Consequences of the disease-related L78R mutation for dimerization and activity of STAT3

The signal transducers and activators of transcription (STATs) are dual function proteins which receive an activating signal from the cell surface and carry it toward the nucleus to induce gene transcription (Stark and Darnell, 2012). The STAT protein family consists of seven members (STAT1, STAT2, STAT3, STAT4, STAT5a, STAT5b and STAT6) that share a common structural organization represented by six distinct and functionally conserved domains: N-terminal domain (NTD), coiled-coil domain (CCD), DNA-binding domain (DBD), linker domain (LD), Src homology 2 domain (SH2) and the C-terminally located transactivation domain (TAD) (Fig. 1A). Numerous ligands are able to activate STAT transcription factors, including cytokines, growth factors and hormones, as well as oncogenic proteins such as mutated tyrosine kinases (Aggarwal et al., 2009). The dysregulated activation of STAT signalling is involved in chronic inflammation (O'Sullivan et al., 2007) and cancer, including blood malignancies and solid tumours (Yu and Jove, 2004).

In the canonical view of JAK-STAT signalling, cytokines are recognized by cell surface receptors that are associated with Janus tyrosine kinases (JAKs). Upon ligand binding the JAKs become activated and phosphorylate tyrosine residues of the receptor. These phosphotyrosine motifs act as docking sites for proteins containing Src homology 2 (SH2) domains, including various adaptor proteins, negative regulators, such as suppressor of cytokine signalling (SOCS) proteins and phosphotyrosine phosphatases (e.g. SHP2), or, most importantly, transcription factors of the STAT family. STATs are recruited to the receptor and are phosphorylated on a single tyrosine residue by JAKs. The phosphorylated STATs form homo- or hetero-dimers through reciprocal interactions between the phosphotyrosine motif of one and the SH2 domain of the other protomer. The dimerized STAT molecules translocate to the nucleus, bind to specific response elements and activate target gene expression (Levy and Darnell, 2002).

In recent years, several publications have reported on preformed (latent, unphosphorylated) STAT dimers (Haan et al., 2000; Kretzschmar et al., 2004; Schröder et al., 2004), STAT tetramers at specific gene promoters (Zhang and Darnell, 2001), but also on conglomerates of higher molecular masses such as statosomes (Ndubuisi et al., 1999), nuclear bodies (Herrmann et al., 2004) or paracrystals (Droescher et al., 2011). Moreover, biological functions of STATs independent of tyrosine phosphorylation have been described. These non-canonical functions have been demonstrated to be involved in controlling heterochromatin and microtubule stability or regulating metabolic functions in mitochondria (Mohr et al., 2012).

STAT3 has been identified as a key player in chronic inflammation, autoimmune diseases and tumorigenesis. In contrast to the cytokine-induced transient activation, STAT3 is persistently activated in a variety of human cancers, mostly as a consequence of deregulated upstream signalling pathways (Yu and Jove, 2004). More recently, mutations targeting structural domains of STAT3 have been associated with some diseases, such as hyper-IgE syndrome (HIES) (Holland et al., 2007), large granular lymphocytic leukaemia (LGL leukaemia) (Koskela et al., 2012) or inflammatory hepatocellular adenoma (IHCA) (Pilati et al., 2011).

Here, we used Förster resonance energy transfer (FRET) to study the dimers of STAT3 before and after activation, to determine the orientations of the protomers in these dimers and to analyse the structural and functional consequences of the somatic L78R mutation found in IHCA.

To investigate the homodimerization of STAT3 with FRET imaging we used eGFP (enhanced green fluorescent protein) as a donor fluorophore and TMRstar coupled to the SNAP tag as an acceptor fluorophore. The target protein was labelled N- or C-terminally with the donor or with the acceptor fluorophore resulting in four different fusion protein constructs (Fig. 1A).

To study the expression and functional properties of the STAT3 fusion constructs, HeLa cells were transfected with plasmids encoding SNAP–STAT3, STAT3–SNAP, eGFP–STAT3 or STAT3–eGFP. Cells were stimulated with IL-6 and soluble IL-6 receptor for 30 minutes or left untreated and lysates were analysed by immunoblotting (Fig. 1B). After cytokine stimulation, all constructs were phosphorylated at tyrosine residue 705, demonstrating that labelling with eGFP or SNAP tag (N- or C-terminally) does not interfere with the phosphorylation of the fusion protein, in good agreement with previously published data from our laboratory (Herrmann et al., 2004) and from others (Huang et al., 2007). For further functional analysis, the tagged STAT3 proteins were expressed in HeLa cells and the nuclear translocation in response to IL-6 and soluble IL-6 receptor stimulation was monitored in real time with confocal microscopy (Fig. 1C). All four constructs accumulated in the nucleus in response to cytokine treatment verifying the fusion constructs as functional molecules for further experiments.

To study the interactions and conformational changes between two fluorescently labelled STAT3 molecules, we measured FRET with the acceptor photobleaching technique (Fig. 2A). Briefly, when a suitable donor and acceptor fluorophore are in close proximity, non radiative energy transfer (FRET) can occur between the molecules. Disruption of the energy transfer by selective bleaching of acceptor molecules (TMRstar) leads to an increase in donor (eGFP) fluorescence due to loss of quenching. To set up the experimental conditions for FRET detection, background and maximal FRET signal were analysed with appropriate negative and positive FRET controls.

As a negative control, eGFP was coexpressed with TMRstar–STAT3 resulting in a very low FRET signal (2.28%±1.01, mean±s.d.) (Fig. 2A, lower panel; Fig. 2B). Additional controls including mitochondrial cytochrome oxidase 8-2 subunit (Cox8A)–TMRstar coexpressed with eGFP or donor (eGFP) and acceptor (TMRstar–STAT3) expressed alone also showed low FRET efficiencies of 1.05%±0.66, 0.69%±1.12 and 0.60%±0.98 respectively (data not shown). Energy transfer efficiencies of TMRstar-loaded SNAP tag fused with an eight-amino-acid linker to eGFP served as a positive control. A FRET efficiency of 17.55%±4.84 was measured, representing the highest detectable energy transfer signal in our setup (Fig. 2A, upper panel; Fig. 2B).

In further experiments, the selection of cells for FRET measurements and microscope settings were based on the positive control (donor and acceptor fluorophores are equimolar), to avoid the influence of different donor-to-acceptor ratios, which beside other parameters (spectral overlap, molecular distance, applied method) have been shown to affect FRET efficiency (Berney and Danuser, 2003).

It has been well established and demonstrated by various techniques that STATs are able to form stable dimers prior to stimulation (Braunstein et al., 2003; Haan et al., 2000; Kretzschmar et al., 2004; Schröder et al., 2004; Stancato et al., 1996). Our previous findings have shown the essential role of the NTD in the formation of unphosphorylated STAT3 dimers (Vogt et al., 2011), similar to other STAT family members (Wenta et al., 2008). However, the conformation and orientation of the protomers in the preformed dimer are still unknown. In our model (Fig. 3A), we propose two possible conformations of unphosphorylated STAT3 dimers: (1) antiparallel, with the two C-terminal regions localized at the opposite ends of the dimer, similar to the conformations found for latent STAT1 (Mertens et al., 2006) and latent STAT5a (Neculai et al., 2005); and (2) parallel, with the C-termini of the protomers located at the same end of the dimer, similar to the activated STAT3 dimer (Becker et al., 1998).

To follow the orientation of the NTDs during dimerization, eGFP–STAT3 and TMRstar–STAT3 (carrying the donor and acceptor fluorophore at the N-terminus of the host molecule) were coexpressed and examined with acceptor photobleaching FRET in fixed HeLa cells (Fig. 3B). Prior to activation, a significant FRET signal (P<0.001) was detectable between the NTDs compared with the negative control in both compartments (nucleus, 5.42%±2.05; cytoplasm, 5.50%±1.91; mean±s.d.), indicating the existence of latent STAT3 dimers and the close proximity of NTDs in this structure.

Upon activation STAT3 molecules get phosphorylated, translocate to the nucleus and act as transcription factors through binding to specific DNA sequences. This activated nuclear STAT3 fraction has completely lost the FRET signal, indicating that the NTDs have moved away from each other to form the activated STAT3 dimer. FRET efficiencies in the cytoplasm were not significantly different (P = 0.829) from untreated cells, indicating the dominance of a non-activated or de-activated STAT3 dimer fraction in this compartment after 30 minutes of cytokine stimulation. These findings are in good agreement with data generated with STAT5a, where stimulation also resulted in a loss of FRET between the NTDs of CFP–STAT5a and YFP–STAT5a (Neculai et al., 2005).

With C-terminally labelled STAT3 (STAT3–eGFP and STAT3–TMRstar) strong FRET signals were detected prior to activation in both cytoplasm and nucleus (Fig. 3B) indicating the close proximity of the C-terminal domains in the preformed dimer (nucleus, 10.09%±1.79; cytoplasm, 10.26%±2.26). FRET analysis revealed an increased energy transfer efficiency only in the nuclear compartment after cytokine treatment, in comparison with untreated cells (P = 0.00016), which is in good agreement with the previously published FRET results on STAT3–CFP and STAT3–YFP constructs (Cimica et al., 2011; Kretzschmar et al., 2004). However, this difference compared to non-stimulated sample is not that high to support an antiparallel orientation, but rather a parallel orientation of the protomers in the activated STAT3 dimer. In support of this hypothesis, measurements on C-terminally truncated STAT3 (ΔTAD-STAT3–TMRstar) and STAT3–eGFP, showed no detectable changes in FRET efficiency after stimulation (data not shown). Similar FRET experiments performed with living HeLa cells gave essentially the same results showing that fixation of cells does not affect FRET efficiencies (data not shown).

To further define the interdomain interactions during dimer formation of latent STAT3, we generated a construct which encodes the NTD of STAT3 and analysed the capability of the isolated NTD to interact with full-length STAT3. In HeLa cells transfected with constructs encoding TMRstar–NTD and eGFP–STAT3 or STAT3–eGFP (Fig. 4A) a strong FRET signal was detectable (nucleus, 9.34%±3.62; cytoplasm, 8.72%±2.94; mean±s.d.) when the donor fluorophore was localized at the N-terminus of the full-length protein. In contrast, transfection of STAT3–eGFP and TMRstar–NTD resulted in low FRET signals (nucleus, 2.64%±2.23; cytoplasm, 2.76%±1.94) indicating that the interaction is localized at the N-terminal region of STAT3. Measurements on N-terminally deleted STAT3 (STAT3-ΔNTD–eGFP), revealed no significant FRET between the two constructs (nucleus, 1.76±1.25%; cytoplasm, 1.45±1.57%), indicating that no other STAT3 domains are involved in the N-terminal interactions. This is in good agreement with GST pulldown assays from cellular lysates, where also no interaction between the N-terminal fragment and other structural domains of STAT3 was detectable (Zhang et al., 2002). Our data demonstrate the capability of the TMRstar–NTD construct to interact with full-length STAT3 and verify the NTD as one interaction surface, similar to previous publications, showing the capability of NTDs of other STATs to dimerize with themselves (Ota et al., 2004).

Next, we wanted to know whether a somatic mutation in the NTD of STAT3 found in human IHCA interferes with the NTD–NTD interaction. For this purpose the point mutant L78R was generated in analogy to the somatic mutation (Pilati et al., 2011) (Fig. 4B). As in case of the wild-type construct we fused a SNAP tag to the mutated NTD and tested the interaction with full-length STAT3 using FRET analysis (Fig. 4C). The single mutation in the N-terminal fragment resulted in the complete loss of FRET between the NTDs of STAT3 (nucleus, 1.62%±1.37; cytoplasm, 1.37%±1.31, mean±s.d.; Fig. 4C) in comparison to wild-type (Fig. 4A). Involvement of the L78 residue in the dimerization of NTDs of STAT3 suggests a dimer interface similar to STAT1 where the NTDs carrying a L78 mutation were properly folded but monomeric (Chen et al., 2003).

To analyse to what extent the isolated NTD interferes with STAT3 dimerization, HeLa cells were transfected with SNAP–NTD and STAT3–YFP. Cellular lysates were analysed by blue-native PAGE and the fluorescence of STAT3–YFP was detected with a fluorescence scanner (Fig. 4D). In cells expressing STAT3–YFP alone, STAT3 monomers and preformed dimers were detected, as has been shown previously (Vogt et al., 2011). Upon overexpression of SNAP–NTD, the dimers disappear in a dose-dependent manner. NTD-mediated interactions are of functional importance on promoters, such as the α2-macroglobulin gene promoter, that require tetramerization of activated STATs at adjacent interferon-γ-activated sequence (GAS) sites (Zhang and Darnell, 2001). Indeed, overexpression of the STAT3 NTD inhibits α2-macroglobulin gene induction in HepG2 cells in response to IL-6, as shown by a reporter gene assay (Fig. 4E, left panel). The NTD harbouring the L78R mutation does not interfere with gene induction. Thus, the interfaces of the STAT3 NTD involved in latent dimer formation and tetramerization on promoters seem to be very similar. Moreover, disturbing NTD interactions by a selective inhibitor can interfere with STAT3 activity on a subset of promoters that respond to unphosphorylated STAT3 (Timofeeva et al., 2013).

To elaborate these findings, two STAT3 constructs were generated carrying the specific point mutation at leucine L78 in full-length STAT3: SNAP–STAT3(L78R) and STAT3(L78R)–SNAP (Fig. 5A). To analyse the capability of L78R mutated STAT3 to form dimers prior to cytokine addition, FRET was measured, in a similar procedure to the wild-type measurements, on two different FRET pair combinations: eGFP–STAT3 and TMRstar–STAT3(L78R), and STAT3–eGFP and STAT3(L78R)–TMRstar (Fig. 5B). In all cases, no significant FRET signal was detectable prior to stimulation clearly indicating that the L78R mutated STAT3 molecules cannot form dimers in the latent state, similar to the N-terminally truncated STAT3 construct (Vogt et al., 2011). With the C-terminally labelled constructs, FRET was detectable after cytokine addition (nucleus, 6.61%±1.82; cytoplasm, 6.70%±1.93, mean±s.d.) demonstrating that the L78R mutation does not prevent the formation of activated STAT3 dimers. As in case of wild-type STAT3, the in vitro findings on fixed cells were repeated and confirmed in experiments with living cells (data not shown). In summary, our findings verify the importance of NTD interactions in stabilizing the preformed STAT3 dimer and highlight the relevance of the L78 residue in the dimer interface.

Upon transfection of HeLa cells, SNAP–STAT3(L78R) and STAT3(L78R)–SNAP were expressed and phosphorylated on Tyr705 after cytokine addition to a comparable extent, indicating that localization of the tag does not interfere with phosphorylation (Fig. 5C). However, although overexpressed, hardly any phosphorylation of STAT3(L78R) could be detected in the absence of a stimulus.

To characterize STAT3(L78R) in more detail, MEF^Δ/Δ cells lacking endogenous STAT3 were stably transfected with fluorescent STAT3(L78R)–CFP–YFP [STAT3(L78R)-CY] and compared with established and well-characterized MEF^Δ/Δ cells expressing wild-type STAT3–CFP–YFP (STAT3-CY) (Vogt et al., 2011) in a live-cell imaging experiment (Fig. 5D). Compared to the preferentially cytoplasmic localization of STAT3-CY wild-type, the L78R mutant is more equally distributed between the cytoplasm and nucleus (Fig. 5D, upper images), similar to the distribution of STAT3 lacking the NTD (Vogt et al., 2011). Upon stimulation, both proteins enter the nucleus with similar kinetics (Fig. 5D, middle panel). Upon removal of the stimulus both proteins redistribute to the initial state after about 3 h (Fig. 5D, lower panel, best seen for cells marked with arrows). Thus, the L78R mutant does not show any gross changes in nuclear accumulation and redistribution.

The stably transfected MEF^Δ/Δ cells were also used to analyse tyrosine phosphorylation of STAT3(L78R)-CY in comparison with STAT3-CY wild-type (Fig. 5E). Impairment of latent dimer formation in the STAT3(L78R) mutant does not have any influence on the sensitivity of the protein upon stimulation of IL-6 neither at low doses of the cytokine (Fig. 5E, upper panel) nor at early time points (Fig. 5E, middle panel). However, upon pulse stimulation with IL-6, phosphorylation of STAT3(L78R) is prolonged compared to wild-type (Fig. 5E, lower panel). Again, no phosphorylation of the L78R mutant is detectable in the absence of a stimulus.

It has been shown that the interaction of the isolated NTDs of STAT3 is of low affinity (millimolar range) compared to other members of the STAT family (micromolar range) (Wenta et al., 2008) suggesting that additional interfaces are involved in the latent STAT3 dimer. FRET studies on SH2-mutated STAT3 (R609Q) revealed the need of an intact SH2 domain for preformed dimerization (Kretzschmar et al., 2004). This specific mutation has a strong influence on the overall functionality of STAT3 and leads to a non functional SH2 domain (Hemmann et al., 1996) preventing receptor recruitment and phosphorylation of STAT3 (Herrmann et al., 2007).

We applied intramolecular FRET (donor and acceptor fluorophores localized at the same host molecule; Fig. 6A) to analyse the effect of the R609Q mutation on the overall STAT3 structure. Double tagged wild-type STAT3 (Fig. 6B) prior (nucleus, 6.68%±1.24; cytoplasm, 6.48%±1.39; mean±s.d.) or after cytokine stimulation (nucleus, 11.39%±1.90; cytoplasm, 13.20%±2.78) showed a similar FRET efficiency to the eGFP–STAT3 and STAT3–TMRstar pair prior to or after activation (data not shown). In the experiments with double-labelled wild-type STAT3 we cannot exclude the intermolecular FRET signal from our detection system. However, the monomeric R609Q mutant showed high intramolecular FRET efficiency values (nucleus, 11.27%±2.16; cytoplasm, 10.95%±2.45) similar to that of the activated wild-type, indicating the same structural organization of this mutant monomer as the activated dimer (Fig. 6C).

These results suggest that mutation of the R609 residue in the SH2 domain not only prevents the interaction between phosphotyrosine motifs and the binding pocket in the SH2 domain, but leads to a complete reorientation of the SH2 domain (detected as high intramolecular FRET efficiency between the N- and C-terminal domains), disrupting the capability of the SH2 domain to additionally stabilize the preformed dimers. The R609Q mutant reflects a STAT3 form that structurally mimics the activated state, rendering the molecule unable to interact with activated and latent STAT3.

Our FRET experiments on latent STAT3 revealed closely associated NTDs that rearrange upon activation and DNA binding as indicated by the loss of nuclear FRET of the N-terminally labelled protomers upon stimulation of cells. Previous work from our laboratory (Vogt et al., 2011) showed that the deletion of the NTD abrogated the formation of latent STAT3 dimers. A similar function of the NTD in formation of latent dimers was shown for STAT5a (Neculai et al., 2005), STAT1 (Mao et al., 2005) and STAT4 (Vinkemeier et al., 1998). FRET experiments with the isolated N-terminal fragment of STAT3 support the idea of homotypic interactions of the NTDs instead of cross reactions with other domains of STAT3. These findings are in good agreement with previously published data based on other STAT family members (Ota et al., 2004; Wenta et al., 2008). Native gel electrophoresis showed that dimerization of nonphosphorylated STAT3 can be inhibited by the coexpression of the wild-type NTD (SNAP–NTD) demonstrating that the preformed dimers are to a large extent stabilized by homotypic NTD–NTD interactions.

Tetramerization of STAT3 has been shown to be essential for the maximal transcriptional activation of the α2-macroglobulin gene promoter (Zhang and Darnell, 2001). This cooperative binding of STAT dimers on specific DNA target sites is also mediated by homotypic interactions between NTDs (Vinkemeier et al., 1996; Xu et al., 1996). Upon overexpression of wild-type NTD, a significant downregulation of the α2-macroglobulin promoter was detectable compared to non-transfected or empty-vector-transfected samples, demonstrating the capability of SNAP-NTD construct to inhibit the tetramerization of STAT3. In agreement with our results, the same effect was achieved with the use of a STAT3-NTD-based inhibitor peptide (STAT3-Hel2A) on an acute-phase response element (APRE) luciferase reporter (Timofeeva et al., 2007).

Additionally, we focused on the crucial residue in the NTD, which is involved in mediating and stabilizing the homotypic interaction. Based on the somatic mutation found in inflammatory hepatocellular tumours (Pilati et al., 2011) targeting the NTD, we generated a construct encoding the N-terminal fragment with a L78R amino acid mutation. No interaction of the mutant NTD with full-length STAT3 protein was detectable, suggesting that the L78 residue is a promising target for further investigations. In support of our findings with the isolated NTDs, FRET measurements on L78R mutated STAT3 showed no detectable dimers prior to activation. These results suggest a similar NTD–NTD interface in STAT3 as was previously proposed for STAT4 and STAT1 (Chen et al., 2003), where the hydrophobic residues L77 and L78 are involved in the homodimerization of these domains. Mutations of these residues to alanine did however not completely prevent latent dimer formation of STAT3 (Vogt et al., 2011). Thus, a stronger disturbance of the dimer interface, such as the replacement of the hydrophic leucine residue by a charged arginine residue, is required to achieve a monomeric population of latent STAT3.

In the context of IHCA, STAT3(L78R) has been reported to be a weakly constitutively active mutant. In our hands, upon transient overexpression in HeLa cells in the presence of endogenous STAT3 wild-type as well as upon stable expression in MEF^Δ/Δ cells lacking endogenous STAT3, hardly any constitutive phosphorylation was detectable. However, under steady-state conditions, a greater portion of the L78R mutant is located in the nucleus compared to wild-type, as has been observed for STAT3 lacking the NTD (Vogt et al., 2011). Sensitivity to IL-6 stimulation is similar for STAT3 wild-type and the L78R mutant. In this respect, STAT3 resembles STAT1 and is different from STAT4, which requires latent dimer formation for activation (Ota et al., 2004). However, upon pulse-stimulation, phosphorylation of STAT3(L78R) is slightly prolonged. This does not translate into a prolonged nuclear localization, most probably because of the different time windows of dephosphorylation (minutes) and redistribution (hours). Interestingly the NTD of STAT3 has been shown to be important for the recruitment of coactivators such as most recently described for FoxP3 (Hossain et al., 2013). It would be interesting to know whether this kind of interaction is also mediated through the L78 interface and whether the disturbance of such interactions through the L78R mutation contributes to IHCA.

The NTD of STAT3 forms dimers through a low affinity interaction (K_d = 3.7 mM) compared to the NTDs of STAT1 and STAT4 (K_d = 23.3 µM and 2.7 µM, respectively) (Wenta et al., 2008) suggesting that other structural regions are involved in dimer stabilization of full-length STAT3. The crystal structure of the unphosphorylated STAT3 core fragment (lacking the NTD and transactivation domain) revealed no evidence of other structural domains being involved in preformed dimer formation, such as the coiled-coil domain or DNA-binding domain (Ren et al., 2008), in contrast to STAT1 and STAT5 (Mao et al., 2005; Neculai et al., 2005). In particular, the interface involving the F172 residue in the anti-parallel STAT1 dimer is not conserved in STAT3 (Ren et al., 2008). Our measurements on C-terminally truncated STAT3 showed that deletion of the transactivation domain does not affect latent dimer formation (data not shown). However, FRET measurements on SH2-mutated STAT3(R609Q) revealed the need for an intact SH2 domain for the formation of latent dimers and suggested that there are SH2 homotypic interactions prior to activation (Kretzschmar et al., 2004). Our intramolecular FRET measurements on STAT3R609Q (Fig. 6B) indicate that this mutation not only affects interaction with phosphotyrosine motifs but disturbs the overall structure of the SH2 domain resulting in a monomer, which mimics the structural organization of the activated STAT3 dimer. This modification possibly interferes with regions in the SH2 domain that are involved in dimerization of latent STAT3 by generating an additional stabilizing interface between the SH2 fragments. Previously published data have demonstrated that the isolated STAT3 SH2 domain interacts with itself, and these dimerized SH2 domains are unable to interact with phosphopeptides (Haan et al., 1999). Interestingly, in HIES, heterozygous mutations (missense or single-codon in-frame microdeletions) in the DNA-binding and SH2 domains of STAT3 (Holland et al., 2007) result in the expression of full length but dysfunctional STAT3 molecules. The mutations targeting the DNA-binding domain have been shown to have dominant-negative effects on the wild-type form (Minegishi et al., 2007), whereas the amino acid changes in the SH2 domain (S611N and F621V), similar to R609Q mutation, abrogate the dimerization of STAT3 upon activation (He et al., 2012).

Our findings identify the SH2 domain as a secondary stabilizer of latent STAT3 dimers leading to a parallel orientation of the protomers similar to that of activated STAT3 (Fig. 7). This is in contrast to the antiparallel dimers of latent STAT1 and STAT5 where the SH2 domains are localized at the opposite ends of the dimer (Mao et al., 2005; Mertens et al., 2006; Neculai et al., 2005). This parallel association could explain why, under some circumstances, latent STAT3 is capable of DNA binding (Nkansah et al., 2013; Timofeeva et al., 2012) and gene induction (Yang et al., 2005).

Differently labelled STAT3 constructs for FRET analysis were generated with the use of pcDNA5/FRT/TO expression vector (Invitrogen, USA). pcDNA5/FRT/TO-STAT3-YFP (Herrmann et al., 2007) was used as a template for cloning the constructs. The labelling tags eGFP and SNAP were derived from peGFP-N1 vector (Clontech, USA) and pcDNA5/FRT/TO-SNAP-YFP (Recker et al., 2011), respectively, and amplified by PCR with the appropriate primers to insert the restriction sites for the cloning procedure. R609Q mutated STAT3 was created from pSVL-STAT3(R609Q)-YFP, described previously (Herrmann et al., 2004). Site directed mutagenesis was performed with the QuikChange II XL Kit (Agilent Technologies, USA) to introduce the L78R mutation (underlined codon) into STAT3 using the following primer pair: 5′-CAAGAGTCCAATGTCCGCTA TCAGCACAACCTTC-3′ and 5′-GAAGGTTGTGCTGATAGCGGACATTGGACTCTTG-3′.

Recombinant human IL-6 (1400 U/µl) and soluble IL-6 receptor were prepared as described previously (Arcone et al., 1991; Weiergräber et al., 1995).

Human cervix adenocarcinoma (HeLa) cells were cultivated in Phenol Red free DMEM-medium (Gibco, Germany), supplemented with 10% FCS and 1% penicillin/streptomycin. Human liver hepatocellular carcinoma (HepG2) cells were cultivated in DMEM/F12-Medium (Gibco, Germany), supplemented with 10% FCS and 1% penicillin/streptomycin. The cells were incubated at 37°C in a humidified atmosphere containing 5% CO₂. Transfection was performed transiently, when cells reached a density of 60–80% confluence, by using the transfection reagent TransIT-LT1 (Mirus Bio LLC, USA), and OPTIMEM (Gibco, Germany) as serum-free medium. Conditions and protocols were used according to the manufacturer's instructions. MEF^Δ/Δ cells stably transfected with STAT3(L78R)-CY have been generated as described earlier for STAT3-CY wild-type (Vogt et al., 2011) using the Flip-In T-REx system (Invitrogen, USA) that guarantees integration of the different transgenes at identical genomic sites.

HeLa cells were cultured on six-well plates. The cells were stimulated with 20 ng/ml IL-6 and 500 ng/ml soluble IL-6 receptor for 30 minutes or left untreated and lysed as described previously (Vogt et al., 2011). The lysates were analysed by SDS-PAGE, western blotting and immunodetection using antibodies directed against phosphotyrosine 705 of STAT3 (Cell Signaling Technology, Danvers, MA) and STAT3 (H190; Santa Cruz Biotechnology, Santa Cruz, CA).

The labelling of the SNAP tag fusion proteins was performed using SNAP-Cell TMRstar substrate, a cell-permeable fluorescent label based on Tetramethylrhodamine (New England Biolabs, USA). The labelling procedure was carried out according to the manufacturer's protocol.

HeLa cells were examined with a LSM 710 confocal microscope (Carl Zeiss, Germany) using a 40×, 1.1 NA water immersion objective. For fixation, cells were seeded and grown on LabTek four-well chamber slides (Nunc, Thermo Fisher Scientific, USA) in Phenol-Red-free medium. After removing the medium, cells were washed two times in phosphate-buffered saline (PBS) containing 1 mM MgCl₂ and 0.1 mM CaCl₂ (PBS⁺⁺). Fixation was performed by addition of 200 µl methanol to the cells, followed by incubation in the dark, for 20 minutes at room temperature. After incubation, the cells were washed once with PBS⁺⁺ and quenched with 50 mM NH₄Cl (diluted in PBS⁺⁺) for 5 minutes. Coverslips were mounted with Immu-Mount (Thermo Fisher Scientific, USA) and the samples were kept in the dark until further investigation. For live-cell imaging, cells were plated on 35-mm glass-bottomed dishes (Ibidi, Germany) in Phenol-Red-free medium 48 hours before the experiment and transfected as described above. Cells were imaged at 37°C and 5% CO₂ in the cell incubator of a Zeiss LSM 710 microscope (Carl Zeiss, Germany).

HeLa cells were transfected with STAT3–YFP and increasing amounts of the SNAP–NTD-encoding expression vector. Cells were left unstimulated and lysed with the NativePAGE™ Sample Prep Kit (Invitrogen, UK) with 1% (final volume) DDM (n-dodecyl-β-maltoside) as detergent. The lysates were mixed with Coomassie Brilliant Blue G-250 and separated on a 4–16% gradient gel using the NativePAGE™ Novex Bis-Tris gel system (Invitrogen, UK), and the protein complexes were separated overnight at 4°C. Following electrophoresis, the gel was analysed with a fluorescence scanner to visualize the STAT3–YFP signal by using a fluorescence scanner (Typhoon, GE Healthcare, UK). The probe was excited with 488-nm laser line and the emission detected using 515–555 nm band filter setting.

HepG2 cells were grown in six-well plates and transfection was performed with 1 µg β-galactosidase expression vector (pCR3lacZ, Pharmacia, Sweden), 300 ng of luciferase reporter construct (α2-macroglobulin) and the indicated plasmid constructs using the TransIT-LT1 transfection reagent (Mirus Bio LLC, USA) according to the manufacturer's instructions. After 24 hours, cells were stimulated for 4 hours, harvested, and luciferase and β-galactosidase activity were measured in triplicates. Luciferase assays were performed using a luciferase assay kit (Promega, Germany) and values were normalized to transfection efficiencies derived from β-Gal expression.