Mutation of two glutamyl residues in the DNA-binding domain results in increased tyrosine phosphorylation of STAT1
In an effort to identify DNA-binding mutants of STAT1 with preserved GAS recognition, we performed a mutational study on the STAT1 molecule and generated numerous point mutants in the DNA-binding domain. A critical glutamic acid residue at position 411 in the full-length protein was found to be conserved in STAT1, STAT2, STAT3 and STAT4 of the human STAT family. Structural data of the DNA-bound STAT1 dimer revealed that the carboxyl group of E411 has a distance of 5.7 Å from the phosphodiester backbone of the co-crystallized DNA double helix and that there is no other residue in the STAT1 molecule to prevent its free access to DNA (Figure
A). This exposed residue on the surface of the DNA-binding domain was mutated to alanine and the corresponding mutant was expressed in HeLa and STAT1-negative U3A cells by transfection with pSTAT1-GFP. STAT1-E411A was normally expressed and no indication of structural instability was detected neither by Western blotting (Figure
B,C) nor gelshift experiments (Figure
D). In response to stimulation of cells with interferon−γ, the E411A mutant was tyrosine-phosphorylated (Figure
B,C) and bound to a single optimal GAS site in the M67 probe similar to the wild-type protein (Figure
Figure 1 Substitution of two residues in the DNA-binding domain of STAT1 transcription factor leads to increased tyrosine phosphorylation and prolonged GAS binding. (A) Localization of critical glutamyl residues 411 and 421 in the DNA-binding domain adjacent to (more ...)
We then performed kinetic studies on tyrosine dephos-phorylation in IFNγ−prestimulated U3A cells which were subsequently exposed to 500 nM of the potent ATP-competitive kinase blocker staurosporine
]. Treatment with the kinase inhibitor resulted in a rapid and complete dephosphorylation of wild-type STAT1 within 15 min, while the E411A mutant exhibited a much lower dephosphorylation rate (Figure
C). Moreover, the ratio of tyrosine-phosphorylated STAT1 to the total intracellular STAT1 pool, which also contained unphosphorylated protein, was elevated in the mutant as compared to its wild-type counterpart (Figure
B,C). Similarly, mutation of another glutamic acid residue in position 421, which also points with its side chain in the direction of the DNA double helix (Figure
A), resulted in defective dephosphorylation and increased DNA-binding activity (Figure
When we tested for cooperative DNA binding resulting from the ability to form stable tetramers on tandem GAS sites by means of EMSA analysis, we found no significant difference between the wild-type and mutant STAT1 (Figure
F). Both variants bound independently to either GAS site, resulting in both fast migrating STAT1/DNA complexes containing a single STAT1 dimer and slow migrating complexes with two dimers. When such complexes were challenged with a 750-fold molar excess of unlabeled M67 duplex oligonucleotides, the tetrameric complexes resisted displacement due to stable tetramerization. In contrast, the dimeric complexes of both wild-type and mutant STAT1 were either totally or partially displaced, indicative of cooperative DNA binding. Thus, substitution of either of the two conserved glutamyl residues in position 411 or 421 of the full-length STAT1 molecule critically impaired the continuous dephosphorylation/rephosphorylation cycle and resulted in elevated and prolonged tyrosine phosphorylation levels. However, binding to an optimal GAS site as well as cooperative DNA binding due to tetramer stabilization was unaltered.
Tyrosine-phosphorylated STAT1-E411A protects co-expressed endogenous STAT1 from inactivation
The partial insensitivity of STAT1-E411A towards the inhibitory effect of staurosporine was independent of the cell type, as prolonged tyrosine phosphorylation was also detected in HeLa cells (Figure
A). Similar to the effect in U3A cells, stimulation with an equal concentration of IFNγ resulted in higher levels of tyrosine-phosphorylated mutant STAT1 as compared to the wild-type. Also in cytokine-stimulated HeLa cells, the ratio of tyrosine-phosphorylated STAT1 to the total STAT1 was increased, indicating that hyperphosphorylation reflects an inherent property of the mutant. In line with the altered kinetics of tyrosine phosphorylation, we found that, also in HeLa cells, DNA-binding activity to the M67 site was enhanced following 45 min of stimulation with IFNγ (Figure
B). Moreover, in the presence of staurosporine the rate of dephosphorylation was decreased in the point mutant as compared to the wild-type, thus confirming that the mutant E411A displayed a prolonged state of DNA binding (Figure
Figure 2 Co-expressed endogenous STAT1 is protected from inactivation through heterodimer formation with STAT1-E411A. (A) Kinetics of STAT1 dephosphorylation in HeLa cells expressing green fluorescent protein-tagged wild-type or mutant STAT1. Cells were stimulated (more ...)
Interferon-prestimulated HeLa cells expressing endogenous STAT1, in addition to either the GFP fusion of wild-type STAT1 or its GFP-tagged mutant, were subjected to the inhibitory effect of staurosporine. In cells expressing STAT1-E411A-GFP, not only did the mutant phospho-protein resist staurosporine treatment much better, endogenous STAT1 was also partially insensitive, as revealed by its prolonged tyrosine phosphorylation (Figure
A) and enhanced DNA-binding activity (Figure
B). Thus, the presence of the E411A substitution protects also co-expressed native STAT1 protein from its rapid inactivation. This finding suggested that the mutant STAT1 protein interacts with endogenous STAT1 in a way that impairs access to the inactivating nuclear phosphatase.
Diminished nuclear export of tyrosine-phosphorylated STAT1-E411A
We then tested whether the nucleocytoplasmic distribution differed between wild-type and the mutant (Figure
A). Cytosolic and nuclear extracts were prepared from either unstimulated or IFNγ−stimulated HeLa cells expressing STAT1-GFP fusion proteins and the levels of tyrosine phosphorylation were subsequently probed by means of Western blotting. It was found that, in nuclear extracts, the amount of phospho-STAT1 was significantly higher for mutant STAT1 as compared to the wild-type, and vice versa, in cytosolic extracts there was slightly more phosphorylated wild-type protein. Thus, the concentration of phospho-STAT1 in the nucleus was higher when the critical glutamyl residue was displaced by alanine, resulting in a more pronounced nuclear retention. Again, the amount of endogenous phospho-STAT1 was higher in HeLa cells expressing the E411A mutant as compared to its wild-type GFP fusion.
Figure 3 Mutation of E411 results in diminished nuclear export of STAT1. (A) Altered nucleocytoplasmic distribution of the E411A mutant. HeLa cells expressing GFP-tagged STAT1 variants were either unstimulated (−) or stimulated with IFNγ (+), as (more ...)
To confirm the altered nucleocytoplasmic shuttling properties of the mutants by a different approach, we performed a permeabilized cell transport assay
B). HeLa cells expressing GFP-tagged wild-type STAT1 or the respective glutamyl mutants were stimulated for 45 min with IFNγ to induce nuclear accumulation of the recombinant fusion proteins. Subsequently, the cells were either immediately fixed or incubated for 6 min with 50 μg/ml digitonin on ice before fixation. Treatment with digitonin at this concentration selectively permeabilized the plasma membrane, thereby, releasing cytoplasmic proteins, while the integrity of the nuclear envelope remained intact. As expected, stimulation with IFNγ resulted in the nuclear accumulation of all GFP-tagged STAT1 variants (Figure
B, top panel). However, permeabilization by digitonin completely abrogated the pre-existing nuclear presence of STAT1-WT-GFP, while the two mutants remained accumulated in the nucleus (Figure
B, bottom panel). Thus, the nuclear export rate of the mutants was critically reduced as compared to the wild-type protein.
By adding a transferable nuclear export signal (NES) to GFP-tagged STAT1
], we have gathered further evidence for an altered DNA binding of the mutants. In resting cells, STAT1-NES-GFP showed a cytoplasmic redistribution as compared to the nearly pancellular localization of STAT1-GFP, which resulted from enhanced nuclear export (Figure
C). Also in contrast to STAT1-WT, the NES adduct failed to accumulate in the nuclei of interferon-stimulated cells, because the enhanced nuclear export rate competed with nuclear retention on DNA. Interestingly, however, nuclear accumulation was fully restored in the additional presence of the E411A mutation. This observation clearly confirms that high-affinity DNA-binding is the underlying phenotype of the E411A mutant.
The mutant E411A exhibits high-affinity GAS binding and has a broad repertoire of non-optimal binding sites
We now performed experiments that were aimed at elucidating the molecular basis behind the altered activation/inactivation cycle of the two STAT1 glutamyl mutants. Putative mechanisms for hyperphosphorylation of STAT1 variants include diminished nuclear import due to mutations in either the dimer-specific nuclear import signal or other regions of the STAT1 molecule, which interact with importin-α, as well as altered binding kinetics to DNA. STAT1 mutants with impaired nuclear import are exposed to the high kinase activity and comparably low phosphatase activity in the cytosol
], and a DNA-binding mutant termed STAT1-dnaplus
has been described, which failed to recognize GAS probes in gelshift assays
]. We found that the glutamyl mutants do not fall into either of these categories, since, upon cytokine stimulation of cells, the mutants were imported normally into the nucleus (Figure
A,B), thus ruling out defective nuclear accumulation as the cause for their hyperphosphorylation. Furthermore, the mutants recognized GAS elements in mobility shift assays, clearly distinguishing them from STAT1-dnaplus
, in which three other residues in the DNA-binding domain were substituted for positively charged residues (Thr327Arg, Val426His, Thr427His). The altered DNA-binding kinetics of the glutamyl mutants was evident in competition experiments employing challenge with excess unlabeled GAS oligonucleotides (Figure
A,B). These experiments clearly revealed that a dramatically reduced dissociation rate from DNA constitutes their underlying phenotype. In the mutants, the release from optimal DNA-binding sites was critically impaired, resulting in a longer half-life of GAS-bound dimers as compared to wild-type STAT1. Thus, the stability of preformed protein-DNA complexes differed significantly between the two mutant STAT1 proteins and their wild-type counterpart.
Figure 4 Replacement of glutamyl residues in position 411 or 421 results in STAT1 mutants with a decreased dissociation rate from DNA and high-affinity binding to non-canonical binding sites. (A) Comparison of the dissociation rates between STAT1-WT and STAT1-E411A (more ...)
In order to compare the sequence requirement for DNA binding between the E411A mutant and wild-type STAT1, we used non-optimal GAS elements as molecular probes in mobility shift assays (Figure
C). Both the wild-type and the mutant bound with high affinity to oligonucleotides containing a single GAS site. However, STAT1-E411A also reacted with a mutated probe which, due to the exchange of two base pairs, contained no consensus GAS element. Although binding to this 2x nonGAS probe was weaker than to either GAS-nonGAS or tandem GAS oligos, there was a detectable formation of DNA-bound STAT1 dimers not requiring an intact GAS site for DNA binding. Thus, in the presence of excess unlabeled GAS oligos, the E411A mutant bound to DNA not only with a higher affinity than the wild-type molecule, but also showed a relaxed sequence requirement for interaction with DNA.
dephosphorylation assays, using whole cell extracts from reconstituted U3A cells in the presence of the STAT1-inactivating Tc45 phosphatase, confirmed that the two glutamyl mutants are indeed DNA-binding mutants (Figure
D). It has been shown that DNA-bound STAT1 is protected from dephosphorylation and barred from nuclear exit
], and we report here that the glutamyl mutants but not the wild-type protein resisted Tc45-catalyzed inactivation. These experiments collectively demonstrate that there must be a considerable amount of mutant phospho-STAT1 interacting with genomic DNA that does not participate in nucleocytoplasmic shuttling and resists inactivation by nuclear phosphatases.
A low dissociation rate from DNA results in prolonged cytokine-induced nuclear accumulation
The experiments presented thus far have shown that mutation of two critical glutamyl residues in the DNA-binding domain results in high-affinity DNA binding and defective tyrosine dephosphorylation of STAT1 upon stimulation of cells with IFNγ. Therefore, we wondered whether the resting distribution and the kinetics of cytokine-inducible nuclear accumulation differed between the mutant and wild-type STAT1 variants. For these experiments, we additionally mutated the glutamyl acid residues at positions 411 and 421 in positively charged lysyl residues and found that the resulting two novel point mutants closely mimicked the corresponding alanine mutant as described above (data not shown). The GFP fusion proteins of all three STAT1 variants (wild-type, E411K, and E421K, respectively) demonstrated a similar localization in resting HeLa cells, namely a pancellular distribution with a slightly elevated cytoplasmic concentration (Figure
A). Replacement of the native glutamic acid residues at position 411 and 421 was without effect on the cytokine-induced nuclear accumulation, since the tyrosine-phosphorylated GFP fusions were imported normally into the nuclear compartment. However, when IFNγ−prestimulated cells were subsequently treated for 60 min with the kinase inhibitor staurosporine, a striking difference between the two point mutants and wild-type STAT1 was detected. In HeLa cells expressing wild-type STAT1, staurosporine caused a rapid collapse of nuclear accumulation, while nuclear localization of the glutamyl mutants persisted despite the presence of staurosporine. Thus, both point mutations significantly retarded the nuclear residence time of STAT1, but did not completely prevent the collapse of nuclear accumulation, since after 120 min of staurosporine exposure the former resting distribution of STAT1 was again achieved (data not shown). Thus, not surprisingly, insensitivity to pharmacological kinase inactivation resulted not only in elevated levels of tyrosine-phosphorylated STAT1 (Figures
), but also in a markedly prolonged phase of nuclear accumulation. Additionally, we found that, in the absence of staurosporine, the nuclear retention time was considerably prolonged for the mutant STAT1 proteins during IFNγ−induced stimulation (data not shown).
Figure 5 STAT1 DNA-binding mutants with decreased dissociation from DNA exhibit a prolonged nuclear accumulation phase. (A) Localization of GFP-tagged STAT1 variants in HeLa cells expressing either wild-type or mutant STAT1, in which glutamic acid residues at (more ...)
To exclude the possibility that the differential nuclear accumulation kinetics seen for the glutamyl mutants is an artefact resulting from the presence of the GFP domain, we confirmed this finding by means of immunocytochemical staining in U3A cells expressing recombinant, untagged STAT1 (Figure
B). Similarly to the GFP adducts expressed in HeLa cells, the respective glutamyl mutants showed an unaltered resting distribution and accumulated normally in the nuclei of IFNγ−stimulated U3A cells. However, after 60 min of staurosporine addition to the cells, the mutant STAT1 molecules were still predominantly localized in the nucleus, whereas the resting distribution of the wild-type protein had already been restored at that time point. Following 90 min of staurosporine exposure, the nuclear accumulation of both mutants had also collapsed, demonstrating that the DNA-binding mutants were less sensitive to kinase inhibition. This finding in U3A cells confirmed that the reduced dephosphorylation rate and prolonged nuclear accumulation are inherent properties of the glutamyl mutants, which result directly from their slow off-rate from DNA.
High-affinity DNA binding crucially impairs transcriptional responses
The impact of high-affinity DNA binding on gene transcription was next investigated. Reporter gene assays were performed to assess the consequence of a decreased dissociation rate from DNA on gene expression. Using a luciferase reporter with a synthetic promoter containing three strong GAS sites separated by 10 bp (3xLy6E), we found that all STAT1 variants tested displayed transcriptional responses upon stimulation of reconstituted U3A cells with IFNγ (Figure
A). However, reporter gene induction was significantly repressed in cells expressing either of the glutamyl mutants as compared to the wild-type protein. STAT1-E411K displayed the lowest reporter gene activation of the mutants under investigation, demonstrating that the transcriptional activity decreased from wild-type > E411A > E411K (p = 0.049 and p < 0.001, respectively). Similar results were also obtained for STAT1-E421K using two reporters containing native fragments from the ICAM-1 promoter, termed pIC339 and pIC1352 (Figure
A). Thus, exchange of the negatively charged glutamyl acid residue at position 411 for either a neutral or positively charged amino acid stepwise diminished the transcriptional response on a reporter gene with a strong cytokine-driven promoter.
Figure 6 Suppressed transcriptional responses of STAT1 mutants with high-affinity GAS binding. (A) Decreased reporter gene activation in U3A cells expressing STAT1 DNA-binding mutants as compared to wild-type protein. U3A cells were transiently transfected with (more ...)
We then used real-time RT-PCR assays to probe the induction of three endogenous IFNγ−responsive genes in transfected U3A cells. Again, the mutants failed to reach the transcriptional activity of the wild-type protein (Figure
B). While constitutively expressed recombinant stat1 mRNA was detected in all samples as expected (p > 0.05), there was a significant reduction in irf1 mRNA synthesis in U3A cells expressing the E411K and E421K mutant as compared to wild-type STAT1 (p = 0.029 and p = 0.003, respectively). Induction of the gbp1 and mig1 gene was also critically impaired or even completely abolished by replacing either of the glutamyl residues (for all comparisons p < 0.03).