Nucleocytoplasmic shuttling of STAT transcription factors

Abstract

The signal transducer and activator of transcription (STAT) proteins have initially been described as cytoplasmic proteins that enter the nucleus only after cytokine treatment of cells. Contrary to this assumption, it was demonstrated that STATs are constantly shuttling between nucleus and cytoplasm irrespective of cytokine stimulation. This happens both via carrier-dependent as well as carrier-independent transportation. Moreover, it was also recognized that cytokine stimulation triggers nuclear retention of dimeric STATs, rather than affecting the rate of nuclear import. In summary, it is increasingly being appreciated that STAT nucleocytoplasmic cycling determines the quality of cytokine signaling and also constitutes an important area for microbial intervention.

Abbreviations

CRM1

chromosomal region maintenance 1

dsNLS

dimer-specific nuclear localization signal

GAS

gamma-activated sites

JAK

Janus kinase

NLS

nuclear localization signal

NPC

nuclear pore complex

SH2

SRC homology 2

STAT

signal transducer and activator of transcription

Introduction

Multicellular organisms utilize an integrated network of cell–cell communications and humeral interactions to coordinate complex cellular processes such as proliferation, differentiation, and homeostasis. Cells recognize external stimuli and transform the signals into a cellular response, which most often result in an alteration in the pattern of expressed genes. Many signal transducers that function as transcription factors have to traverse the barrier of the nuclear envelope in order to gain access to specific target genes within the nuclear compartment. The Janus kinase (JAK)-signal transducer and activator of transcription (STAT) pathway is regarded as a paradigmatic model for such a direct signal transduction, because it transmits information received from extracellular polypeptide signals without the interplay of second messengers directly to target promoters in the nucleus [1].

The STAT proteins comprise a family of evolutionarily conserved transcription factors and in mammalian cells seven known STAT proteins were identified, denoted STAT1, STAT2, STAT3, STAT4, STAT5a, STAT5b, and STAT6, all of which are activated by a distinct set of cytokines and growth factors [1]. These proteins consist of several conserved functional domains. The amino terminal N-domain is responsible for tetramerization of all STATs (with the probable exception of STAT2), and this domain also regulates receptor recognition and phosphatase recruitment for some STATs [2–5]. The N-domain is followed by a coiled-coil domain implicated in protein–protein interactions [6], a DNA binding domain [7], a linker domain that participates in DNA binding [8], an SRC homology 2 (SH2) domain that mediates dimerization and receptor binding [9], and a carboxy-terminal transactivation domain [10].

Best characterized is the role of STAT proteins in cytokine signaling. Upon binding of extracellular ligands such as interferons or interleukines to their cognate receptors, receptor-associated Janus kinases, of which four have been described in mammalian cells (JAK1, JAK2, JAK3 and TYK2), undergo tyrosine autophosphorylation and transphosphorylate tyrosine-containing motifs on the intracellular receptor chains, thus creating docking sites for the SH2 domain of STAT molecules [11]. Subsequently, the JAKs catalyze the phosphorylation of a single tyrosine residue in the carboxy terminus of STAT proteins [10,12]. The tyrosine-phosphorylated STATs detach from the intracellular receptor tail and homo- or heterodimerize due to reciprocal phosphotyrosine-SH2 interaction ([1] and Fig. 1). Before exposure of cells to cytokines the STAT molecules are nontyrosine phosphorylated, but may assemble into dimeric and higher order complexes [13,14]. Structurally and functionally these aggregates remain sparsely characterized. Therefore, throughout this review we will use the term ‘dimer’ as shorthand for ‘tyrosine-phosphorylated dimer’.

Figure 1

STATs at the cell membrane. A schematic representation of the events leading to the tyrosine phosphorylation (activation) of STATs. The activation of receptor-associated JAK kinases after cytokine stimulation results in tyrosine phosphorylation of the receptor. The STATs dock to these sites via their SH2 domains and become tyrosine phosphorylated concomitantly. The activated STATs detach and homo- or heterodimerize.

A characteristic but until recently poorly understood phenomenon associated with cytokine stimulation of cells is the inducible and transient accumulation of STAT proteins [10]. Once in the nucleus, STAT dimers can directly bind to nonameric DNA sequences known as gamma-activated sites (GAS) in the promoter region of cytokine-responsive genes resulting in gene transcription [7]. Several years ago, Yoneda and coworkers showed that cytokine stimulation with concomitant dimerization of tyrosine-phosphorylated STATs induces their association with importin transport factors [15]. Next, we will describe what is presently known about the molecular basis of this process.

Requirements for cytokine-induced nuclear import of STATs

Macromolecules and ions alike have to traverse the nuclear membrane through specialized structures called nuclear pore complexes (NPCs) [16]. The NPCs constitute high-order octagonal channels that are an integral part of the nuclear envelope. They are composed of proteins called nucleoporins which are present in multiples, and some of them contain hydrophobic phenylalanine/glycine (FG)-rich repeat motifs [16]. Macromolecules exceeding a molecular mass of ≈ 40 kDa are generally barred from freely crossing the nuclear membrane by random diffusion [17]. Thus, the NPCs function as selectivity filters by restricting the transport of some macromolecules, while allowing the rapid translocation of others.

Detailed mechanistic insight has been acquired into translocation mechanisms that rely on transport receptors of the karyopherin superfamily of proteins [18]. Karyopherins mediate either import into or export from the nucleus and they are therefore also called importins or exportins, respectively. They recognize loosely conserved sequence motifs on the surface of their substrates (also called cargoes). These signals allow the association with cargo proteins and the subsequent passage of the complex through the nuclear pore. Importins and exportins, although structurally related, differ in their sequence requirements for cargo association, as nuclear localization signals (NLS) are usually rich in basic residues, while nuclear export signals are characterized by the presence of hydrophobic residues, usually leucines [19]. It is believed that the karyopherins act as chaperones during nucleocytoplasmic translocation. Passage through the pore appears to require weak and transient binding to the nucleoporin FG repeats, an interaction that by itself was shown to occur independently of metabolic energy [20,21]. Energy consumption, however, confers directionality to this process, which therefore was also termed active transport. The driving force behind the active translocation is created by Ran-GTPase nucleotide exchange factors, which are distributed asymmetrically between cytosol and nucleus [22]. Nucleotide hydrolysis by RanGAP, the cytoplasmically localized RanGTPase-activating protein, results in high levels of RanGDP in the cytosol. In the presence of RanGDP, importins are loaded with substrates and may translocate through the NPC into the nucleus, while the export receptors are liberated from their cargo molecules in this environment. The reverse reactions take place in the nucleus. Here, a high RanGTP/RanGDP ratio is maintained by the guanine nucleotide exchange factor RCC1, which catalyzes the conversion of RanGDP to RanGTP. RanGTP was demonstrated to promote both the disassembly of importin/cargo complexes and the association of exportins such as chromosomal region maintenance 1 (CRM1) with their cargoes [19].

At present, the overwhelming majority of examples of protein nucleocytoplasmic shuttling belong to this active mode of translocation. STAT proteins have also been demonstrated to utilize components of this Ran-dependent nuclear import machinery [15,23]. The karyopherin importin β (p97) has been identified as the carrier that transports importin α complexed with STATs into the nuclear compartment ([15,23] and Fig. 2A). In interferon-stimulated cells dimerized STAT1 and STAT2 bind directly to importin α5 (NPI-1/hSrp1), a karyopherin that contains 10 armadillo repeats [15,24,25]. Only the very C-terminal armadillo repeats 8 and 9 bind to STAT1 homodimers and STAT1-STAT2 heterodimers, whereas classical NLS sequences interact with repeats 2–4, 7 and 8 [26].

Figure 2

STATs at the nuclear envelope. (A) Carrier-dependent import. Phosphorylated STAT dimers expose a dimer-specific nuclear localization signal and associate with importin α. Through importin β-mediated interactions with the interior of the nuclear pore (NPC) this complex migrates into the nucleus. The complex disassembles after the bindung of RanGTP. The exact stoichiometry and order of events have not been established. (B) Carrier-dependent export. Unphosphorylated STATs can bind to the exportin CRM1 via leucine-rich nuclear export signals and traverse the NPC. RanGTP enhances the interaction of CRM1 with cargo proteins. In the cytoplasm, the nucleotide hydrolysis of RanGTP leads to release of the cargo. (C) Carrier-independent nucleocytoplasmic translocation. For the STATs, the majority of translocation events occur via direct interactions with proteins of the nuclear pore. The resulting nucleocytoplasmic cycling proceeds independently of metabolic energy.

The binding site for importin α5 on the STAT1 dimer has been mapped to an unusual dimer-specific nuclear localization signal (dsNLS) within the DNA binding domain [24,25]. The homologous sequence in the DNA binding region of STAT3 was later reported to also function as an NLS for the dimer [27]. It is interesting to note that binding of STATs to importin α5 does not appear to pose an obstacle to promoter binding and transcription, as STAT-target DNA can disrupt the importin α5 complex with STAT1 [25]. The dsNLS differs from conventional import signals in some respects (Fig. 3). First, it does not resemble the consensus sequence of classical mono- or bipartite NLSs, which consist of one or two arginine/lysine-rich clusters of basic amino acids separated by a spacer region ranging from 10 to up to about 40 residues [28,29]. The STAT1 dsNLS, in contrast, contains only a few positively charged residues. Another distinguishing feature of the STAT dsNLS is its nontransferability, because it functions only in the context of the STAT dimer, but not autonomously as is typical for conventional NLSs [28,29]. In addition, the STAT amino termini also appear to provide signals for the cytokine-inducible nuclear localization as judged from the inability of amino terminal deletion mutants to accumulate in the nucleus [30]; and residues in the coiled-coil domain seem to contribute to carrier-dependent nuclear import of some STATs [27].

Figure 3

The dimer-specific nuclear import signal (dsNLS) of STAT1. A short stretch from the DNA binding domain of STAT1 harbors overlapping export and import activities. Notably, the import activity is observed only in the native STAT dimer, whereas the export activity is readily observable in the isolated peptide. Residues that were demonstrated to be important for export (of isolated peptides) are depicted in a white box, residues that are required only for import (of the dimer) are boxed in dark grey. Residues, mutation of which affected both import and export, are shown in a light grey box. For comparison, the homologous sequences of other STATs are listed: D, Drosophila; h, human.

The canonical model of the JAK-STAT pathway stated that unphosphorylated STATs are cytoplasmic and do not participate in nucleocytoplasmic shuttling. However, this model has been challenged by the observation that some STAT family members undergo constitutive shuttling between the nuclear and cytosolic compartments even in the absence of cytokine stimulation. A growing body of evidence indicates that the nucleocytoplasmic cycling of STAT proteins is much more dynamic than initially thought. In the following we will describe and discuss the recent advances, which make necessary a fresh look at the principles of cytokine signaling.

Continuous nucleocytoplasmic cycling of STATs

Loss-of-function mutations of the STAT1 dsNLS block nuclear entry of tyrosine-phosphorylated STAT1 [29]. As anticipated, the dsNLS mutants failed to activate interferon-inducible STAT target genes despite their unperturbed dimerization and DNA binding abilities. Moreover, the import defect was associated also with the loss of cytokine-induced nuclear accumulation. Despite that, ample amounts of unphosphorylated dsNLS mutants of STAT1 were found in the nucleus of unstimulated cells [29]. This was taken as the first indication that unphosphorylated STATs used nuclear import mechanism(s) that deviated from the importin-dependent translocation described for the phosphorylated dimer. Further hints came from the observation of nuclear pools of monomeric STAT1 and STAT3 in a variety of unstimulated primary cells or established cell lines [31,32]. Point mutations in either the SH2 domain or the tyrosine residue in position 701 that completely prevented the signal-dependent dimerization had no effect on the intracellular STAT1 localization in resting cells [31,32]. The direct visualization of STAT1 nucleocytoplasmic shuttling in resting cells was made possible by the intracellular microinjection of precipitating anti-STAT1 IgG [29]. Strikingly, upon the microinjection of a specific antibody, but not of an unspecific immunoglobulin, STAT1 was depleted from the noninjected compartment [29]. This assay was used to perform time-course experiments to assess the nucleocytoplasmic flux rates of endogenous STAT1 in unstimulated cells [33]. It was found that the antibody-induced STAT1 clearance was rapid and complete in about 30 min, irrespective of whether the antibody was injected into the cytoplasm or the nucleus (Fig. 4A–C). Moreover, while energy-depletion of cells precluded nucleocytoplasmic transport of karyopherin-dependent cargo proteins, the unphosphorylated STAT1 continued to exchange between nucleus and cytosol under this condition [33]. Thus, constitutive nucleocytoplasmic shuttling continued in the absence of metabolic energy and an intact RanGTP gradient. High exchange rates between the nuclear and cytoplasmic STAT pools were reported also for STAT3 and STAT5 [34,35].

Figure 4

Nucleocytoplasmic shuttling of STAT1 in resting and cytokine-stimulated cells. Antibody microinjection assays with an unspecific STAT3 antibody (A) or a specific STAT1 antibody (B–D). After antibody injection the cells were incubated for 30 min at 37 °C, before fixation and immunocytochemical detection of endogenous STAT1. The site of injection was marked by the coinjection of fluorescine-conjugated bovine serum albumin. Arrows point at the injected cells. The control in (A) demonstrated that the STAT1 distribution is not affected by microinjection of an unspecific antibody. The injection of a STAT1-specific antibody revealed the constitutive cycling of STAT1 in resting cells (B,C). Cytoplasmic injection of anti-STAT1 depleted endogenous STAT1 from the nucleus (B), whereas nuclear delivery of anti-STAT1 caused STAT1 accumulation in the nucleus (C). In (D) the cells were treated with interferon-γ for 60 min to induce the nuclear accumulation of STAT1, before anti-STAT1 was injected into the cytosol of the indicated cell. After another 30 min, nuclear STAT1 was substantially diminished in the injected cell. Note the continued nuclear accumulation in the neighboring cells.

These findings were complemented by import assays with digitonin-permeabilized cells that retain an intact nuclear envelope, but which are devoid of cytoplasmic proteins such as importins [36]. These experiments revealed that exclusively unphosphorylated STAT1 could enter the nucleus in the absence of cytosolic proteins, whereas tyrosine-phosphorylated STAT1 dimers required both metabolic energy and added cytosol for nuclear import. Identical observations were also made for unphosphorylated STAT3 and STAT5 [33]. Moreover, it was found that the carrier-free transport is saturable and appears to occur through direct contacts between STAT proteins and FG repeat-containing nucleoporins [33]. Interestingly, in vitro alkylation with N-ethyl-maleimide of a single cysteine residue in the STAT1 linker domain precluded the translocation across the nuclear membrane, suggesting that the functionally poorly characterized linker domain plays a fundamental role in carrier-independent nucleocytoplasmic shuttling [33]. Although the structural details that determine the carrier-free passage of STATs through the nuclear pore remain to be established, it was shown that truncated STAT mutants that lack the amino- and carboxy-termini entered the nucleus with identical kinetics as the full-length molecule. The nuclear export rate of these truncation mutants, on the other hand, was reduced [33], which indicated that the structural requirements are complex and possibly affect transport in a direction-specific manner. Taken together, STATs use two different import pathways: before cytokine stimulation, unphosphorylated STATs migrate via a carrier-free mechanism that involves direct interactions with nucleoporins. Nuclear import of tyrosine-phosphorylated STAT dimers, on the other hand, is dependent on importins, Ran, and metabolic energy. Both pathways operate simultaneously in cytokine-stimulated cells and it appears that phosphorylation-induced dimerization is the switch from facilitated diffusion to carrier-mediated translocation (Fig. 2). Notably, only one third of the STAT1 molecules are tyrosine phosphorylated at any moment during cytokine stimulation [37].

Work in our laboratory identified a functional leucine-rich nuclear export signal in STAT1 and demonstrated its role in vivo, thus showing that nuclear export of STAT1 was occurring [38]. In the meantime, further putative leucine-rich nuclear export signals have been identified in varying locations in STAT1 [39], STAT3 [40], and STAT5 [35], as well as in Dictyostelium STATa [41], and STATc [42]. Of note is the fact that characterization of the STAT export signals remains incomplete, as export activity in the full length molecule has not been demonstrated yet for some of them. Interestingly, a biphasic regulation was described for STATa in which extracellular cAMP initially directs nuclear import of tyrosine-phosphorylated STATa and phosphorylation of amino terminal serine residues catalyzed by glycogen synthase kinase-3 promotes its subsequent export [41]. This raises the intriguing possibility of flux modulations via post-translational modifications also for mammalian STATs. However, the respective phosphorylation sites are not conserved.

While the CRM1-mediated nuclear export was initially implicated only in the termination of cytokine-induced nuclear accumulation of STATs, it is now clear that this export pathway operates constitutively [33]. Preincubation of resting cells with the CRM1 inhibitor leptomycin B did not cause the nuclear accumulation of STAT1, which by some was taken as an indication that STATs do not shuttle in resting cells [39]. In addition, it was noted that leptomycin merely attenuated the cytoplasmic relocation after cytokine-induced nuclear accumulation, but did not cause a complete block [38]. As described above, this phenotype is explained by the existence of a carrier-independent and hence leptomycin-insensitive nuclear export mechanism [33]. STATs are predominantly cytoplasmic in resting cells, although STAT- and cell type-specific differences were reported [32]. For STAT1, the underlying molecular mechanism was determined to entail the cooperative action of both the carrier-free and the CRM1-dependent translocation mechanism (Fig. 2B,C). It was found that inactivation specifically of CRM1 or generally of energy-consuming transport pathways caused a nuclear relocation, resulting in a pancellular STAT1 distribution [33]. Whether retention mechanisms such as the complexation with cytoplasmic anchoring factors also contribute to the cytoplasmic accumulation in resting cells is currently unclear.

As was mentioned already, cytokine stimulation of cells triggers a dramatic translocation of STATs into the nucleus. This phenomenon, which depending on the stimulus and its intensity can last for several hours, was initially believed to reflect an exclusively nuclear residence of STATs. However, nuclear accumulation was recognized to be a highly dynamic process, as the rapid nucleocytoplasmic cycling of STATs continues even during the accumulation phase. In the following we will outline how dimerization, the STAT/DNA dissociation rate, and the activity of a nuclear phosphatase were identified as the crucial players that control retention and accumulation of STATs in the nucleus.

The STAT/DNA dissociation rate is a central integrator of cytokine signaling

Novel insight into the readily observable cytokine-stimulated nuclear accumulation of STATs has been gained in the recent past. It was long known that dimerization of phosphorylated STATs is an absolute requirement for an observable accumulation in the nucleus [10]. However, it has become clear that the concurrent switch to carrier-dependent transport is not the cause of nuclear accumulation, as mutants were generated that were imported normally in response to cytokine stimulation, but that nevertheless were not capable of nuclear retention [43]. Based on in vivo labeling experiments and subcellular fractionations, it was previously proposed that the duration of STAT nuclear accumulation was influenced by the activity of tyrosine phosphatases [37]. Several phosphatases, some of them nuclear, have been demonstrated to affect the rate of STAT dephosphorylation in vivo[44]. Alternatively, ubiquitination followed by degradation was proposed to terminate STAT signaling in the nucleus [45].

Recent work unambiguously demonstrated that tyrosine-phosphorylated STAT1 is incapable of nuclear exit and has to be dephosphorylated in order to leave the nuclear compartment [4,43]. This fact constitutes the basis of the cytokine-induced nuclear accumulation of STATs. The importance of reduced export for the induced nuclear accumulation was also shown for a STAT protein from Dictyostelium[42]. While the nuclear accumulation can last for several hours, the nuclear phosphatase activity results in almost instantaneous dephosphorylation. Therefore the question arises as to the mechanisms that defer tyrosine dephosphorylation. Surprisingly, this mechanism was determined to be DNA binding. It was found that the sequence-specific off-rate from DNA was correlated with the half-life of the phosphorylated protein [43]. STAT dimers that were bound to high-affinity GAS sites resisted dephosphorylation better, as compared to STAT molecules bound to non-GAS sites (Fig. 5). Thus, contrary to the previous assumption that dephosphorylation releases STATs from DNA, it was the other way around, and DNA binding protected STATs from the enzyme activity. This conclusion was supported by measurements of the intranuclear mobility of STAT1 in the presence and absence of phosphatase activity [4,43,46]. Even if the phosphatase activity was blocked, the mobility of STAT1 remained close to the diffusion limit. Normally, however, owing to their high DNA off-rate [2], the protection from dephosphorylation conferred by DNA binding does not last for the entire time of nuclear accumulation. In vivo, the half-life of phosphorylated STAT1 and STAT3 was shown to not exceed 15–30 min even on a target promoter [37,47]. Thus, the apparently constant level of nuclear accumulated STAT molecules is maintained by constant nuclear export and successive re-import [48,49]. The resulting nucleocytoplasmic cycling during nuclear accumulation was clearly demonstrated by cytoplasmic trapping of STAT1 after antibody microinjection ([43] and Fig. 4D). The central role of dimerization for nuclear retention of STATs was confirmed by a STAT1 mutant that had lost its ability to recruit the inactivating phosphatase TC45 [43,50]. Exchange of a single amino acid residue in the amino terminal domain could reverse the defective nuclear accumulation of a DNA binding mutant without rescuing the DNA binding phenotype [4]. These observations also contradicted a competing model for nuclear accumulation, which stated that DNA binding was a necessary prerequisite for nuclear accumulation [39].

Figure 5

STATs in the nucleus. STAT binding sites on DNA differ strongly in terms of their DNA off-rate, which is lowest at optimal target sites (GAS). Enzymatic dephosphorylation of STATs is possible only when the molecule is off DNA. Thus, the activity of the STAT dimer is extended at promoters with optimal STAT binding site(s).

Thus, the coupling of dephosphorylation and nuclear retention to the sequence-specific DNA off-rate constitutes a regulatory mechanism that integrates at least three important determinants of cytokine signaling. These are the half-life of the transcriptionally active STAT dimer, the duration of promoter occupancy, and finally the ability to link nuclear activity to the activity of cytokine receptors in the cell membrane.

STAT nucleocytoplasmic transport in disease

It is increasingly becoming clear that nucleocytoplasmic cycling of signal transducers is an intricate process that affects signaling in many ways. It is therefore not surprising that several viral proteins, such as the V proteins from Nipah and Hendra viruses, both of which cause zoonotic diseases in animals and humans, have been shown to interfere with the nucleocytoplasmic translocation of STAT proteins ([51–53]; reviewed in [54]). The interferon antagonistic activity of these paramyxovirus V proteins included the cytoplasmic sequestration of STAT1 and STAT2 in high molecular mass complexes. It was shown that Nipah and Hendra V proteins alter the subcellular distribution of STAT1 in resting cells and prevent nuclear import of both STAT1 and STAT2 in interferon-stimulated cells. Thus, inhibition of nucleocytoplasmic shuttling constitutes a viral strategy to evade the antiviral effects of interferons. In addition, impaired interleukine-12-dependent nuclear translocation of STAT4 was reported in a patient with recurrent mycobacterial infection [55]. These first examples demonstrate already that nucleocytoplasmic transportation of STATs can offer novel possibilities also for medical intervention.

Acknowledgements

The authors' research on this subject is funded by grants from the Deutsche Forschungsgemeinschaft, the EMBO-Young-Investigator-Program and the Bundesministerium für Bildung und Forschung (BioFuture).