W. H. Lamers, AMC Liver Center, Academic Medical Center, University of Amsterdam, Meibergdreef 69-71, 1105 BK, Amsterdam, the Netherlands. Fax: + 31 20 5669190, Tel.: + 31 20 5665948, E-mail: firstname.lastname@example.org
Glutamine synthetase (GS) is expressed at high levels in subsets of cells in some tissues and at low levels in all cells of other tissues, suggesting that the GS gene is surrounded by multiple regulatory elements. We searched for such elements in the 2.5-kb upstream region and in the 2.6-kb first intron of the GS gene, using FTO-2B hepatoma and C2/7 muscle cells as representatives of both cell types and transient transfection assays as our tools. In addition to the entire upstream region and entire intron, an upstream enhancer module at −2.5 kb, and 5′, middle and 3′ modules of the first intron were tested. The main effects of the respective modules and their combinatorial interactions were quantified using the analysis of variance (anova) technique. The upstream enhancer was strongly stimulatory, the middle intron module strongly inhibitory, and the 3′-intron module weakly stimulatory in both hepatoma and muscle cells. The 5′-intron module was strongly stimulatory in muscle cells only. The major new finding was that in both cell types, the upstream enhancer and 5′-intron module needed to be present simultaneously to fully realize their transactivational potencies. This interaction was responsible for a pronounced inhibitory effect of the 5′-intron module in the absence of the upstream enhancer in hepatoma cells, and for a strong synergistic effect of these two modules, when present simultaneously in muscle cells. The main difference between hepatoma and muscle cells therefore appeared to reside in tissue-specific differences in activity of the respective regulatory elements due to interactions rather than in the existence of tissue-specific regulatory elements.
Glutamine synthetase (GS; EC 22.214.171.124), the enzyme that catalyses the ATP-dependent conversion of glutamate and ammonia into glutamine, is expressed in a tissue-specific and developmentally controlled manner. GS functions to remove ammonia or glutamate, or to produce glutamine. Cells that function primarily to remove glutamate or ammonia, contain very high GS levels (30–160 µm), whereas cells that synthesize glutamine contain much lower levels (1–8 µm) . Another highly characteristic and functionally important feature of GS is its topographic distribution: in organs in which GS is present at relatively high concentrations, it is usually expressed in a subset of cells only, whereas in organs in which it is present at low concentrations, it is expressed in the majority of cells. Examples of the first group of organs are the pericentral hepatocytes in the liver, the astrocytes in nervous tissue, the epithelial cells of the caput epididymis, and the gastric antrum. Examples of the second group are adipocytes and muscle cells (for a review, see ).
Because of these interorgan differences in distribution and cellular concentration of GS, and because only a single functional copy of the GS gene is present per haploid genome in rodents [2–4], it is to be anticipated that the regulation of GS expression is complex . Studies aimed towards delineating the transcriptional regulation of GS expression in rodents have thus far revealed upstream enhancer elements at −6.0 kb and −2.5 kb, and intron enhancer elements at +0.35 kb and +1.6 kb, by transient or stable transfections to cultured cells [5–8]. The sequence of the far-upstream mouse GS enhancer that is active in adipocytes  is 80% similar to that of the far-upstream rat GS enhancer and, like the rat far-upstream enhancer, also maps at −6.0 kb in the Celera Discovery System mouse genome database. The upstream enhancer at −2.5 kb confers pericentral localization to reporter gene expression in the liver of transgenic mice . The significance of the far-upstream and intron elements for the in vivo expression pattern of GS remains to be assessed.
In the liver, genes are expressed in a porto-central gradient. Studies with transgenic animals have shown that many of these gradients in gene expression, including that of GS , are determined at the transcriptional level. Porto-central gradients in gene expression have been distinguished into dynamic and stable gradients . The dynamic type of zonation is characterized by adaptive changes in expression in response to changes in the metabolic or hormonal state, whereas the stable type of zonation, of which GS is an example [12,13], is characterized by the virtual absence of such adaptive changes. A relatively simple model to explain such a stable expression pattern is to assume a ‘double-lock’ regulatory mechanism, meaning that GS expression depends on the synergistic interaction of two or more factors .
The observation that very high levels of GS are present in subsets of cells in some organs, and moderate-to-low levels in all cells of other organs, in combination with the hypothetical ‘double-lock’ mechanism to account for the stable expression of GS in pericentral hepatocytes suggested that multiple regulatory modules would control GS expression and that at least some of these modules would be interdependent with respect to their regulatory activity. To test this hypothesis, we examined the combinatorial effects of modular deletions in the distal upstream and first intron regions of the rat GS gene on reporter gene expression in hepatoma and muscle cells. We now report that interactions of upstream and intronic regulatory elements do indeed determine the degree of activation of the GS promoter and that these interactions differ quantitatively between cells from hepatic and muscular origin.
Materials and methods
Sequence of the first intron of GS
The nucleotide sequences of the upstream region and of the first 1938 nucleotides of the first intron of the rat GS gene were reported [5,6]. This intron sequence ends at the EcoRI restriction site (Fig. 1). The remaining 877 nucleotides of the first intron of GS were sequenced in both orientations. The sequence data has been deposited with the EMBL nucleotide sequence data bank and is available under accession number AF170107.
Construction of plasmids
Rat GS genomic DNA sequences were cloned into the vector pSPluc+ (Promega). The starting construct (Fig. 1, construct Q) was made by inserting the genomic GS segment from −2520 bp (relative to the transcription start site) to +2774 bp (corresponding to position +132 in the GS cDNA, that is, the translation start site in the second exon) between the HindIII and NcoI sites in the polylinker upstream of the luciferase cDNA. The 305 bp bovine growth-hormone polyadenylation sequence (the XbaI–PvuII fragment from pcDNA3; Invitrogen) was inserted between the XbaI and EcoRV sites in the polylinker downstream of the luciferase cDNA. The other constructs were generated by modular deletions of construct Q. The upstream modules were: the entire region downstream of HindIII (−2520), the region downstream of EcoRV (−2148), the region downstream of PstI (−965), or the upstream enhancer element (−2520 to −2148 ). The first intron was subdivided into three modules: a 5′SmaI–BamHI fragment (+153 to +856), a middle BamHI–SmaI fragment (+856 to +1791) and a 3′SmaI–BamHI fragment (+1791 to +2712). The extended GS promoter (BglII (−368) to the transcription-start site , the first exon, a minimized first intron, containing its 5′-most portion (+119 to +153) and 3′-most portion (+2712 to +2760), and the second exon up to the translation start site (+132) were present in all constructs. The construct carrying only these elements (construct A) was used as reference construct.
For the transfection assays, the respective constructs were purified in CsCl gradients or on Nucleobond columns (Machery-Nagel, Düren, Germany).
FTO-2B rat hepatoma cells  were cultured in DMEM/F-12 medium (Gibco), supplemented with 10% (w/v) foetal bovine serum (Gibco). C2/7 cells (generously provided by M. Buckingham, Institut Pasteur, Paris, France) are a subclone of the C2 cell line that was originally derived from Soleus muscle of adult C3H mice . These cells were cultured in DMEM (Gibco), supplemented with 10% (w/v) foetal bovine serum. All cells were cultured at 37 °C in humidified air containing 5% CO2. Cell lines were tested monthly for contamination with mycoplasms.
Exponentially growing FTO-2B cells were transfected by electroporation  and C2/7 cells were transfected at the myoblast stage by the calcium-phosphate method . In both cases 20 µg supercoiled plasmid was used. Cotransfection with 5 µg of the vector pRSVcat  enabled correction for differences in transfection efficiency. After electroporation, the cell suspension was divided into two equal parts, one being grown in culture medium and the other in culture medium supplemented with 100 nm dexamethasone. Sixteen h after transfection, the cells were washed with NaCl/Pi and given fresh medium. In the case of the C2/7 cells, the concentration of foetal bovine serum was reduced to 1% to induce the formation of myotubes. Harvest of the cells was carried out 48 h after transfection in the case of FTO-2B cells, and 72 h in the case of C2/7 cells, that is, when all cells were fused into myotubes. Cells were lysed in 100 mm KH2PO4/K2HPO4 pH 7.6, 0.1% (v/v) Triton X-100 buffer and tested for chloramphenicol acetyltransferase activity , luciferase activity  and protein concentration (bicinchoninic acid reagent; Pierce).
Splicing of modified first intron
Constructs carrying the different modules of the GS first intron were tested for proper splicing of the mRNA. RT-PCR was carried out with primers in the first exon (+1 to +18) and in the luciferase-coding region (+78 to +60 in the luciferase cDNA). All constructs generated correctly spliced mRNAs (data not shown).
Correction for experimental variation and statistics
The transactivation potential of the tested DNA constructs was expressed as the ratio between their luciferase activity (light units·mg protein−1) and the chloramphenicol acetyltransferase activity (units·mg protein−1) of the cotransfected pRSVcat construct. The data were collected from 33 experiments with FTO-2B cells and 13 experiments with C2/7 cells. In each experiment, different combinations of constructs were tested. The number of transfections per construct was 8–20. Interexperimental variation in reporter gene activity was removed using log-transformed values and the general linear model/anova without interaction (spss version 10.0.7; SPSS Inc.).
The activity of a specific construct (X,Y) containing upstream module (X) and intron module (Y) can be modelled to consist of the sum of a basal activity (produced by the promoter and minimized intron), the effects of the respective upstream (X) and intron (Y) modules, and the interaction (X,Y) between these modules:
In this model, the value of the main effects of the upstream and intron modules, and that of their interactions can be calculated with an approach based on the analysis of variance (anova) technique. To normalize the data, the activity of reference construct A was set to 100 arbitrary units (AU) in these calculations. The activity of the respective modules and their interactions, including 95% confidence intervals, was expressed relative to construct A. Whenever a difference is mentioned in the text, it is significant at the 5% level.
We based the design of our analysis of the regulatory properties of sequences in the upstream region and within the first intron of the rat GS gene on the assumption that two or more interdependent regulatory elements were responsible for transactivation of the GS promoter . For this reason, the study was designed to reveal which DNA sequences do interact with respect to transactivation of this promoter. We also wished to avoid that changes in the position of the regulatory sequences might affect the regulatory behaviour of the DNA modules. For that reason, the experimental constructs were made by modular orthotopic additions to construct A (Fig. 1).
To delineate regulatory elements upstream of the GS structural gene, the upstream sequence present in construct A was extended to −965 nucleotides (construct B), to −2148 nucleotides (construct C), or to −2520 nucleotides (construct M) (Fig. 1, panel I). None of these upstream modules significantly enhanced the activity of the GS promoter in either FTO-2B hepatoma cells or C2/7 myotubes. Previous experiments had shown that the upstream region was able to transactivate the heterologous thymidine kinase (TK) promoter, and that this activity was localized between −2520 and −2148 bp . When placed directly in front of the GS promoter, this distal upstream enhancer element (construct H, Fig. 1, panel III) caused a small but significant increase in reporter gene expression (1.4-fold in FTO-2B and 1.8-fold in C2/7 cells).
The transactivational capacity of the first intron of the GS gene was tested as such (construct G), as a 703-bp 5′ intron module (construct D), a 935-bp middle intron module (construct E), a 921-bp 3′ intron module (construct F), or after deleting all intron sequences except 35 nucleotides at the 5′ end and 48 nucleotides at the 3′ end (construct A, the reference construct) (Fig. 1, panel II). In FTO-2B cells, the entire intron (construct G) decreased reporter gene activity significantly to 40% of that of construct A. The inhibitory activity was found to reside in the 5′ and middle intron fragments (constructs D and E). In muscle cells, the entire intron depressed reporter gene activity to 50% of that of reference construct A. When the intron fragments were tested individually, the 5′ intron fragment (construct D) was without effect on the promoter, whereas the middle fragment slightly decreased reporter gene activity (to 70%) and the 3′ fragment (construct F) stimulated promoter activity 1.9-fold.
Interactions between upstream and intron regulatory modules
When different combinations of upstream and intron sequences were tested for transactivation of the GS promoter, the highest activities were observed for constructs containing the upstream enhancer (constructs H-L), whereas the lowest activities were consistently associated with the presence of the middle intron module (constructs E, J and O) (Fig. 1, panels II–IV). The effect of partner choice appeared to matter most for constructs containing the 5′-intron module (constructs D, I and N). These findings demonstrated that the degree of transactivation of the GS promoter depended to a substantial degree on interactions between the upstream and intron regulatory sequences. We therefore used an approach that is based upon the anova technique to quantify the main (that is, ‘intrinsic’) effects of upstream and intron modules, and to segregate these effects from those due to interaction between the respective elements. In this approach, the activity of construct A was set at 100 AU.
FTO-2B hepatoma cells
The computation of the main effects revealed that the presence of the upstream enhancer increased promoter activity with 87 AU in hepatoma cells, whereas this number was slightly lower (67 AU) for the entire upstream region (Fig. 2, upper panel). Both effects were statistically significant. The 5′- and 3′-intron modules both increased promoter activity with 23 AU. The middle intron module decreased promoter activity with 84 AU. Although the effect of the entire intron on promoter activity was not significant (14 AU), it neutralized the negative effect of its middle fragment. The actual activity of the respective constructs often resulted from less than additive effects between the upstream and intron modules. Such negative interactions were observed for constructs containing the upstream enhancer, but lacking the 5′-intron fragment (constructs H, J and K), and vice versa (constructs D, G). Furthermore, the effects of the upstream region and the entire intron were not additive (construct Q). The other combinations did not show significant interactions, meaning that the main activities of their components accounted for the observed effect. These findings demonstrate that the simultaneous presence of the upstream enhancer and the 5′-intron module is necessary for full transactivation of the GS promoter in hepatoma cells.
C2/7 muscle cells
The upstream enhancer significantly increased promoter activity in muscle cells with 121 AU, whereas this number was not significant (13 AU) for the entire upstream region (Fig. 2, lower panel). The intron modules all had significant effects on the promoter: the 5′- and 3′-intron modules increased promoter activity with 127 AU and 47 AU, respectively, whereas the middle intron module decreased promoter activity with 58 AU. The entire intron did not affect promoter activity significantly. The inhibitory effect of the middle intron module therefore outweighed the strongly stimulatory effects of the 5′- and 3′-intron modules. The interaction between the upstream enhancer and the 5′-intron fragment produced a more than additive transactivational effect on the promoter. Similar to liver cells, the stimulatory effect of either element was largely lost if the other element was absent. Furthermore, and again similar to liver cells, the effects of the upstream region and the entire intron were not additive. These findings show that the upstream enhancer and the 5′-intron modules are mutually dependent for full activity of the GS promoter in both liver and muscle cells.
Comparison of FTO-2B with C2/7 cells
The comparison of both cell lines revealed that the main activities of the upstream enhancer and 5′-intron modules, as well as their interaction, were higher in C2/7 cells than in FTO-2B cells. Both cell types resembled each other in that the upstream enhancer and 5′-intron module had to be simultaneously present for the highest level of reporter gene expression, whereas significant negative interactions were observed if either element was absent. Apparently as a result of the latter effect the upstream enhancer increased the inhibitory effect of the middle intron module, but did not support the stimulatory effect of the 3′-intron module in both cell types. In C2/7, but not in FTO-2B cells, the upstream enhancer and the 5′-intron modules lost their stimulatory activity in the context of the upstream region and the entire intron, respectively. As the presence of both the entire upstream region and the entire intron negatively affected promoter activity in both cell types, the 5.3-kb region encompassing the entire upstream region and first intron, was threefold less active in muscle than in hepatoma cells. The differences between hepatoma and muscle cells therefore can be explained by tissue-specific differences in activity of the respective regulatory elements due to interactions rather than in the use of distinct, tissue-specific regulatory elements.
Glucocorticoid sensitivity of the regulatory sequences
All modules were tested for sensitivity to glucocorticoid treatment. Only constructs containing the middle intron module showed a threefold induction of reporter gene activity when tested in C2/7 cells (data not shown). In hepatoma cells, no effects of the hormone were observed.
We have studied the capacity of the upstream region and first intron of the GS gene to transactivate its promoter, as well as the effect of interactions between these regions on GS promoter activity. Furthermore, we aimed to determine if any of the regulatory elements in these regions behaved differently in cells which can express high levels of GS (hepatocytes) and in cells which do express low levels of the gene (muscle). The application of the anova technique allowed us to segregate the main (intrinsic) activities of the respective elements from the effects of their interactions. Using this approach, the upstream enhancer was identified as a strongly stimulatory element in both hepatoma and muscle cells, whereas the 5′-intron module was strongly stimulatory in muscle cells only. In both cell types, however, the upstream enhancer and 5′-intron module depended on each other for effective transactivation of the GS promoter. Other intriguing findings were that the upstream enhancer lost most of its activity when present in the context of the entire upstream region in muscle cells, but not in hepatoma cells. Furthermore, the inhibitory effect of the middle intron module appeared to be constitutive in hepatoma cells, but dependent on glucocorticoids in muscle cells. Apparently, both positive and negative elements, and extensive interactions between them, regulate GS promoter activity.
In addition to transcriptional control, translatability of the GS mRNA and stability of the GS protein appear to be important post-transcriptional levels of control [1,22]. In fact, we did show that it is necessary to consider these post-transcriptional control levels when analysing the expression of GS in transgenic animals [23,24]. However, as both the reporter gene and the transcription termination and polyadenylation sequences that were used are not normally expressed in either liver or muscle, post-transcriptional control is an unlikely level of control to explain the observed differences between hepatoma and muscle cells.
The similarity of the activity of constructs A, B and C argues against the presence of an inhibitory sequence within the upstream region. Nevertheless, the activity of the upstream enhancer in muscle cells and, to a lesser extent, in hepatoma cells, is clearly mitigated in the context of the entire upstream region, i.e. by the interposition of 1780 bp. We interpret the increase in activity of the upstream enhancer when positioned in close proximity to the promoter as the consequence of a distance effect (see [25,26]): apparently, the upstream enhancer has difficulty contacting the promoter when the 1780 bp intervene. We have previously observed such a distance effect for the carbamoylphosphate synthetase enhancer in conjunction with the TK promoter, but not with the carbamoylphosphate synthetase promoter alone .
The 5′ and middle intron modules of the first intron of GS correspond tp two DNaseI-hypersensitive sites . Three studies [6–8] have analysed the enhancer activity of these modules in conjunction with the heterologous TK promoter . The 5′ intron module behaves as a conditional enhancer element when positioned downstream of the promoter (this study), but as a constitutive stimulatory element when tested upstream of the TK promoter . In this configuration, the activity of the 5′ intron element resided between positions +153 and +627 . In contrast, the strong and consistently inhibitory effect of the middle intron module on GS promoter activity does not appear to be context sensitive, as it was also observed when tested upstream of the TK promoter . This inhibitory activity resided in a 325-bp fragment (position +1466 to +1791) and was, similar to our finding, relieved by glucocorticoids . A putative GRE (glucocorticoid-responsive element) was identified at position +1656 to +1670. The middle intron regulatory element in the mouse GS gene  was tested in stable transfection assays in a differentiating adipocyte cell line. Its core activity was found to be limited to a 310-bp fragment, the sequence of which corresponds with that of position +1450 to +1752 in the rat GS intron. This sequence was found to contain C/EBP and HNF3 consensus-binding sites at position +1580 to +1592. The middle intron regulatory module may therefore qualify as a glucocorticoid-responsive unit (see ). The presence of an inhibitory GRU (glucocorticoid-responsive unit) in the GS gene and a strongly stimulatory one in the carbamoylphosphate synthetase gene  may explain the frequently reciprocal behaviour of both genes with respect to expression .
Co-operative interactions in the binding of transcription factors to arrays of response elements within an enhancer module appear to be the rule rather than the exception. The explanation for these co-operative effects is that the binding of a factor to an element within such an array entails an increase in the affinity of adjacent elements for their corresponding transcription factors. Due to the presence of protein–protein interactions within an enhancer–promoter complex, the transcription factor-binding sites do not have to be adjacent [29,30]. However, co-operative interactions between distant enhancer modules as now reported for GS are described infrequently. Reported examples include the synergistic interaction between a far-upstream and an upstream enhancer , between an upstream and an intron enhancer [32,33], and between an intron and a downstream enhancer . Interestingly, the GS gene itself may present yet another example of an interaction between distant regulatory modules, as both the −6.0 kb far-upstream enhancer and the middle intron element enhance reporter gene activity in stably transfected adipocytes [7,9]. Notwithstanding this association, it remains to be shown that these two elements do indeed interact. Whether the cooperative interactions between distant enhancers obey the same rules as observed for elements within a single enhancer and for enhancer–promoter interactions, remains to be established.
In transgenic animals, the spatio-temporal expression pattern of a reporter gene that is driven by the GS upstream region, revealed several discrepancies between the expression of the endogenous GS gene and the reporter gene [23,24]. This finding suggested that one or more regulatory elements that were not present in this transgene, were responsible for the expression pattern of endogenous GS. Furthermore, our modelling of gene expression patterns in the liver had predicted that an interaction between at least two regulatory elements was necessary to generate the remarkably stable expression gradient of GS . The present study has identified the upstream enhancer and the 5′-intron module as two such interacting regulatory elements.