential effect of above mentioned viral promoter/enhancer elements on transgene expression from a metal-inducible promoter2829 and to test whether an insulator element derived from the chicken β-globin locus can be used to shield this promoter from undesired interference by viral enhancers. The insulator effect was analyzed after adenoviral gene transfer in vitro and in vivo. Viral transduction studies were performed with E1-deleted, first-generation vectors.

Results

Interference by viral regulatory elements

The heterologous promoter used in this study represents a core promoter containing a transcription start site and TATA box combined with five copies of metal-responsive elements (MREs)2829 (Figure 1a). In the non-induced stage, only minimal basal transcription is initiated from the core promoter. In the presence of heavy metals, MRE binding transcription factors are activated and induce transcription. This promoter was placed in front of the human α1-antitrypsin (hAAT) cDNA followed by the bovine growth hormone polyadenylation signal (bPA). Unspecific interference by viral enhancers can be assessed by measuring the basal and induced hAAT expression in the presence and the absence of viral sequences within transfected test plasmids. In addition, this inducible promoter allows, on the one hand, for testing whether the insulator/s can block interference by viral enhancers as measured in changes of basal expression and, on the other hand, for analyzing whether insulators affect the intrinsic activity of the MRE promoter as measured by its inducibility.

To analyze the impact of viral sequences on inducible gene expression, the MRE-hAAT expression cassette was cloned into pBS (pBSMRE) and into a modified pΔE1sp1A with Ad sequences 22–341 containing most of the left ITR and the E1A enhancer as well as nucleotides 3523–5790 which include the pIX promoter (pAdMREa/b) (see Figure 1b). These constructs were transfected into 293 cells and hAAT concentrations were measured 2 days later in the culture supernatant in the presence and absence of ZnSO4 (Figure 2a). HAAT expression from the pBSMRE construct was slightly over background in the absence of ZnSO4 and 50- to 55-fold higher in the presence of ZnSO4. The level of induced hAAT expression was comparable with that from the strong RSV promoter (pAdRSV), which was not responsive to ZnSO4 induction. Non-induced, basal hAAT expression was significantly higher from the adenoviral shuttle plasmids, pAdMRE-a and -b, containing the hAAT expression cassette in either orientation suggesting transactivation of the core promoter by viral enhancer/s. The inducibility of the MRE-hAAT cassette cloned in the rightward orientation with respect to the 5′Ad ITR (pAdMREa) was abolished since hAAT expression levels in the non-induced and induced stages were comparable. Expression from plasmids containing the leftward orientated cassette (pAdMREb) was approximately five-fold inducible compared with approximately 50-fold in constructs without Ad sequences. Interestingly, significant hAAT expression was observed for pΔE1sp1A-based constructs only containing a rightwards-orientated hAAT cDNA without the MRE promoter (pAdhAATa) (Figure 2a). No hAAT expression was detected if the hAAT cDNA was in the context of pBS without adenoviral sequences. This suggests that nt 22–341 contain either an intrinsic promoter or upstream regulatory elements, which when assembled together with a suboptimal transcription start site present within the hAAT cDNA form an active promoter. Similarly elevated basal expression levels which differed quantitatively depending on the orientation of the transgene cassette were observed with recombinant adenoviruses (Figure 2b). The cis-activity mediated by Ad nt 1–341 was even more pronounced in adenoviral vectors than on plasmid level. The promoterless vector with the hAAT cassette in rightward orientation (Ad.hAATa) yielded about one third of the hAAT expression of the corresponding vector with promoter (Ad.MREa).

Figure 2
figure 2

Activation of hAAT expression by viral sequences. This shows absolute hAAT levels after adjustment of transfection efficiencies as well as the ratios between ZnSO4-induced (+) and non-induced (−) expression (indicated on the right). Shown is the average hAAT concentration ± s.e.m (n > 3). HAAT expression from the MRE hAAT cassette was compared with expression from the RSV promoter, which is widely used for high-level gene expression after adenovirus-mediated gene transfer. The induction ratios for promotorless constructs were not applicable (n/a). (a) Plasmids were transfected into 293 cells in parallel in two sets of dishes. 150 μM ZnSO4 was added to one set of dishes to induce the MRE promoter. Induced and basal hAAT levels were measured 2 days after transfection in culture supernatants. All test plasmids were spiked with pCMVlacZ to correct for differences in transfection efficiency (see Materials and methods). Average transfection efficiency was approximately 80% as measured by the number of β-gal-positive cells. Expression levels from the RSV-hAAT cassette cloned into the pΔE1sp1A derivative were similar for both transgene cassette orientations and did not respond to ZnSO4 addition (not shown). (b) 208f Cells were infected with adenovirus at an MOI of 100. Induction with ZnSO4 was started 6 h after infection and hAAT analysis was performed at 48 h after infection. Constructs and viruses are presented in schematic form as outlined in Figure 1. Note: pAd.hAATa/b and the viral vectors Ad.hAATa/b do not contain heterologous promoters.

Full size image

Effects of insulators in vitro

The HS-4 insulator was selected because it appeared to function within mammalian cells in a cell-type independent manner and to lack evident enhancer/promoter elements.23 This was a prerequisite for our aim to improve regulated gene expression in adenoviral vectors. We used the 1.2 kb fragment derived from the 5′ end of the chicken β-globin locus containing the HS-4 region, hereafter designed as HS-4 insulator. In order to exert its enhancer-blocking activity, the insulator must be inserted between the enhancer and the promoter of interest. As outlined above, Ad vectors contain the E1A enhancer upstream of a transgene inserted into the E1 region and the E2, E3, and E4 enhancers downstream of the insertion site. This implies that the promoter or expression cassette should be flanked on both sides by insulators. To prove this, pΔE1sp1A-based constructs were generated containing one or two HS-4 insulators flanking the MRE-hAAT expression cassette (Figure 3). Using only one-sided insulation was expected to allow for better differentiation between the impact of upstream and downstream viral enhancers on transactivation. An additional aim was to analyze whether or not the orientation of the insulator with regard to the MRE promoter and the viral sequences influences its function. Therefore, a total of 16 different constructs with all possible combinations and orientations of transgene cassette and insulator/s were generated and tested in transfection experiments.

Figure 3
figure 3

Adenoviral shuttle vectors containing HS-4 insulators. The detailed structure of individual elements is described in Figure 1. HS-4 is the 1.2 kb insulator element derived from the chicken β-globin locus. The arrow indicates the orientation of the HS-4. Elements labeled with ‘→’ have the 250 bp GC-rich core region on the left side of the fragment. Sixteen different pΔE1sp1A-based shuttle vectors containing all possible insulator/transgene cassette combinations were generated and tested in transfection experiments.

Full size image

Interestingly, constructs with the HS-4 element in 3’→5’ orientation upstream of the MRE promoter demonstrated high basal expression levels suggesting that an intrinsic cis-activating activity was present in the HS-4 fragment when cloned in this constellation (data not shown). To investigate this effect further, transfection studies with pBS-based constructs containing the insulator in different orientation linked to the MRE-hAAT cassette were performed (Figure 4). Basal expression from the pIns1/4 was 10 to 20-fold higher than from constructs with other insulator/transgene cassette combinations. Due to this interfering effect, all constructs containing this particular insulator/expression cassette constellation were excluded from further studies.

Figure 4
figure 4

Effect of different orientations of HS-4 elements on hAAT expression. To investigate better the presence of cis-activating activity within the HS-4 element, the insulator was cloned together with the MRE-hAAT cassette into pBS in different orientations and hAAT concentrations were measured with and without induction (n = 3).

Full size image

The most informative pΔE1sp1A-based constructs were used to produce the corresponding recombinant adenoviruses and tested in vitro (Figure 5). 208f Cells were infected with a multiplicity of infection (MOI) of 100 p.f.u./per cell and hAAT expression was analyzed 48 h later in parallel with and without ZnSO4 induction. A MOI of 100 was used because this was the lowest MOI that allowed for transduction of 100% of 208f cells as testes with a lacZ expressing adenovirus (data not shown). Transduction studies were performed in 208f cells, because this cell line did not support significant Ad replication, in contrast to most hepatoma cell lines.30

Figure 5
figure 5

Effect of insulators on hAAT expression after adenovirus infection in vitro. The most informative pΔE1sp1A derivatives shown in were used to produce first-generation adenoviruses. 208f Cells were infected with a MOI of 100, which was sufficient to transduce all cells. HAAT expression was measured with and without induction as described in Figure 2. Absolute hAAT concentrations and induction ratios are shown (n > 3). Note: Induction studies with Ad vectors could not be performed in E1A expressing 293 cells due to development of cytopathic effects.

Full size image

The vector with the rightwards-orientated cassette (Ad.MREa) demonstrated higher basal expression levels resulting in lower induction ratios than the vector with the expression cassette cloned in leftward orientation (Ad.MREb). One insulator inserted downstream of the rightwards cassette (Ad.Ins1/1a) reduced basal expression. This effect was more pronounced with one insulator inserted upstream of the rightwards cassette (Ad.Ins1/3a). This indicates that the Ad region nt 1–341 contain enhancer as well as a distance depending cis- activating activity, which was described earlier for phAATa/b constructs (Figure 2). Two insulators surrounding the rightwards cassette (Ad.Ins2/1a) have a clear additive shielding effect resulting in induction factors that were approximately 40-fold higher than the corresponding insulatorless control (Ad.MREa). The double insulator combination affected the absolute hAAT levels upon activation compared with the control Ad.MREa. However, while the levels for activated expression dropped only by a factor of four, the two insulators in Ad.2/1a reduced the non-induced, basal expression by more than a factor of 200 compared with the insulatorless control.

When inserted between the leftward-orientated cassette and downstream Ad sequences (AdIns1/3b), one insulator can improve the inducibility of the MRE promoter and shield against downstream enhancers. In this case, the induction factor was approximately 19-fold compared with approximately 12-fold for the corresponding construct without the insulator (pAdMREb). One insulator inserted between the upstream Ad sequence and leftward cassette (AdIns1/1b) blocked the interference by the enhancers present in Ad region nt 1–341 when compared with AdMREb. This indicates that interfering enhancer/s are present upstream as well as downstream of the expression cassette. Again, the vectors with two insulators flanking the leftward cassette showed the best induction factors. For Ad.Ins2/1b the levels of activated expression were approximately seven-fold reduced whereas the basal expression was inhibited by approximately 30-fold to nearly undetectable levels.

In conclusion, comparing the induction ratios from Ad vectors with a leftwards-orientated hAAT cassette (Ad.MREb, Ad.Ins1/1b, Ad.Ins1/3b), the shielding effect of one insulator against downstream enhancers (Ad.MREb versus Ad.Ins1/3b) and upstream located viral enhancers (Ad.MREb versus Ad.Ins1/1b) becomes evident. Two insulators have a clear additive enhancer-blocking effect in adenoviruses (Ad.Ins2/1a/b, Ad.Ins2/2b). Importantly, both hAAT cassette orientations gave similarly high induction factors (approximately 40-fold) when shielded with two insulators in contrast to the corresponding vectors without insulators (no induction and 12-fold induction for rightwards and leftwards directed expression cassettes, respectively). Two insulators affected the absolute expression levels upon activation. This effect was more pronounced for certain double insulator-transgene cassette combinations (Ad.Ins2/1b) than for others (Ad.2/2b). However, the inhibitory effect of double insulators was much less for activated expression than for basal expression resulting in greatly improved induction factors compared with insulator-less vectors.

In an attempt to better characterize the structural elements within the 1.2kb HS-4 elements that are involved in insulation, we performed analyses with the double insulator virus that had demonstrated the best insulation and induction rates (Ad.Ins2/2b). Two derivatives of Ad.Ins2/2b were generated that contained truncated HS-4 elements flanking the transgene cassette (Figure 6). These DNA fragments were deleted for the 250 bp insulator core region, which, according to recent publications, accounts for a large portion of the enhancer blocking activity.24 Viruses with truncated HS-4 elements (Ad.ΔIns1 and Ad.ΔIns2) demonstrated lower induction rates than Ad.Ins2/2b. Basal expression from Ad.ΔIns1 and Ad.ΔIns2 was comparable with that from the non-insulated vector, Ad.MREb indicating less efficient insulation. This suggested that the insulator effect was greatly diminished when the functional important 250 bp HS-4 core region was removed from the 1.2 kb fragment. The data with vectors containing the 0.5 and 1.0 kb DNA fragment give an example that the insertion of random ‘stuffer’ DNA flanking the expression cassette will not exert the same shielding effects as the complete HS-4 element.

Figure 6
figure 6

Shielding effect of truncated HS-4 elements. The adenovirus vector Ad.ΔIns1 or Ad.ΔIns2 contained a 1.0 kb or 0.5 kb fragment of the original HS-4 elements used in Ad.Ins2/2b (see Figure 5). The restriction sites HindIII and AflII used to produce the Ad.Ins2/2b derivatives containing the truncated HS-4 fragments are indicated in the upper chart. Ad.MREb is the corresponding non-insulated vector with the transgene cassette in leftward orientation (see Figure 5). 208f Cells were infected with a MOI of 100. HAAT expression was measured with and without induction. Absolute hAAT concentrations and induction ratios are shown (n > 3).

Full size image

Insulator effect in vivo

Our ultimate goal was to test whether insulators can shield the MRE promoter when the vectors were applied in vivo. Therefore, the adenovirus vectors used before in the in vitro studies were injected via tail vein infusion into C57Bl/6 mice. This mouse strain is known for a persistent hAAT expression, which appears not to be affected by cellular or humoral immune responses against viral proteins or hAAT.31 This allows for performing induction studies over longer time-periods. Earlier studies have demonstrated that after systemic vector application, 90% of infused adenoviral genomes are found in the liver, predominantly in hepatocytes.32 hAAT expression from the MRE promoter was induced by ZnSO4 added to the drinking water. Serum hAAT levels were measured before and after induction (Figure 7). The data obtained in vivo are widely consistent with the key conclusions drawn based on the in vitro studies. However, there were a number of interesting differences between in vitro and in vivo studies: the basal expression without ZnSO4 induction was generally lower and the expression levels after induction were higher in vivo resulting in approximately four-fold greater induction factors. This is surprising because one would expect that traces of heavy metals present in the blood or hepatocytes would result in higher background levels. The cis-activating activity located within nt 1–341 was not detectable in vivo (Ad.hAATa/b). The highest induction ratios (230-fold) were obtained with vector containing a leftwards-orientated transgene cassette flanked by insulators. Activated expression from double-insulator vectors, particularly from those containing the leftwards orientated was less effected in vivo than in vitro. For the best variant (2/1b), the levels of induced expression dropped only by a factor of 1.6 whereas the levels of basal expression decreased by approximately 12-fold.

Figure 7
figure 7

Effect of insulators on hAAT expression in vivo. Adenovirus (2 × 109 p.f.u.) was administered to C57Bl/6 mice by tail vein infusion. From independent experiments, it was known that induced hAAT expression from the MRE promoter reached a plateau at day 7 and maintained this level for the time of analysis (3 months). Ten days after adenovirus infusion, the basal hAAT concentration was determined in serum from animals. After obtaining this blood sample, 25 mM ZnSO4 was added to the drinking water and induced hAAT levels were measured 7 days later. Absolute hAAT expression levels before and during ZnSO4 induction and induction ratios are shown at a (n = 3). Notes: hAAT concentrations are shown at a logarithmic scale. The same virus preparations were used for in vitro and in vivo studies.

Full size image

Induction of hAAT expression was maintained at the same high level as long as ZnSO4 was applied to the animals (not shown). Furthermore, we demonstrated that 10 days after removal of ZnSO4 from the drinking water hAAT expression dropped again to basal levels (Figure 8). A repeated administration of ZnSO4 resulted in the same induction factor and kinetics as observed for the initial induction round. Expression from the MRE promoter can also be induced by cadmium, bismuth, silver, cobalt, copper, mercury or nickle.28 In our hands, intraperitoneal injection of CdSO4, 20 and 6 h before analysis allowed for a quick induction of transgene expression, however absolute hAAT levels were approximately 10-fold lower than after induction with ZnSO4. The reason for this remains to be clarified.

Figure 8
figure 8

Induction kinetics in vivo. (a) Repeated induction of transgene expression with ZnSO4. The intervals of ZnSO4-administration (given with drinking water) are indicated. Representative data are shown for Ad.Ins1/3b (n = 3). (b) CdCl2 was administered to another group of animals. CdCl2 was injected intraperitoneally into all animals 20 and 6 h before obtaining the serum sample for the first time-point after induction. Data for viruses Ad.Ins1/3a and Ad.Ins2/1a in two individual animals are shown.

Full size image

In summary, the adenovirus-based, heavy metal-inducible MRE promoter used in this study is a straightforward and reliable system to obtain up to 230-fold induction of transgene expression in vivo when the transgene cassette is flanked on both sides with insulators.

Discussion

The goal of this study was to evaluate the interference of viral promoter/enhancer elements on the transgene expression from an inducible promoter and to test whether HS-4 insulators can improve inducible gene expression from first-generation adenovirus vectors.

Interference of adenoviral sequences with the activity or specificity of heterologous promoters is an expected problem that was reported earlier. Adenoviral sequences present in pΔE1sp1B including the E1A enhancer and the pIX promoter affected the activity and tissue specificity of transgene expression from a muscle-specific promoter as tested in plasmid transfection.33 The inhibitory effect was observed for both orientations of the transgene cassette. Friedman et al34 reported that cell type-specific expression from the rat albumin promoter within the adenovirus genome in human hepatoma cells was low compared with that of the endogenous albumin gene. Similar observations were described by Babiss et al35 for the β-globin promoter. A number of reports noted that gene expression from tissue-specific promoters in adenovirus vectors was not limited to the corresponding endogenous tissue but occurred in other cell types as well.35363738 In consensus with earlier reports, the present study demonstrated that adenoviral sequences affect basal as well as induced expression from a metal-dependent promoter. This effect was seen regardless of the orientation of the transgene cassette, indicating interaction with viral enhancers located upstream and downstream of the insertion site. Interference with viral elements was quantitatively more pronounced with the transgene cassette cloned in rightwards direction. In this orientation, a high basal expression was detected in vitro even in the absence of a heterologous promoter. This was attributed to an URE-mediated cis-activating activity, which was present within adenovirus genome nt 1–341 in addition to an intrinsic enhancer activity. An insulator inserted between nt 1–341 and the rightwards-orientated MRE-hAAT cassette blocked this interfering activity. Since the insulator did not contain canonical polyadenylation signals, the latter activity was most probably caused by UREs rather than by an active transcription start site present within nt 1–341. In support of this hypothesis, it is notable that the 12 bp long, 5′ non-translated region of the hAAT cDNA fragment used in our constructs contained the consensus sequence for initiator elements (YA+1YTCYYY).39 Initiators are core promoters that allow for efficient initiation of transcription in conjunction with upstream regulatory elements or enhancers. In conclusion, the strong cis-activating activity present in nt 1–341 appears in part to be caused by upstream regulatory elements, which in conjunction with a potential initiator present in the hAAT cDNA assemble into an active promoter. Clearly, mapping of transcription start sites is required to prove this hypothesis. Importantly, the nt 1–341 region contains the packaging signal and is therefore essential for propagation of all recombinant adenoviruses. This includes high-capacity, ‘gutless’ vectors, which contain Ad5 nt 1–440 and 35 818–35 935.1

We selected the chicken HS-4 insulator because it is the only insulator derived from mammalian genomes that demonstrated a clear enhancer-blocking activity in a heterologous context.222324252627 In earlier studies, we have unsuccessfully attempted to use the Drosophila ‘gyspy’ insulator for shielding heterologous promoters in Ad vectors (DS, AL, unpublished results). Vectors with a leftwards-orientated hAAT cassette demonstrated higher induction factors than the rightwards-directed cassette on a plasmid and an adenoviral level. This is consistent with earlier reports indicating that unspecific activation by the E1A regulatory region could be reduced when the transgene expression cassette was inserted in the leftward, 3′→5′ orientation with respect to the Ad 5′ITR (for review see Ref. 40). However, this report as well as others,3337 demonstrated significant interference even if the transgene expression unit is cloned in leftward orientation towards the 5′ Ad ITR. Despite quantitative differences, independent of the orientation of the hAAT cassette, vectors with insulators had greater induction ratios based on reduced basal expression and higher expression levels after induction. Insertion of one HS-4 element on either site of the transgene cassette demonstrated that expression from the MRE promoter was affected by upstream (eg E1A) enhancers as well as by viral elements located downstream of the insertion site (eg E2, E3, E4 or MLP). Logically, two insulators flanking the transgene cassette conferred maximal insulation and induction ratios which were approximately 230-fold in vivo and approximately 40-fold in vitro. Our data suggest that transgene cassettes flanked by direct repeated HS-4 insulators (Ad.Ins2/1a and Ad.Ins2/1b) were better shielded against cis-activation in vitro and in vivo than transgene cassettes flanked by inverse repeated HS-4 elements (Ad.Ins2/2b).

For the same recombinant virus, basal expression levels and induction ratios differed between in vitro studies performed in 208f cells and in vivo studies which mostly reflect transgene expression within mouse hepatocytes. Activation of viral enhancers/silencers by tissue-specific transcription factors may account for these differences.3441 Importantly, the HS-4 element appeared to function in a cell-type-independent manner and exerted its insulating effect in rat 208f cells and in mice. Since the function of enhancers as well as of insulators24 involves DNA binding proteins specific for a given cell type, enhancer or insulator, caution should be exercised in generalizing our results for other regulated systems.

The insulator/s prevented interference between the MRE promoter and viral enhancers/silencers, however, certain insulator/transgene combinations also had decreased absolute hAAT expression levels upon induction, which was more pronounced in the in vitro studies. This may suggest that in the induced stage, viral enhancers probably contribute to transcriptional activation from the MRE promoter in addition to the specific activation by metal-responsive elements and that insulators would block this additional cis-activation. Alternatively, in these cases the HS-4 element/s may affect, to a certain degree, the interaction between the MREs with the corresponding core promoter. Importantly, the HS-4 insulator/s reduced the levels of basal expression to a much greater extent than the levels of activated expression resulting in higher induction factors compared with the insulator-free control. While levels of basal expression in vitro decreased by factor 200 or 30 (for Ad.Ins2/1a or Ad2/1b), levels of induced expression dropped only four- or seven-fold, respectively. In vivo basal expression was reduced by factor 15 or 12, whereas expression levels upon activation were only four- or 1.7-fold lower, for Ad.Ins2/1a or Ad.Ins2/2b, respectively. Notably, the absolute levels of induced hAAT with double-insulator constructs in vivo were approximately 1500 ng/ml, which is in the range of the Rous sarcoma virus (RSV) or phosphoglycerate kinase (PGK) promoters used for high-level in vivo gene expression.42 Also, for most inducible systems (eg for expression of toxic gene products) a minimized basal expression is more important than a maximal, activated expression.

The HS-4 fragment used in this study had a length of 1.2 kb. The relatively large size of this element is probably not problematic when used in ‘gutless’ vectors, however, may reduce the cloning capacity of first-generation vectors. Much of the insulator activity of HS-4 domain is contained within a 250 bp core element.21 Our studies with truncated HS-4 fragments support the functional importance of this core element. Current efforts are focused on testing whether this small core element as mono- or multimer exerts similar effects as the 1.2 kb fragment.

In constructs where the HS-4 element was linked in 3′→5′ orientation to the MRE promoter, a significant intrinsic cis-activating activity originating from the HS-4 fragment was detectable. While lacking a classical TATA or initiator sequence, the G+C-rich HS-4 region contains a number of binding sites for transcription factors23 which may function as UREs. This may account for the elevated basal expression observed with this particular insulator/promoter constellation. From the data presented in Figure 4, it cannot be concluded that the blocking activity of the HS-4 fragment is directional. It is generally thought that insulators act independent of their orientation as was demonstrated for the scs or scs′ elements.43

It is known that enhancer activity is approximately 50-fold higher on chromatin assembled plasmid DNA than that on naked DNA.44 In this context, the issue of whether or not the process of insulation requires the involvement of chromatin and chromosomal context of the transgene is controversially discussed in the literature. It has been shown that scs or scs′ can block enhancer-activated transcription on plasmid DNA microinjected into Xenopus laevis oocytes.4546 However, this assay cannot definitively discriminate the transcription activity from plasmid DNA fully assembled or unassembled into mini-nucleosomes. The incoming adenoviral genome has a supercoiled structure and is associated with core proteins VII, V and IVα2. Several studies indicate that once the viral genome is translocated to the nucleus, the viral core structure is replaced by a nucleosome-like structure involving cellular histones.474849 Similarly to the adenoviral genome, transfected plasmids associate with histones and are packaged into a nucleosome-like structure shortly after transfection. Therefore, while our study demonstrates that insulators can act in cis on episomal DNA this does not exclude that chromatin-like structures are functionally important for insulators.

The concrete mechanism of action of HS-4 insulators is unclear. The insulation effect cannot be simply reduced to termination of transcription from cryptic promoters present within the adenoviral genome. There are no canonical poly-adenylation signals within the 1.2 kb HS-4 elements. Furthermore, these HS-4 elements have been incorporated into retrovirus vectors without affecting the virus titers, which indicates that transcription of the retrovirus genome is not prematurely terminated by the HS-4 elements.2627 In vitro, part of the activation of basal expression by the Ad region 1–341 is caused by upstream regulatory elements that form together with an initiator sequence in the hAAT fragment, an active promoter. This promoter configuration can be disrupted by an HS-4 element as it can probably be disrupted by any random DNA fragment. However, this effect cannot fully account for the reduction of basal expression by HS-4 elements in vitro. A clear cis-activation by adenoviral enhancers that can be blocked by HS-4 elements is evident if one compares hAAT levels for AdMREb versus Ad.Ins1/3b or AdMREa versus Ad.Ins1/3a. More importantly, the promoter activity formed by joining the Ad1–314 and the hAAT fragment was not observed in vivo. At the same time, there was unspecific activation of the MRE promoter by Ad enhancers resulting in high basal expression from AdMREa and AdMREb in vivo. This basal expression could be reduced by more than 10-fold by two insulators flanking the expression cassette demonstrating a clear enhancer-blocking effect of insulators at a magnitude that is in agreement with other studies on HS-4 elements.222425 Importantly, studies with HS-4 elements truncated for the 250 bp-core region underscored the enhancer-blocking function of HS-4 elements in Ad vectors. Taken together, our data demonstrated that the shielding effect of HS-4 elements was in part due to an enhancer-blocking activity.

The HS-4 insulator was recently used in a regulatable, helper-dependent adenovirus.2 In this study, a duplicate of the insulator was employed to block the nonspecific interaction between the heterologous promoters of two expression cassettes present within the same vector. However, the 2.4 kb HS-4 dimer significantly affected the activity and/or specificity of the two promoters in vivo and did improve the inducible expression system. As suggested in the present report, the 3′→5′ orientation of the HS-4 domain exerts cis-activating activity, which may have contributed to the problems encountered by Burcin et al.2

Ring et al38 had previously shown that the specificity of an ERBB2 promoter to ERBB2 expressing breast cancer cells was lost when the promoter was used in adenovirus vectors. The authors speculated that ‘cryptic’ transcription start sites within the Ad genome were not responsible for this loss of tissue specificity. In a recent study by Vassaux et al50 from the same group, an ERBB2 promoter-thymidine kinase (tk) expression cassette was flanked on both sides by bovine growth hormone polyadenylation signals (bPA) in order to increase selectivity of tk expression to breast cancer cells. Specificity of tk expression was assessed based on sensitisation to gancyclovir or RT-PCR. The data suggest that the bPA-containing vector did not significantly express TK in ERBB2 negative cells, whereas the vector without bPAs did. Our results support their observation that UREs (particularly within Ad nt 1–341) located in proximity to the heterologous promoter can contribute to nonspecific cis-activation and that this interaction can be disrupted by inserting a DNA spacer between the UREs and the promoter. However, it is unlikely that bPAs will block cis-activation by distant viral enhancers. Our study, albeit limited to one expression cassette, suggests that this can be achieved by HS-4 insulators in vitro and in vivo.

Our data demonstrate that the double insulator combinations allowed for an up to 230-fold and 40-fold induction of hAAT expression in vivo and in vitro, respectively. This, in combination with barely detectable basal hAAT expression makes the technically straightforward, inducible MRE promoter flanked by insulators practically important. For example, minimized basal expression may allow for the generation of viruses with potential cytotoxic proteins or proteins that affect the adenovirus life cycle. In addition, as a proof of principle, our study suggests that the employment of insulators may be important for incorporating other inducible systems into adenovirus vectors including the Tet-,51 or RU 486-regulated52 systems where the absolute induction factor critically depends on the level of basal expression in the non-induced stage. Furthermore, insulators may improve the specificity and/or activity of tissue-specific promoters in adenovirus vectors.

Materials and methods

Plasmid constructs

Sequences of the metal-responsive promoter and the HS-4 element were taken from the plasmids pMRENeo2829 (gift from Richard Palmiter, University of Washington) and pJC5–4 (GenBank accession No. U78775)2325 (gift from G Felsenfeld, NIH), respectively. To facilitate cloning, the MRE, HS-4 element, and the human α1-antitrypsin (hAAT) cDNA were subcloned in the cloning vectors pBSK(+) (Stratagene, La Jolla, CA, USA) and pSL1190 (GenBank accession No. U13866). A HS-4 containing 1.2 kb XbaI fragment and a MRE promoter containing 200 bp SacII/SacI fragment were inserted into the corresponding sites of the pSL1190 polylinker to generate pSLMRE and pSLJCa and pSLJCb with the HS-4 element in 5′→3′ (a) or 3′→5′ (b) orientation, respectively. The 250 bp GC-rich region of HS-4 including the unique HindIII site is referred to as being the element's 5′ terminus. Accordingly, elements labeled ‘→’ contain the GC-rich region and the HindIII site on their left side. A 1.7 kb HindIII/XhoI fragment from pBShAAT42 containing the hAAT cDNA and the bovine growth hormone gene polyadenylation signal was cloned into the HindIII/XhoI sites of pBSK(+) (pBShAAT). pMREhAAT was obtained by ligating a 200 bp MRE promoter comprising SacII/SacI fragment to a 2.7 kb SacI/XmnI fragment of pBShAAT and a 1.9 kb SacII/XmnI fragment of pBSK(+). The intermediate constructs pJCMREa and pJCMREb were constructed by inserting a HS-4 containing 1.2 kb XbaI fragment and MRE containing 200 bp SacII/SacI fragment into the polylinker of pSL1190. The intermediate constructs pJChAATa and pJChAATb were obtained by inserting the HS-4 comprising 1.2 kb KpnI/SalI fragments of pSLJCa/b into pBShAAT opened by KpnI/XhoI digestion. Insertion of a 250 bp MRE containing fragment into the SacII/SpeI sites of pJChAATa/b formed the constructs pIns1/1 and pIns1/2. Fusion of a 1.6 kb SpeI/KpnI fragment of pMREJCa/b to a 2.6 kb SpeI/XmnI fragment of pBShAAT and a 1.9 kb KpnI/XmnI fragment of pBSK(+) resulted in pIns1/3 and pIns1/4. The constructs pIns2/1, pIns2/2, pIns2/3 and pIns2/4 were constructed by three fragment ligation of a 1.6 kb SpeI/KpnI fragment of pMREJCa/b, a 3.8 kb SpeI/XmnI fragment of pJChAATa/b, and a 1.9 kb KpnI/XmnI fragment of pBSK(+). Plasmid vectors containing adenoviral sequences were based on pΔE1sp1A. The polylinker of pΔE1sp1A was modified by insertion of the sequences AGCTTGCGGCCGCTTACGCGGTACCT and CTAGAGCGGCCGCATACGCGGTACCA adding new restriction sites (pAd+ and pAd−). Insertion of a 3.1 kb KpnI/NotI fragment of pIns1/1 and pIns1/2 into the corresponding sites of pAd+ and pAd− resulted in pAdIns1/1a, pAdIns1/1b, pAdIns1/2a, and pAdIns1/2b. Fusion of a 4.3 kb KpnI fragment of pIns1/3, pIns1/4, pIns2/1, pIns2/2, pIns2/3, and pIns2/4 to the KpnI site of pAd+ in both orientations produced the constructs pAdIns1/3a, pAdIns1/3b, pAdIns1/4a, pAdIns1/4b, pAdIns2/1a, pAdIns2/1b, pAdIns2/2a, pAdIns2/2b, pAdIns2/3a, pAdIns2/3b, pAdIns2/4a, pAdIns2/4b. Insertion of a hAAT cDNA containing 1.6 kb XbaI/XhoI fragment into the corresponding sites of pAd+ resulted in pAdhAATa. Insertion of hAAT cDNA containing 1.6 kb KpnI/HindIII fragment from pBShAAT into the corresponding sites of pAd− resulted in pAdhAATb. pAdMREa was constructed by fusing the MRE containing 400 bp ClaI/SpeI fragment of pSLMRE to pAdhAATa opened with ClaI/XbaI. Ligation of a 600 bp MRE containing HindIII/SalI fragment to pAd+ opened with KpnI and to a hAAT cDNA containing 1.6 kb Kpn/HindIII fragment resulted in pAdMREb. The construction of pAd/RSVhAAT was described earlier.42

Adenovirus vectors

All adenovirus vectors were generated by homologous recombination with pJM17 (Microbix, Toronto, Ontario, Canada) in 293 cells. The shuttle vectors pAdIns1/1a, pAdIns1/1b, pAdIns1/3a, pAdIns1/3b, pAdIns2/1a, pAdIns2/1b, pAdIns2/2a, pAdIns2/2b, pAdhAATa, pAdhAATb, pAd.MREa and pAdMREb were cotransfected with pJM17 into low passage 293 cells by calcium phosphate coprecipitation as previously described.15 The plaque titers of all viruses were determined on 293 cells. The presence of replication-competent adenovirus and contamination with endotoxin in virus preparations was excluded by tests described earlier.53 Viruses with a titer of 5 × 1011 p.f.u./ml were stored at −80°C in 10 mM Tris-Cl (pH 8.0)–1 mM MgCl2–10% glycerol.

Tissue culture

293 (Human embryonic kidney cells) and 208f (rat fibroblasts) cells were cultured in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (Gibco, BRL, Grand Island, NY, USA) and Pen/Strep. FBS was pretreated to eliminate potentially interfering heavy metal traces by filtration through Chelex 100 resin (BioRad, Hercules, CA, USA). Fifty ml of FBS were applied to a column matrix formed of 5 g resin. The procedure was repeated five times. To evaluate the promoter activity in vitro, expression constructs were transfected into 293 cells using Qiagen superfect transfection reagent (Qiagen, Valencia, CA, USA). 106 Cells were seeded in six-well dishes and 24 h later, 2.5 μg expression plasmid were cotransfected with 0.25 μg pCMVlacZ. Six hours after transfection, 150 μM ZnSO4 was added to a subset of culture dishes to induce the MRE promoter. Forty-eight hours after transfection the concentration of hAAT in the supernatant was evaluated by ELISA. Transfection efficiencies were determined by quantification of β-galactosidase activity in cell lysates using a chemiluminescent β-gal reporter gene assay (Boehringer Mannheim, Mannheim, Germany). Induction studies with Ad vectors could not be performed in E1A expressing 293 cells due to development of cytopathic effects. To analyze the promoter activity within viral vectors, 5 × 105 208f cells were seeded in six-well dishes 24 h before infection. Cells were infected in 0.5 ml culture media containing adenovirus (MOI 100). Six hours after infection, 150 μM ZnSO4 was added to a subset of cells to induce the MRE promoter. One hundred per cent transduction efficiency under these conditions was verified by infection with Ad.β-Gal and X-gal staining 48 h after infection. Forty-eight hours after infection the concentration of hAAT in the supernatant was analyzed by ELISA.

Animals

Animal studies were performed in accordance with the institutional guidelines set forth by the University of Washington. All animals were housed in SPF facilities. Four- to 5-week-old female C57/BL6 mice were purchased from Jackson Laboratories (Bar Harbor, ME, USA). For in vivo transduction experiments 2 × 109 p.f.u. adenovirus diluted in 200 μl DMEM were injected via tail vein infusion. Blood samples were obtained by retroorbital bleeding and hAAT concentrations were determined by ELISA. CdCl2 triggered expression was induced by intraperitoneal injection of 1 μg CdCl2 per gram bodyweight 20 and 6 h before the first bleeding time-point. ZnSO4-induced expression was mediated by supplementing drinking water with 25 mM ZnSO4 for 7 days per induction round. Mice were supplied with regular drinking water for 10 days between induction rounds.

ELISA

hAAT concentrations were determined by ELISA as previously described.42 The detection limit of the assay was 500 pg/ml. Culture supernatants were used undiluted for hAAT detection. HAAT concentration in serum samples represents the weighted average of different dilutions from every sample.