*Review* **Nucleocytoplasmic Shuttling of STATs. A Target for Intervention?**

#### **Sabrina Ernst 1,2 and Gerhard Müller-Newen 1,\***


Received: 3 September 2019; Accepted: 13 November 2019; Published: 19 November 2019

**Abstract:** Signal transducer and activator of transcription (STAT) proteins are transcription factors that in the latent state are located predominantly in the cytoplasm. Activation of STATs through phosphorylation of a single tyrosine residue results in nuclear translocation. The requirement of tyrosine phosphorylation for nuclear accumulation is shared by all STAT family members but mechanisms of nuclear translocation vary between different STATs. These differences offer opportunities for specific intervention. To achieve this, the molecular mechanisms of nucleocytoplasmic shuttling of STATs need to be understood in more detail. In this review we will give an overview on the various aspects of nucleocytoplasmic shuttling of latent and activated STATs with a special focus on STAT3 and STAT5. Potential targets for cancer treatment will be identified and discussed.

**Keywords:** STAT3; STAT5; nuclear pore complex; nuclear transport receptors; nucleocytoplasmic shuttling; cancer; targeting

#### **1. Aim and Scope**

STAT (Signal transducer and activator of transcription) proteins can be seen as intracellular messengers that relay signals sensed at the plasma membrane to chromatin and genes in the nucleus. To achieve this, STATs must pass the nuclear envelope through nuclear pore complexes (NPCs). Thus, passage through the NPC is an essential step in the sequence of events from activation of STATs at cytokine receptors to DNA-binding and target gene induction. As detailed in the reviews and articles of this Special Issue of Cancers, deregulated activation of STAT3 and STAT5 contributes to various cancers in many ways. Thus, STAT3 and STAT5 proteins have emerged as promising therapeutic targets. Protein–protein interactions involved in nucleocytoplasmic transfer of STATs have not been exploited yet as molecular targets for intervention. Detailed knowledge of the involved molecules and mechanisms is an essential prerequisite for successful and specific targeting. In this review, we will first describe the general mechanisms involved in import and export of proteins in and out of the nucleus, concentrating on those which are most relevant for transcription factors. We will then focus on the mechanisms involved in nucleocytoplasmic shuttling of STAT3 and STAT5 and finally assess possible molecular targets for specific intervention.

#### **2. General Mechanisms of Nucleocytoplasmic Transport of Proteins**

#### *2.1. The Nuclear Pore Complex*

To enter or exit the nucleus, proteins must pass through the nuclear pore complex (NPC) [1]. NPCs are huge macromolecular assemblies (about 120 MDa in humans) made up of multiple copies of Nucleoporins (NUPs). More than 30 different NUPs have been identified that are built into the NPC as

multiples of eight (8–64) resulting in the eight-fold rotational symmetry of the NPC [2]. The NPC can be seen as a channel that allows selective transfer of cargo and at the same time forms a soft barrier for free diffusion of macromolecules larger than about 30 kDa [3,4], preventing their access to the nucleus without permission. The barrier is formed by phenylalanine-glycine (FG)-repeats that protrude from certain NUPs into the lumen of the channel [5,6].

How exactly the FG-repeats form a selective permeability barrier is not completely understood and several models are currently being discussed [7,8]. One of the most prevalent is the selective phase model that relies on interactions between the FG-repeats creating a sieve-like meshwork with hydrogel-like properties, which would explain the observed mass exclusion limit [9]. Selectivity for cargo allowed to pass might result from phase separation that prevents passage of macromolecules that are unable to mix or interact with the selective phase made up by the FG-repeats [8].

The import/export pathways through the NPC involve soluble nuclear transport receptors (NTRs) that bind cargo in conjunction with the Ran-GTP/GDP cycle. NTRs can interact with FG-repeats [10] and facilitate passage of bound cargo through the NPC. According to the selective phase model, interaction of NTRs with the FG-repeats leads to local disturbance of the meshwork allowing the NTR/cargo complex to travel almost freely between cytoplasm and nucleoplasm [11]. The energy-consuming Ran-GTP/GDP cycle provides directionality of the transport through control of cargo to NTR binding which is differently regulated in nucleoplasm vs. cytoplasm [12,13].

NTRs, also known as Karyopherins, can be subdivided in Importins and Exportins, facilitating nuclear import and export, respectively. Biportins have also been described that support transport of cargo in both directions with imported and exported cargo being distinct [14].

#### *2.2. Importins*

The best characterized Karyopherin is Importin-β1 which either binds cargo directly or indirectly through interaction with adapters such as α-Importins or Snurportin-1 [15]. The transcription factors Snail1 [16,17] and SREBP2 [18] are among the cargoes directly bound by Importin-β1. Snurportin-1 is best known for its involvement in the nuclear import of spliceosomal snRNPs [19]. The interaction site of Importin-β1 with FG-repeats has been mapped [20,21] and is different from the well-defined interaction sites with cargoes and adapters [22]. This means that cargo-loaded Importin-β1 can interact with FG-Nucleoporins of the NPC and thereby facilitate passage of the Importin-β1/adapter/cargo complex.

The import pathway using α-Importins as adapters to connect cargo with Importin-β1 has been intensively studied and is now known as the classical import pathway [23]. Accordingly, the term classical nuclear localization signal (cNLS) refers to linear sequence motifs of cargoes that bind to α-Importins. The cNLS can be further subdivided into monopartite cNLS and bipartite cNLS. Monopartite cNLS consist of a short stretch of basic amino acid residues, the sequence PKKKRRV of the SV40 large T-antigen being the first identified [24]. The first bipartite cNLS has been found in the Nucleoplasmin protein of Xenopus laevis, consisting of two short stretches of basic amino acids, both essential for its function, separated by a few less relevant amino acids [25].

In humans, seven Importin-α isoforms have been identified: Importin-α1, -α3, -α4, -α5, -α6, -α7 [23] and the most recently discovered Importin-α8 [26]. All α-Importins are mainly built up of ten Armadillo (ARM)-repeats resulting in a structurally conserved solenoid protein domain [27]. The cNLS of cargo binds along a groove on the inner concave surface of α-Importins. Although the cNLS binding region is quite conserved, the α-Importins are specific for a set of cargo proteins with some considerable overlap. This specificity results in part from preferential binding of certain NLS but also from the three-dimensional context in which the NLS is presented by the cargo [28]. Apart from cNLS, so called non-classical or atypical NLS have been identified, which do not fit in the cNLS consensus motifs [29]. A short sequence termed Importin- β binding (IBB) domain precedes the solenoid domain of α-Importins. In the absence of cargo, the IBB folds back to the NLS binding site. Upon binding of cargo, the IBB is replaced by the cNLS of the cargo protein. The now exposed IBB binds to Importin-β1, facilitating the nuclear import of the ternary Importin-β1/Importin-α/cargo complex [30].

Transportin-1, also known as Importin-β2, also imports a broad spectrum of cargoes, including transcription factors and mRNA-binding proteins. Transportin-1 binds cargo directly through a broad range of loosely related NLS that are different from cNLS [31]. Symportin-1 has been identified as an adapter for nuclear import of some ribosomal proteins by Transportin-1 [32]. Transportin-3 recognizes arginine-serine (RS)-rich NLS found in proteins typically involved in mRNA metabolism such as the Splicing Factor 2 (SF2) [33]. Importin-13 is closely related to Transportin-3 but mediates both protein import and export. Therefore, by definition Importin-13 can be regarded as a Biportin. Some transcription factors are among the many cargoes identified so far. Cargo binding by Importin-13 relies on the recognition of folded domains rather than linear NLS [34,35].

Some import pathways have been identified that do not rely on β-Importins. Among those, Calmodulin-mediated nuclear import seems to be most relevant for transcription factors [36]. Finally, some proteins including transcription factors enter the nucleus independent of NTR or other carrier proteins, e.g., through direct interaction with NUPs [37].

#### *2.3. Exportins*

Crm1 (Chromosome region maintenance 1; in the systematic nomenclature of Karyopherins designated as Exportin-1 or Xpo1) is the most promiscuous Exportin, which mediates the export of about 1000 substrates, including various RNAs, ribonucleoproteins, and transcription factors [38]. Most protein cargo of Crm1 contains a leucine-rich nuclear export signal (NES) of 8–15 amino acids [39–41]. Like all β-Karyopherins, Crm1 is built up of HEAT-repeats (Huntingtin, Elongation factor 3, protein phosphatase 2A and TOR kinase). Each HEAT-repeat consists of two α-helices connected by loops of varying length. Similar to the ARM repeats of α-Importins, the HEAT-repeats of β-Karyopherins form a solenoid protein domain [42] with a slightly curved superhelical structure of high conformational flexibility. In all of the available structures of cargo bound to Crm1, the NES of the cargo fits into a hydrophobic groove between HEAT repeats 11 and 12 (out of 20) at the outer convex surface of Crm1 [43]. Binding at the outer convex surface of Crm1 allows accommodation of cargo of various sizes.

Other Exportins, such as Exportin-5, Exportin-t, and Exportin-2 (better known as CAS for Cellular Apoptosis Susceptibility), bind their cargo on the inner concave surface. Accordingly, these Exportins are more selective. Exportin-5 mediates export of several types of RNA but not protein; Exportin-t is even more restricted to the export of tRNAs [44]. CAS is specialized for the nuclear export of α-Importins, which is needed to maintain the classical import pathway [45]. Exportin-4, a distant member of the β-Karyopherin family, originally described to mediate nuclear export of the translation initiation factor 5A [46] and the transcription factor Smad3 [47], has more recently been described to be involved in the nuclear import of Sox transcription factors [48], thus acting as a Biportin. Another Biportin, Exportin-7 (also known as RanBP16) [49], has recently been described to act as a broad-spectrum Karyopherin with about 200 export and 30 import substrates [50].

#### *2.4. Ran-GTP*/*GDP Cycle*

In principle, through interaction with the FG-repeats, β-Karyopherins can pass the NPC in both directions [51]. Directionality of nucleocytoplasmic transport is achieved through coupling of cargo binding to β-Karyopherins with the energy-consuming Ran-GTP/GDP cycle [12,52]. Ran is a small G-protein of the Ras superfamily of GTPases that exists in a GTP- or GDP-bound state. The GTP-bound state results from replacement of bound GDP with GTP that requires the activity of a Ran-specific, chromatin-bound guanine-nucleotide exchange factor (RanGEF, also known as RCC1), which is predominantly located in the nucleus. Therefore, the Ran-GTP concentration in the nucleus is high. The GDP-bound state is generated through release of the β-phosphate of bound GTP by hydrolysis, which requires the activity of a ran-specific GTPase-activating protein (RanGAP). Through interaction with the Nucleoporin Ran-binding protein 2 (RanBP2, Nup358), RanGAP is located predominantly at the cytoplasmic surface of the NPC. Therefore, in the cytoplasm close to the nuclear envelope, the Ran-GTP concentration is low and the Ran-GDP concentration is high [53,54].

Cargo binding to Importins in the cytoplasm is Ran-independent. After passage of the Importin/cargo complex through the NPC, release of cargo in the nucleus is induced upon Ran-GTP binding. In contrast, Ran-GTP-binding to Exportins is required for cargo loading in the nucleus. After passage of the Exportin/cargo/Ran-GTP complex through the NPC, RanGAP-mediated GTP-hydrolysis results in release of RanGDP and cargo into the cytoplasm [55].

As mentioned above, NTRs facilitate nucleocytoplasmic transport through transient interactions with the hydrophobic FG-repeats within the channel of the NPC. In principle, other proteins with hydrophobic surface patches should to some extent also be able to disturb the FG-repeat meshwork, allowing them to shuttle between the cytoplasm and the nucleus in an NTR- and Ran-independent, passive manner. Indeed, in a recent study [56] it has been found that a continuum exists with gradual differences in the ability of proteins to pass the NPC in an active (i.e., NTR- and Ran-dependent) or passive manner. Depending on their surface properties, even large proteins can leak into the nucleus without involvement of NTRs.

#### **3. STAT1—Using a Side Track for Nuclear Import**

STAT1 is deeply involved in the antiviral response triggered by endogenously produced Interferons but also in the therapeutic responses to exogenously administered Interferons in cancer treatments and anti-viral therapies [57]. STAT1 is best known for its tumor suppressive activity in cancer but some tumor-promoting effects have also been documented [58]. IFNγ (type II Interferon) induces the activation of STAT1 through phosphorylation of Y701 by IFNγ receptor-associated Janus kinase 1 (JAK1) and JAK2, leading to dimerization and nuclear accumulation of the transcription factor [59]. Among all STAT family members, the mechanism of nuclear accumulation of the activated STAT1 dimer is best understood. Activated STAT1 interacts with Importin-α5 [60], but in an unconventional manner. Instead of employing the binding site for cNLS, which would involve ARM repeats 2–4 and 6–8 as the major and minor binding sites, respectively [22], STAT1 binds to the more C-terminally located ARM repeats of Importin-α5 [61] involving a critical tyrosine residue (Y476) located in ARM repeat 10 [62]. Accordingly, the activated STAT1 dimer does not expose a cNLS but a dimer-specific surface area to interact with Importin-α5. This surface has been termed dimer-specific NLS (dsNLS) and includes regions within the N-terminal domain (NTD) [63] and the DNA-binding domain (DBD; see Figure 1a for a scheme of the domain structure of STAT proteins; see Table 1 for an overview on putative NLS and NES sequences of STAT proteins; and see Figure 1b showing the corresponding putative NLS and NES motifs in the STAT1 dimer structure) [64]. Hence, Importin-α5 can be displaced from STAT1 by DNA-binding [65,66]. Most interestingly, this unique mode of interaction is exploited by the Ebola virus VP24 protein, that interacts with ARM repeats 8–10 of Importin-α5, preventing binding and nuclear translocation of activated STAT1 [67]. Interaction of Importin-α5 with cNLS cargo remains unaffected by VP24 binding. This example shows that, in principle, nuclear accumulation of a STAT transcription factor can be selectively blocked.

**Figure 1.** Structures of STAT proteins with putative nuclear localization signals (NLS) and nuclear export signals (NES) highlighted as listed in Table 1. In this context, "putative" means that the corresponding sequences do not fulfill classical NLS or NES functions but are required for nuclear transport or interaction with nuclear transport receptors. (**a**) General scheme of structural domains of STAT proteins (NTD, N-terminal domain; CCD coiled-coil domain; DBD, DNA-binding domain; LD, linker domain; SD, SH2 domain; TAD, transactivation domain). Numbers refer to amino acid positions. (**b**–**d**) Structures of individual STAT proteins as ribbon representations (left) or space-filling representations (middle and right). Domains are stained according to the coloring of the scheme in (**a**), DNA is shown in pink. Putative NLS and NES are marked in red and cyan, respectively. The corresponding references are listed in Table 1. PDB IDs: STAT1, 1BF5; STAT3, 1BG1; STAT5A, 1Y1U. Images of structures were generated with PyMOL.


**Table 1.** NTRs, NLS and NES involved in transport of STAT proteins.

<sup>1</sup> n.d., not determined; <sup>2</sup> in this context, "putative" means that the corresponding sequences do not fulfill classical NLS or NES functions but are required for nuclear transport or interaction with NTRs.

#### **4. STAT3—Acting on Many Stages**

#### *4.1. Role of STAT3 in Cancer*

STAT3 is involved in cancer in multiple ways, as detailed in several reviews and articles of this Special Issue of Cancers. Besides acting as an oncogene in a cell autonomous manner [80,81], activation of STAT3 in the tumor microenvironment contributes to metastasis, angiogenesis, cancer stem cell maintenance, and immune evasion [82]. Therefore, blocking STAT3 activity in cancer is of particular interest. However, it should also be noted that in some cancers STAT3 has tumor suppressive activities [83,84]. Several steps in the STAT3 activation pathway are potentially targetable and thus might be used to interfere with STAT3 activation. These include phosphorylation on Y705 by (oncogenic) tyrosine kinases, dimerization through phosphotyrosine/SH2 domain interactions, nuclear translocation, DNA-binding to GAS (γ-interferon activated sequence) elements, and interaction with cofactors. However, passage of the activated STAT3 dimer through the NPC has only rarely been considered as a potential molecular target for intervention [85,86]. This might be due to the fact that import mechanism for activated STAT3 and the interactions of STAT3 with NTRs have been analyzed in some detail but are not completely understood as outlined below.

#### *4.2. Nuclear Import*

Compared to STAT1, the mechanism of nuclear import of the activated STAT3 dimer is less well defined. Reports on interactions with α-Importins are contradictory in part. In one of the first studies, interactions of STAT3 with Importin-α1, -α3, and -α5 were detected [70]. Additionally, interactions of STAT3 with Importin-α5 and -α7, but not with Importin-α1, -α3, and -α4, were found [69]. In both reports, interactions were only seen upon activation of STAT3. Other groups found interactions with Importin-α3 and -α6 [71] and Importin-α3, -α5, and -α7 [72] being independent of activation of STAT3. In this context it is worth noting that expression of Importin-α6 is restricted to testis [87]. Therefore, Importin-α6 is less relevant for nuclear import of ubiquitously expressed transcription factors such as STAT3 and STAT5. Based on import assays with permeabilized cells, the MgcRacGAP protein bound to Rac has been suggested as an NLS-containing adapter for Importin-α-mediated nuclear import of STAT3 and STAT5 [88]. In another study using an MgcRacGAP inhibitor and siRNA mediated knockdown, the functional role of MgcRacGAP for nuclear import of STAT3 could not be substantiated [89]. These discrepancies might be the result of different experimental conditions and cellular systems used. For instance, subcellular distribution of STAT3 analyzed by immunofluorescence heavily depends on the applied fixation method [90]. Alternatively, the different findings indicate that STAT3 employs a different import system depending on the cytokines used for its activation, e.g., Interleukin-6 (IL-6) vs. Epidermal Growth Factor (EGF). Clearly, more work is required for better understanding.

Functional relevance of Importin-α5 in nuclear import of activated STAT3 has been demonstrated, as well as the involvement of Importin-β and Ran [70]. Moreover, it has been shown that the epitope on Importin-α5 employed in the interaction with STAT3 is different from the one used for the interaction with STAT1. A mutation in ARM repeat 10 of Importin-α5 blocks interaction with STAT1 [62] but does not affect the interaction with STAT3 [72]. Correspondingly, the epitopes used by STAT3 and STAT1 for interaction with Importin-α5 are also different. Amino acids in the DBD of STAT1 are involved in the recognition of Importin-α5 [65,66], connecting Importin-binding and nuclear presence with DNA-binding activity of STAT1 [91]. In STAT3, mutation of R214 and R215 located in the coiled-coil domain (CCD) abolishes binding of Importin-α5 and nuclear accumulation in response to stimulation [69]. Liu et al. [71] defined the sequence D150-K163 as being indispensable for nuclear import of STAT3, which is also located in the CCD. Nuclear accumulation of activated STAT3 does not require DNA-binding activity [72]. Common to STAT3 and STAT1 is the involvement of the N-terminal domain in stimulation-dependent nuclear import [63,74,75]. The essential role of the NTD for active nuclear import of STAT3 is further supported by the observation that the tumor suppressor ARHI

(A Ras Homologue member I/DIRAS3) blocks nuclear translocation of phosphorylated STAT3 via direct interaction with the NTD [92].

It seems that Importin-α5 is a major determinant for nuclear import of STAT3 and STAT1 but the molecular interfaces of the interactions are considerably different. This opens an avenue for specific interventions. However, to achieve a selective block of nuclear import of STAT3, the structural details of the interactions with NTRs must be characterized in more detail. It should be noted that no classical NLS or NES sequences have been identified in STAT3 in the sense of transferable functional motifs. The above-mentioned amino acids in the putative NLS and NES are most probably important for the integrity of domain structures or the part of epitopes that interact with NTRs in a non-conventional manner [75], as exemplified for STAT1 [62] (Figure 1c showing putative NLS and NES motifs in the activated STAT3 dimer structure).

#### *4.3. Nuclear Export and Nucleocytoplasmic Shuttling*

In the course of defining nuclear export signals of STAT3, it was found that pharmacological inhibition of Crm1-mediated nuclear export results in partial nuclear accumulation of STAT3 independent of cytokine stimulation [79]. From these and other studies it became evident that the subcellular distribution of latent STAT3 with high cytoplasmic and low nuclear concentration is the result of a steady-state of constitutive nuclear import and rapid export independent of phosphorylation at Y705 [93,94]. Determinants of stimulation-dependent nuclear import, such as the NTD and the R214/R215-motif, are not involved in basal nuclear import of latent STAT3 [75]. Unphosphorylated STAT1 has also been detected in the nucleus [95], and the mechanisms of nuclear import of latent and activated STAT1 were also found to be different [64]. Nuclear import of unphosphorylated STAT3 might be independent of NTRs and could involve direct interactions with FG-repeat NUPs of the NPC as observed for STAT1 [96]. From the three sequence motifs identified as putative NES in STAT3, at least the one containing L525/L528 is involved in constitutive nucleocytoplasmic shuttling [75,79].

It has been firmly established that STAT3 and STAT1 form homodimers in the absence of the activating tyrosine phosphorylation resulting in so called preformed dimers [97,98]. The phosphotyrosineindependent dimerization involves homotypic interactions of the NTDs [99]. However, basal nucleocytoplasmic shuttling of STAT3 does not require the NTD [75] and is not sensitive to a single point mutation (L78R) that prevents preformed dimer formation [100]. Not only is dimer formation not required for basal shuttling, but monomeric STAT3 shuttles even faster [75]. The increased shuttling rate of monomeric STAT3 could be attributed to the smaller size compared to the preformed dimer. However, a recent study suggests that hydrophobic surface patches that interact with FG-repeats of FG-Nups might be a stronger predictor for NTR-independent passage of the NPC than size [56]. Thus, accessible hydrophobic surfaces that are masked in the dimer may facilitate faster shuttling of monomeric STAT3 in an NTR-independent manner.

Since basal and stimulation-dependent nuclear import of STATs occur through different mechanisms, they could, in principle, be targeted independently of each other. However, the functional relevance of nuclear presence of unphosphorylated STAT3 and STAT1 (U-STATs) is not entirely clear. Nuclear U-STATs might be involved as cofactors in gene regulation independent of their DNA-binding activity [101].

#### **5. STAT5—Leukemia and More**

Important functional roles in hematopoiesis downstream of hematopoietic cytokines such as Erythropoietin and Thrombopoietin have been attributed to STAT5. Consequently, deregulated STAT5 signaling is prominently implicated in myeloproliferative diseases and leukemias [102]. The involvement of STAT5 in solid tumors is also well documented, both as an oncogene and a tumor suppressor [103], as detailed in several reviews and articles of this Special Issue of Cancers. In particular, in myeloproliferative neoplasms and leukemias, STAT5 has been identified as a promising therapeutic target.

Compared with the many reports on STAT3 and STAT1, only a handful of studies dedicated to the molecular mechanisms governing nucleocytoplasmic distribution of STAT5 exist. In one of the first studies [104], a sequence motif in the DNA-binding domain of STAT5B (V466-I469) was identified whose mutation prevented Growth Hormone-induced nuclear accumulation and DNA-binding. However, from this finding it cannot be concluded that DNA-binding per se is required for nuclear accumulation of activated STAT5B because the mutation could also interfere with binding of NTRs, which was not investigated in this study. We found that another mutation that affects DNA-binding of STAT5A does not impair nuclear accumulation upon Epo stimulation, in agreement with the DNA-binding independent nuclear accumulation of activated STAT3 [72]. In general, nuclear accumulation of STATs through retention by binding to nuclear structures seems to be of minor importance. If nuclear accumulation was dependent on retention of STATs on subnuclear structures, nuclear accumulation would be saturable, meaning that as soon as all binding sites are occupied, the remaining STAT molecules would not accumulate in the nucleus. This is not what is observed in the above-mentioned experiments with transfected cells: Even upon forced overexpression, STATs almost completely accumulate in the nucleus upon stimulation. This view is supported by the results of fluorescence recovery after photobleaching (FRAP) experiments that have been exemplarily performed on STAT1-GFP. Upon nuclear accumulation induced by Interferon treatment, STAT1-GFP freely diffuses through the nucleoplasm with the exclusion of nucleoli [105]. In this and another study [93], the involvement of the cytoskeleton in directed nuclear import of STAT1 and STAT3 was also excluded.

In one of the reports on binding of α-Importins to STAT3, binding to STAT5A and STAT5B was analyzed in parallel but no interaction could be detected [69]. This is in agreement with our own observations made with co-precipitation experiments from cellular lysates using Importins fused to GST (glutathione-S-transferase). Accordingly, no cNLS has been detected in STAT5. In another study, interaction of STAT5A with Importin-α3 has been shown [73], again supporting the notion that the outcomes of in vitro assays (mostly co-precipitations) for studying the interaction of STATs with NTRs are very sensitive to the experimental conditions used.

Similar to STAT3 and STAT1, constitutive nucleocytoplasmic shuttling of STAT5 independent of tyrosine phosphorylation has been detected [77]. A functional role of nuclear U-STAT5 has been established in megakaryocyte differentiation [106]. Basal and cytokine-induced import of STAT5A requires an intact CCD [73] and was abolished upon deletion of eight amino acids in the CCD (L142–E149, see Table 1 and Figure 1d showing putative NLS and NES motifs in the STAT5A monomer structure; a structure of the activated STAT5 dimer is not available) [76]. Interaction of STAT5A with Importin-α3 occurs in an unconventional manner involving the CCD [73]. As for STAT3 and STAT1, stimulation-dependent but not basal nuclear translocation requires the NTD of STAT5 [77]. In a study that tested all sequence motifs that might function as a cNES, a region between aa578 and aa675 in STAT5B comprising the SH2 domain was identified to be required for nuclear export [77]. Another region important for Crm1-mediated export was identified involving L119 and L133 in the NTD of STAT5A [73].

A discrepancy between nuclear localization of STAT5A and STAT5B was found in response to phosphorylation of the critical tyrosine residue (Y694 in STAT5A and Y699 in STAT5B) by Src family kinases (SFK). While phosphorylation of Y699 in STAT5B, as well as Y705 in STAT3, by Src leads to nuclear accumulation of the transcription factors, phosphorylation of Y694 in STAT5A by Src does not [107]. Later it was found that this observation is of some relevance in Bcr-Abl-positive chronic myeloid leukemia (CML), where STAT5 is predominantly localized in the cytoplasm despite being phosphorylated at the critical tyrosine residue [108]. Cytoplasmic retention of STAT5A in response to activation by Bcr-Abl was found to be mediated by SFK such as Src and Hck. Specific inhibition of SFKs resulted in nuclear accumulation of STAT5A, enhanced STAT5 target gene expression, and increased colony formation of CML cells [109]. SFK interfere specifically with dimerization of activated STAT5A and thereby prevent nuclear accumulation [110]. In another report, phosphorylation of S779 has been suggested as an additional requirement for the nuclear translocation of STAT5A in Bcr-Abl positive

cells [111]. These intriguing examples show that even the highly homologous transcription factors STAT5A and STAT5B differ in some aspects of nucleocytoplasmic shuttling and these differences might be relevant for disease.

#### **6. Perspectives for Therapeutic Interventions**

Nucleocytoplasmic transport has been recognized as a target for cancer therapy [112,113] in part based on the observation that components of the nuclear transport machinery are differentially expressed in transformed cells [114]. These changes in expression can be relevant for disease. It has been shown that Crm1-mediated nuclear export contributes to drug resistance in multiple ways [115]. For example, Topoisomerase II inhibitors such as doxorubicin require nuclear Topoisomerase activity. As a drug resistance mechanism, Topoisomerase II can be exported from the nucleus in a Crm1-dependent manner [116]. Treatment with Crm1 inhibitors therefore sensitizes cancer cells to doxorubicin treatment [117]. Only recently, in two independent studies, was Crm1 convincingly identified as a synthetic lethality gene in different types of cancer [118,119]. Earlier clinical studies already showed that the highly efficient Crm1 inhibitor leptomycin B is not well tolerated by patients [120]. Less toxic small molecule inhibitors of Crm1 termed Selective Inhibitors of Nuclear Export (SINE), such as selinexor, are currently used in clinical trials in patients with hematological malignancies and solid tumors [112]. Most recently, selinexor in combination with dexamethasone has been approved for the treatment of refractory multiple myeloma. In this context, it has already been recognized that a more selective export inhibition could be more effective and reduce the serious side effects of treatment [115]. There are only a few inhibitors for some other NTRs available (Importin-β1, Importin-α/β1 heterodimers, and Transportin-1) which have not been explored in such detail yet [121].

For transcription factors that respond to extracellular cues, control of nucleocytoplasmic distribution contributes to efficient signal transduction. High cytoplasmic concentration of the latent transcription factor increases the sensitivity to stimulation, which usually occurs at the cytoplasmic face of membrane-bound receptors. After stimulation and activation of the transcription factor, high nuclear concentration increases the efficiency of DNA-binding and thus gene induction. However, the molecular mechanisms involved in the control of subcellular distribution vary considerably between different transcription factors. For instance, the NLS of NF-κB is masked by IκB [122]. Upon sensing of inflammatory mediators by membrane-bound receptors, IκB is phosphorylated and ubiquitinated resulting in its proteasomal degradation. The now exposed NLS facilitates α-Importin-mediated nuclear accumulation of NF-κB [123]. Likewise, the NLS of some nuclear receptors, such as the Glucocorticoid Receptor and Estrogen Receptor, are masked. Ligand binding to the receptors results in release of Hsp90 and exposition of the NLS [124]. Similar cytoplasmic interaction partners that would prevent nuclear translocation of latent STAT proteins have not been identified yet. Various other mechanisms involving SUMOylation, ubiquitination, mono-ADP-ribosylation, and acetylation have been described to control the nucleocytoplasmic distribution of transcription factors [22,125]. Beside the well-established Karyopherin-mediated transport, other non-conventional transport mechanisms exist [126]. Nucleocytoplasmic transport of β-Catenin is regulated independently of Ran and classical transport factors [127,128]. At the same time, β-Catenin acts as a transport receptor for its transcription factor partner LEF-1 [129]. The regulatory 14-3-3 proteins have also been suggested to be involved in such a piggy-back mechanism of nuclear transport [130,131], which has so far not been described for STAT proteins. Because of these various transport modes, specific interference should be feasible in principle.

In the case of STAT transcription factors, a non-classical NLS is generated through phosphorylationinduced dimerization (dsNLS) that is recognized by α-Importins or other yet uncharacterized factors. Because this mechanism is quite unique, a specific intervention through targeting the dsNLS/α-Importin interface should in general be feasible, as exemplified for STAT1 [67]. Extended surface areas are involved in the STAT/NTR interaction, including the NTD, DBD and CCD. The NTD seems to be a common denominator of nuclear import of activated STATs. Because the NTDs of different STATs vary

in their sequence, a specific targeting should be possible, as exemplified by the specific interaction of ARHI with the NTD of STAT3, but not with the NTDs of STAT5 and STAT1 [92]. The impact of the ARHI/STAT3-NTD interaction on Importin-binding was further analyzed in a study applying homology modelling and molecular dynamics simulation [132]. Indeed, targeting of the NTD of STAT3 has been reported, however, with a focus on the role of the NTD as a tetramerization domain at enhancers and its function in gene induction [133]. The more recently solved structure of the NTD of STAT3 [134] should pave the way for a closer inspection of the surface areas involved in nuclear import by mutation of surface-exposed residues. For nuclear import of STAT5 and STAT3, data concerning use of the classical α-Importin/β1-Importin pathway are conflicting. Here, a fresh approach is needed to identify the NTRs involved, including α-Importins/Importin-β1-independent pathways mentioned in Section 2 of this review.

Reduced export of the imported STAT dimers also contributes to nuclear accumulation because export and redistribution requires dephosphorylation [135]. Blocking nuclear export of STATs as a therapeutic approach seems at first sight counterintuitive because it might trap the activated transcription factor in the nucleus where it could drive persistent oncogenic gene induction. However, STAT1, STAT3, and STAT5 are dephosphorylated in the nucleus by TC45, the 45 kDa isoform of T Cell Protein Tyrosine Phosphatase (TC-PTP) [136–138]. For STAT1 it has been shown that only the dephosphorylated form leaves the nucleus for reactivation at the receptor [91]. Exit out of the nucleus is required for reactivation of STATs [139]. Thus, STAT-mediated gene induction could be inhibited by trapping the dephosphorylated, transcriptionally inactive STATs (in the sense of canonical STAT signaling) in the nucleus through blockade of nuclear export. This has already been shown for STAT3 using the Crm1 inhibitor ratjadone A [140]. The more clinically advanced SINE have also been tested. These compounds inhibited STAT3-mediated Survivin expression in breast cancer models [141]. To interfere more specifically with export of STATs, the Exportins or Biportins involved must be characterized and the molecular interfaces of the STAT/NTR interaction mapped in detail.

Mechanistically, the export of dephosphorylated STATs most probably corresponds to the exit mechanisms involved in constitutive shuttling of latent STATs. Treatment of cells with specific inhibitors for the Exportin Crm1 often leads to only a partial nuclear accumulation of latent STATs. This means that either passive NTR-independent export occurs through direct hydrophobic interactions with FG-Nups or that other Exportins are involved. Specifically for STAT5A, a Crm1-independent export signal in the DBD has been mapped [73]. Redundancy in cargo recognition by NTRs in the sense that one cargo is transported by several NTRs is not unusual. As outlined in Section 2, broad-specificity Exportins or Biportins have been identified in recent years, worthy of analysis with respect to their involvement in the nuclear export of STAT proteins.

#### **7. Conclusions**

Taken together, effective inhibition of oncogenic STAT activity can in principle be achieved through specific blockade of both nuclear import of phosphorylated STAT dimers and export of dephosphorylated STATs out of the nucleus. However, this requires a deeper understanding of the protein–protein interactions involved in these processes. First, all NTRs facilitating nucleocytoplasmic shuttling need to be unambiguously identified for the individual STAT proteins. These studies should also include those NTRs that have been characterized in more detail in recent years. Then, the interactions between STATs and NTRs must be mapped in more detail using all structural data available. The ultimate goal would be to solve the structure of STAT/NTR complexes by X-ray crystallography or cryo-electron microscopy. Furthermore, the contribution of NTR-independent nuclear transport of STATs should be evaluated through assessment and mutation of surface patches as delineated by Frey et al. [56]. Based on solid structural and functional data, specific blockade of nucleocytoplasmic transport of individual STATs by tailor-made inhibitory molecules might be feasible.

**Author Contributions:** S.E. edited the manuscript, prepared figure and table and searched the literature, G.M.-N. drew the conceptual framework and wrote the manuscript.

**Funding:** This research was funded by the Interdisciplinary Center for Clinical Research (IZKF) Aachen within the Faculty of Medicine at RWTH Aachen University.

**Acknowledgments:**Thisworkwas supported by the ConfocalMicroscopy Facility, a core facility of the Interdisciplinary Center for Clinical Research (IZKF) Aachen within the Faculty of Medicine at RWTH Aachen University. The authors would like to thank Wolfram Antonin and Bernhard Lüscher for critical reading of the manuscript. The authors apologize to all colleagues whose contributions have not been cited.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

### *Review* **TYK2: An Upstream Kinase of STATs in Cancer**

#### **Katharina Wöss, Natalija Simonovi´c, Birgit Strobl, Sabine Macho-Maschler and Mathias Müller \***

Institute of Animal Breeding and Genetics, University of Veterinary Medicine Vienna, A-1210 Vienna, Austria; Katharina.Woess@vetmeduni.ac.at (K.W.); natalijabozovic@gmail.com (N.S.);

birgit.strobl@vetmeduni.ac.at (B.S.); sabine.macho-maschler@vetmeduni.ac.at (S.M.-M.)

**\*** Correspondence: mathias.mueller@vetmeduni.ac.at

Received: 30 September 2019; Accepted: 2 November 2019; Published: 5 November 2019

**Abstract:** In this review we concentrate on the recent findings describing the oncogenic potential of the protein tyrosine kinase 2 (TYK2). The overview on the current understanding of TYK2 functions in cytokine responses and carcinogenesis focusses on the activation of the signal transducers and activators of transcription (STAT) 3 and 5. Insight gained from loss-of-function (LOF) gene-modified mice and human patients homozygous for *Tyk2*/*TYK2*-mutated alleles established the central role in immunological and inflammatory responses. For the description of physiological TYK2 structure/function relationships in cytokine signaling and of overarching molecular and pathologic properties in carcinogenesis, we mainly refer to the most recent reviews. Dysregulated TYK2 activation, aberrant TYK2 protein levels, and gain-of-function (GOF) TYK2 mutations are found in various cancers. We discuss the molecular consequences thereof and briefly describe the molecular means to counteract TYK2 activity under (patho-)physiological conditions by cellular effectors and by pharmacological intervention. For the role of TYK2 in tumor immune-surveillance we refer to the recent Special Issue of Cancers "JAK-STAT Signaling Pathway in Cancer".

**Keywords:** tyrosine kinase 2; JAK family of protein tyrosine kinases; signal transducer and activator of transcription; cytokine receptor signaling; gain-of-function mutation; tumorigenesis

#### **1. TYK2-Mediated Cytokine Signaling and Activation of STAT3 and STAT5**

TYK2 was the first identified member of a family of non-receptor kinases later termed Janus kinases (JAK), which additionally comprises JAK1-3 [1,2]. JAKs are associated with cytokine and growth factor receptors and activate STAT (STAT1-4, STAT5A, STAT5B, STAT6) family members [2,3]. JAKs share four functional domains (from N- to C-terminal): (i) a four-point-one, ezrin, radixin, moesin (FERM) homology domain; (ii) an atypical Src-homology 2 (SH2) domain, both facilitating protein-protein interactions (PPIs); (iii) a kinase-like or pseudokinase (JAK homology (JH) 2) domain negatively regulating the kinase activity; and (iv) a tyrosine kinase (JH1) domain which, upon conformational changes at ligand bound receptors, increases its catalytic activity by trans-/autophosphorylation of its activation loop [2,4].

To date, the requirement for TYK2 in signaling has been shown for numerous cytokines, including distinct interleukin (ILs) and interferons (IFNs), which comprise several subtypes (i.e., type I and III IFNs). The heterodimeric cytokine receptor complexes are composed of four distinct TYK2-associated receptor chains (IFNAR1, IL-12Rβ1, IL-10R2, and IL-13Rα1) and a respective second receptor chain associated either with JAK1 or JAK2, which serves as the signal transducing chain harboring STAT docking sites. Usually, these sites contain critical tyrosine residues that are phosphorylated by JAKs upon receptor complex activation (Figure 1). TYK2 also associates with the gp130 receptor chain, yet there is no evidence that gp130-utilizing cytokines rely on TYK2 for signal transduction [5,6]. Note that comprehensive reviews [2,7] provide lists of various other receptors utilizing TYK2-STAT signaling; however, TYK2-STAT activation/utilization is frequently only biochemically assessed by

phosphorylation of critical tyrosine residues and cannot be put on a level with dissected downstream cellular activities. Here we review the cytokines which clearly transduce the TYK2 phosphorylation events into downstream physiological changes (Figure 1).

**Figure 1.** Cytokine receptor families signaling with the participation of TYK2 and JAK1 or JAK2. Cytokines are depicted only upon appearance in humans and mice and proof of TYK2 dependency. The color codes indicate the major STAT(s) activated by the respective cytokines. STAT1-STAT2 heterodimers combine with IFN regulatory factor (IRF) 9 and form the interferon-stimulated gene factor 3 (ISGF3) complex; \* STAT activation is dependent on cell type or of less clear biological relevance.

The biological relevance for TYK2-dependent cytokines activating STAT3 is best established for the IL-10R2 utilizing IL-22 [8,9] and the IL-12Rβ1-utilizing IL-12 and IL-23 [10–12]. IL-22 is a central cytokine in tissue-barrier function, wound healing, and epithelial homeostasis and repair. Cancer promoting, as well as restraining, functions were described [13,14]. IL-23 is a key mediator of inflammation, bridges innate and adaptive immune responses, and is known to support tumorigenesis and metastasis [15,16]. IL-12 is central in promoting cell-mediated immunity to infection and cancer [12]. However, this anti-carcinogenic function can be counteracted by IL-12-STAT3-promoted production of pro-carcinogenic IL-23 [17]. While STAT3 is activated by type I and III IFN stimulation in various cell types, its biological functions in the IFN responses are less clear. Growing evidence suggests that STAT3 is a negative regulator of type I IFN activities, thereby providing a pro-viral and pro-survival cellular program [18]; there is, however, also a report on an opposite, i.e., anti-viral activity of STAT3 [19]. The role of TYK2 in IL-10 signaling through STAT3 is not entirely clear and may be cell type- or context-dependent [6]. The double-edged role of IL-10 in immunity and cancer is reviewed elsewhere [9,20]. IL-19, IL-20, IL-24, and IL-26 (absent in mice) constitute a subfamily within the IL-10 cytokine family and signal primarily through activation of STAT3 [9]. Activation of TYK2 at the respective receptors has not been formally shown but can be inferred from the receptor-chain composition. As this subfamily constitutes relatively recently discovered cytokines, cellular responses are still poorly defined, and we refer to recent publications and reviews for a potential cancer

connection [21–23]. Lastly, without specification of the cytokines involved, TYK2 via STAT3 was reported to be crucial for the mediation of cell death in an auto-inflammatory context [24].

STAT5, in contrast, is not among the primarily activated STATs downstream of TYK2 (Figure 1) and occurs dependently on cell type and differentiation stage, in response to type I and III IFNs [25,26], IL-10R2-, and IL-12Rβ1-receptor family cytokines [9,12]. Neither a cytokine-TYK2-STAT5 axis nor its significance have been established under physiological conditions.

#### **2. Aberrant Expression and**/**or Activity of TYK2 in Cancers**

The JAK-STAT pathway is recognized as a core cancer pathway [27] and directly contributes to all hallmarks of cancer [28]. Oncogenic JAK activity can originate from aberrant JAK expression, deregulated upstream signals, GOF mutations, or generation of fusion proteins, as well as loss of negative feedback regulation [2,29–31]. Initially, cancer research focused on JAK1-3, while the TYK2 impact on disease was predominantly studied in inflammatory and (auto-)immune diseases [32,33]. Table 1 summarizes the literature on constitutive or hyperactivated TYK2, as well as GOF-mutated TYK2 and the resulting activation of STATs in cancers.

#### *2.1. Aberrant TYK2 Levels*

In vitro studies with overexpressed JAKs revealed that aberrant TYK2 levels lead to cellular transformation with constitutive phosphorylation of STAT3 [34]. An unusually high expression of TYK2 associated with or causative for carcinogenesis (reviewed [35]) was described for various cancer cell lines and samples from patients suffering from prostate [36,37], ovarian [38], cervical [39], and breast cancer [40,41], as well as malignant peripheral nerve sheath tumors (MPNST) [42,43]. Conflictingly, lowered TYK2 levels in tumor samples and sections (tumor cells and stroma) are generally considered to be an unfavorable prognostic marker (e.g., [44], www.proteinatlas.org). This is supported by a recently published meta-analysis of JAKs and STATs in hepatocellular carcinoma (HCC) patients, where normal or higher TYK2 levels correlated with longer survival and were found in healthy tissue [45]. The underlying reason for these conflicting reports may be attributed to the anti-proliferative/pro-apoptotic and/or tumor surveillance properties of TYK2 [5], as well as the undetermined tumor cell intrinsic and extrinsic state of TYK2. The important role of TYK2 in immune-surveillance is also in line with findings in patients who carry mutated TYK2 alleles which lead to loss of TYK2, lowered TYK2 levels [46], or expression of kinase-inactive TYK2 [47–49], and that show primarily immunodeficiencies. Nonetheless, proteomics suggested that low TYK2 facilitates local metastasis in breast cancer [50], and a comprehensive screen for protein tyrosine kinase variants in numerous cancer cell lines identified splice variants that render TYK2 inactive [51]. On a molecular mechanistic level, the cell intrinsic tumor-promoting consequences of low TYK2 or LOF of TYK2 currently remain elusive.

#### *2.2. Aberrant Activation of TYK2*

A comprehensive list of receptors (over-)expressed in various cancer types which allows us to deduce putative upstream signals involved in hyperactivation of TYK2 was compiled recently [7]. Primary hematological neoplasm (ALCL, anaplastic large cell lymphoma; T-ALL, T cell acute lymphoblastic leukemia) patient samples and cell lines were shown to be dependent on TYK2 activated by upstream IL-10 and/or IL-22 signals and established an upregulation of anti-apoptotic BCL2 family members via STAT1 and/or STAT3 [52,53]. A similar high TYK2-STAT1/3-BCL2 axis was found in MPNST [43]. Cytotoxic T-lymphocyte-associated antigen 4 (CTLA4, CD152) is mainly expressed on T cells and is a well-established immune checkpoint. CTLA4 signaling is initiated through binding to CD80 (B7-1) or CD86 (B7-2) on the surface of antigen-presenting cells (APCs). Ectopic expression of CTLA4 was found on diverse B-cell lymphoma. Mechanistically, it was established that CD86-CTLA4 engagement resulted in recruitment/activation of TYK2, which, in turn, led to a STAT3-driven tumor-promoting transcriptional program [54]. A STAT-independent involvement of

activated TYK2 in fibroblast growth factor 2 (FGF-2) mediated escape from drug-induced death was reported for a sarcoma cell line [55].

#### *2.3. TYK2 Mutations*

Oncogenic JAK2 with the prominent JAK2V617F mutation found in over 50% of myeloproliferative neoplasia (MPN) patients [56] is the paradigm for the understanding of structure/function relations of JAK activity [57–59] and for the general alertness of the cancer field for mutated JAK family members as potential oncogenes. TYK2 joined the club of GOF-mutated JAKs causative for patient hematopoietic malignancies only recently: In 2013, the first TYK2 GOF point mutations were found in T-ALL cell lines and characterized to have transforming capacity via STAT1 and a BCL2 family member [53]. With respect to biochemical studies, the first GOF mutation of TYK2 was V678F, which is the homologous mutation to JAK2V617F [60,61]. Until now, this mutation was not found in patients. The only mutation reported in a public cancer genome database (www.stjude.cloud) for this residue is the V678L mutation, albeit with unknown structure/function consequences. Point mutations at the TYK2 locus are distributed throughout the whole gene body, with GOF mutations—similar to the other JAKs—primarily accumulating in the JH1 and JH2 domains ([2,5] and see public databases, e.g., Genomic Data Commons of the National Cancer Institute [62,63], Catalogue of Somatic Mutations in Cancer (COSMIC [64], and cBioPortal for Cancer Genomics [65,66]).

In addition to the somatic cancer cell mutations, two GOF TYK2 germline mutations (P760L and G761V) were found in pediatric patients developing several de novo leukemias. These mutations are located in the JH2 pseudokinase domain of TYK2 and are predicted to attenuate the negative regulation on the JH1 kinase domain, leading to constitutively activated TYK2 [67].

A prominent germline TYK2 mutation is P1104A/V, which was first found to be associated with solid and hematopoietic cancers [68,69] and later with immunological and inflammatory disorders (reviewed in [5]). While analyzing MPNST tumor samples, it was proposed that TYK2P1104A is an unfavorable prognostic marker for the disease [42]. Notably, this study solely genotyped the somatic cancer cells and overlooked that this mutation impairs TYK2 catalytic activity; cellular signaling, however, is not completely abrogated, and the detected induction of BCL2 expression might favor an anti-apoptotic program [69,70]. Recent studies show that TYK2P1104A is a LOF mutation, because patients homozygous for this allele are either susceptible to microbial infection or protected from autoimmune disease [47,49,71]. These mechanistic and phenotypic features of TYK2P1104A were confirmed in independent mouse models [48,71].

#### *2.4. TYK2 Fusion Proteins*

Chromosomal rearrangements account for a number of driver kinase fusion genes in cancer [72–74]. The first fusion kinase involving a JAK was TEL-JAK2, consisting of a 3 portion of JAK2 and a 5 region of TEL, a member of the ETS transcription factor family [75]. This chromosomal translocation is found in T-ALL in patients [75] and transgenic mice expressing TEL-JAK2 develop T-cell leukemia [76]. In vitro studies with a TEL-TYK2 fusion showed constitutive activation of STAT1/3/5 and transforming capacities [77], albeit respective translocations have not yet been identified in patients. As observed for GOF-mutated JAKs, JAK2 kinase fusions occur most frequently compared to the other JAKs, which suggests that the JAK2 locus is a mutation and rearrangement hotspot [56,78,79]. The first leukemia patients carrying TYK2 fusion genes described were combinations of the TYK2 kinase domain and a part of the pseudokinase domain with 5 portions of nucleophosmin (NPM) 1, polyadenylate binding protein (PABPC) 4, or the transcription factors MYB or NFκB2 [80–82]. Structurally and mechanistically, the TYK2 fusion proteins lack the negatively regulating function of the pseudokinase (JH2) domain leading to a GOF kinase activity and hyperactivity of STAT3 and depending on the cellularity also STAT1 and 5 (reviewed in [5,58,59]).

Subsequent analysis of patient samples and cell lines [83–87] and screening of cancer data sets revealed more than 50 chromosomal *TYK2* rearrangements found mostly in hematological, but also in solid cancers [88]. For the fusions, it is currently not known if they contribute as driver oncogenes to early tumorigenesis or are rather the result of genomic instability at later tumor stages [89]. Recently, chromothripsis was identified as a new type of chromosomal rearrangement during carcinogenesis. Based on a single chromosome-shattering event and DNA repair complex, intra- and interchromosomal rearrangements, such as fusion genes, are produced within a few cell cycles. If the fusion event(s) allow for growth or survival advantages, a cancer driver gene might be generated [90,91]. Chromothripsis was assigned to genomic alterations in childhood cancer [92], and mechanistically it is caused by defects in the nuclear envelope composition or formation and failures during mitosis [93]. It is tempting to speculate that the remarkably high number of described TYK2 fusions were—at least in part—generated through chromothripsis and thus might act as driver mutations.

#### **3. Tumor-Promoting Activities of (Hyper-)Active TYK2**

The molecular contribution of TYK2 signaling and known protein–protein interactions to the hallmarks of cancer were reviewed previously [5,28]. Here, we highlight the latest findings on the consequences of TYK2 hyperactivity in cancer cells.

#### *3.1. TYK2 Activation of (Oncogenic) STAT Signaling*

As shown in Figure 1, the heterodimeric cytokine receptors with engagement of TYK2 are capable of activating all STATs. Hyperactive, GOF-mutated TYK2 or TYK2 fusions in oncogenic settings preferentially lead to aberrant activation of STAT1, STAT3, and STAT5. The oncogenic potential of STAT3 and STAT5 was recognized early on and is well documented [94,95]. STAT1 was initially considered to exert tumor suppressor functions, and its oncogenic potential emerged more recently [96–98].

STAT1/3/5 were found hyperactivated in patient-tailored cell lines with activated TYK2 [53], as well as carrying somatic or germline TYK2 GOF mutations [53,67] or TYK2-NPM1 and -NFkB2 fusions [80,82]. In other tumor samples or experimental tissue culture settings, STAT3 only, or other dual combinations of activated STAT1/3/5, are described (see Table 1).

Interestingly, TYK2 does not only phosphorylate the major phosphorylation site Y705 in STAT3, but also Y640, which represses STAT3 activation [99]. This phosphorylation site in STAT3 is often mutated in cancers [100,101]. Neither the general (patho-)physiological impact nor the contribution to malignancies of this phosphorylation event are currently known.


**Table 1.** (Hyper-)active TYK2, GOF-, or LOF-mutated TYK2 and STAT activation in various cancers and cancer cell lines.


**Table 1.** *Cont*.

— Unrelated to disease, in vitro findings in stable cell lines, (1) found in patient samples and primary material, (2) in vitro findings endogenous TYK2 expression, (2 \*) in vitro findings exogenous TYK2 expression, (2 \*\*) in vitro findings exogenous claudin expression, (3) phosphorylated mutated TYK2 protein, (3 \*) phosphorylated wildtype TYK2 protein, (4) high levels of wildtype TYK2, (4 \*) high levels of mutated TYK2, (5) reduced levels of phosphorylated mutated TYK2, (6) reduced levels of wildtype TYK2, and (?) not specified if wildtype or mutated TYK2. Please note that some references did not study the activation of all STATs and that not all described STATs in the table are active in all cell systems used. Ph = Philadelphia, and n.d. not determined.

#### *3.2. TYK2 Stimulation of Tumor Cell Invasion*

The families of tight junction proteins claudins (CLDNs) and of matrix metalloproteinases (MMPs) are central for the invasion of tumor cells and, in consequence, metastasis formation [106,107]. Recent studies show that, in liver and lung carcinoma, high levels of CLDN9/12/17 caused activation of TYK2 and STAT1/3 and promoted metastasis [102,104,105]. The promoters of various MMP genes harbor STAT binding sites, and many MMPs are transcriptionally activated through TYK2-associated cytokine receptors [108,109]. Gene-targeted mice revealed that TYK2 and STAT1 are required for expression of MMP2/9/14 under inflammatory conditions [110]. Biochemical studies showed that, dependent on context and inflammatory conditions, MMP1/3 induction involves STAT1 alone [108] or also STAT3 [111]. In a hematopoietic tumor TYK2-STAT3 induced MMP9 and tumor cell invasiveness [54] and in a solid tumor TYK2-STAT3 signaling induced MMP1 expression [103].

The urokinase-type plasminogen activator (uPA)/receptor (uPAR) system is central for a cascade of proteolytic events, including activation of MMPs, which allow for tumor cell migration and metastasis [112]. Signaling via uPAR involves TYK2 and PI3K [113], and, at the post-transcriptional level, TYK2 inhibits the accumulation of plasminogen activator inhibitor (PAI) 2 [114]. In prostate cancer, high levels of TYK2 correlate with invasion and metastasis [36,37]. In an ovarian cancer cell line pY-STAT3 co-localizes with TYK2 and JAK2 at focal adhesions, and hyperactive STAT3 was shown to promote cancer cell motility [38]. Without providing molecular details, a mouse model for aggressive lymphoma showed reduced tumor cell invasiveness upon loss of TYK2 [115]. In addition, without providing molecular insights, a siRNA screen assessing the role of the tyrosine kinome in metastasis formation identified TYK2 as a promoter of invadopodia, which are cellular structures

characteristic for tumor cell migration [116,117]. Connexin43 (Cx43) is the most widely expressed member of a large family of transmembrane proteins involved in gap junction formation. Cx43 can be both pro- and anti-tumorigenic, e.g., by promoting invasion and metastasis and by acting as a tumor suppressor [118,119]. TYK2 was found to play a dual role in regulation of Cx43: On the one hand, TYK2 is capable of directly phosphorylating Cx43, thereby decreasing its stability; on the other hand, angiotensin II-activated TYK2 increased Cx43 levels in a STAT3-dependent manner [120]. This regulatory loop has not yet been studied in the context of carcinogenesis. Furthermore, knockdown of TYK2 reduced migration of breast cancer cell lines [50].

#### *3.3. TYK2 Prevention of Apoptosis*

IFNs in general are capable of promoting apoptosis of cancer cells [121]; hence, provided that IFN stimulus and responsiveness in the tumor is given, TYK2 acts tumor suppressive. Tumor cells are able to resist cell death by upregulation of anti-apoptotic BCL-2 family members [122,123]. TYK2 was shown to drive either in a STAT1- and/or a STAT3-dependent manner or in a STAT-independent but ERK1/2-dependent manner high expression of BCL-2 [43,53,55] or its family members BCL-2L1 [54] and MCL1 [52,55]. In contrast, an in vitro study demonstrated that TYK2 physically interacts with SIVA-1 and promotes SIVA-1 mediated apoptosis, as well as inhibits BCL-2 [124].

#### *3.4. TYK2 Crosstalk to Oncogenes and Proto-Oncogenic Pathways*

In a mouse model of ALCL, as well as in patient cells, TYK2 showed co-operativity with the oncogenic fusion kinase NPM-ALK [52]. In contrast, no co-operation of TYK2 with mutated FLT3-ITD or JAK2V617F in MPN mouse models was found [125,126]. The latter is consistent with the observation that, in JAK2V617F MPN patients (see below) resistant to pharmacological JAK2 inhibition, only JAK1, and not TYK2, leads to heterodimeric STAT activation, despite both kinases show equal tyrosine phosphorylation at the activating loop [127]. This is to be expected, since, in contrast to the other JAKs, loss of TYK2 at heterodimeric JAK-associated cytokine receptors leads only to a partial impairment in signaling [5,6], and, as experimentally described for the IFNAR receptor, TYK2 is the subordinated JAK at cytokine receptors [128,129].

Early biochemical studies suggest that, upon type I IFN treatment, TYK2 interacts with various proto-oncogenes, including the guanine nucleotide exchange factor 1 VAV, the E3 ubiquitin-protein ligase C-CBL, and the SRC family tyrosine kinases FYN and LYN [130–134]. The importance of these PPIs for tumorigenesis is currently unknown. In cancer samples or cell lines, TYK2 was found to cooperate with other oncogenic effectors and pathways, such as the RAF/ERK [53,55,61], MAPKs [135], PIM1/2 [84], and PI3K/AKT/mTOR pathway [36,53,61]. Reported solely in the context of skin inflammation is the TYK2-STAT3 requirement for expression of IκBζ (encoded by *NFKBIZ*) [136]; however, emerging reports suggest cell-intrinsic oncogenic, as well as tumor-suppressive, functions of IκBζ [137].

The mapped and predicted PPIs of TYK2 based on proteomics [138,139] and next generation sequencing (NGS) are accessible at various open-source databases (for a review, see [140]). The TYK2 kinase domain and a STAT3-based reporter system were used to establish the first mammalian two hybrid kinase substrate sensor (KISS) screening platform [141,142]. These databases and the screening approaches should be systematically exploited to further define and fine tune the TYK2 interactome in health and disease.

#### **4. Deactivation and Stabilization of TYK2 under (Patho-)Physiological Conditions**

JAK activity is counter regulated by molecule-intrinsic events, such as post-translational modifications (PTMs) and the inhibitory function of the pseudokinase domain [143] as well as by extrinsic inhibitory regulators, such as suppressor of cytokine signaling (SOCS) proteins and protein tyrosine phosphatases (PTPs) [144].

Databases [145,146] provide curated PTMs, but with the exception of the well described activating phosphotyrosines, there is still a lack of information on the properties of JAKs that are defined by PTMs. For TYK2, ubiquitination and phosphorylation are detected at multiple residues and discussed in the context of stability/decay (PhosphoSitePlus®, [146]), albeit the (patho-) physiological relevance is unknown.

SOCS proteins are encoded by STAT target genes and are negative feedback inhibitors of JAK signaling. SOCS1 and 3 are the most potent JAK inhibitors because, in addition to recruitment of JAKs to E3 ubiquitination/degradation mediated by all SOCS family members, they also harbor a kinase inhibitory region (KIR), which efficiently shuts down JAK activity by binding to the JH1 domain [147]. Activated JAKs and cytokine receptor chains are dephosphorylated by multiple PTPs [148]. The current literature regarding deactivation of TYK2 by SOCS1/3, the PTPs PTB1B and SHP1, as well as the global impact of SOCS and PTP family members in cancer are reviewed elsewhere [5,149–151].

In vitro studies showed that in hematopoietic tumor cells the PTP SHP1 suppresses growth via accelerating the TYK2 protein degradation [152]. In lung cancer cells, overexpression of the E3 ubiquitin ligase seven-in-absentia-2 (SIAH2) accelerates the proteasomal degradation of TYK2, thereby attenuating STAT3 signaling [103].

HSP90 is a chaperone supporting folding, stability, and function of many client proteins, including JAKs and STATs [153–155]. Cancer cells frequently use HSP90 to stabilize and/or increase the function of numerous oncogenes, and HSP90 inhibitors have been studied as anticancer drugs for more than two decades [156,157]. Physical interaction of HSP90 with TYK2 was demonstrated in cancer cell lines and confirmed in a proteome-wide assessment of the HSP90 interactome [158,159]. HSP90 inhibitor treatments in various tumor settings showed beneficial effects by reducing the activity of TYK2 or its fusion proteins [158,160,161].

An emerging field is the involvement of noncoding RNAs in the regulation of the JAK-STAT pathway in carcinogenesis [162–164]. Recently, the long noncoding RNA (lncRNA) MEG3 in concert with a microRNA (miR-147) was reported to modulate JAK-STAT signaling in chronic myeloid leukemia (CML). Interestingly, the lncRNA was found to physically interact with TYK2, JAK2, and STAT3, thereby diminishing the activity level of STAT3 (and STAT5) [165].

#### **5. Pharmaceutical TYK2 Inhibition**

The first selective JAK inhibitor (JAKinib) to be tested in humans was tofacitinib, which potently inhibits JAK3 and JAK1, and, to a lesser extent JAK2, and has little effect on TYK2 [166]. Historically, JAKinibs were developed as immunosuppressive drugs for the clinical use in organ transplants and autoimmune diseases [167]. The success story of ruxolitinib, a JAK2 and JAK1 inhibitor which was the first JAKinib approved for treatment of a hematopoietic malignancy, pushed the perception of JAKinibs as anticancer drugs [168,169]. For insight in development and clinical use, as well as side effects of JAKinibs, we refer to the most recent reviews [170–173].

TYK2inibs are mainly envisaged as therapeutics for treatment of autoimmune and inflammatory diseases [33,174], in which JAKinib selectivity is currently considered not to be of utmost importance [175]. As for the other JAKinibs, the first generation TYK2inibs are directed to the JH1 domain and compete with ATP in binding to the enzymatic pocket. These inhibitors are potent in inhibiting wildtype (overexpressed) TYK2, mutated (hyperactive) TYK2, and TYK2 fusion proteins harboring the JH1 domain. Since the JAKs show high homology in the JH1 domain, it is hard to develop ATP-competing inhibitors with high selectivity for one particular JAK family member [170,172]. A next-generation inhibitor of TYK2 is directed against the JH2 domain and recently passed the phase II clinical trial for psoriasis treatment [176]. A comprehensive report on the high selectivity and the biological effects of this TYK2inib in mouse models, as well as its efficacy in human cells collected from autoimmune patients, was recently published [177]. JH2-specific TYK2inibs are currently further improved, and additional compounds are being developed [178–181]. The only TYK2inib reported and successfully tested to block TYK2 activity in an oncogenic setting is a JH1-specific TYK2inib [135]. Notably, JH2 domain inhibitors might not be working for treatment of diseases driven by TYK2 fusion genes missing parts of the JH2 domain.

#### **6. Conclusions and Future Perspectives**

Since the discovery of TYK2 and the JAK-STAT signaling paradigm in the early 1990s, enormous progress has been made in the structural and functional understanding of the linear JAK-STAT axis and the crosstalk of JAKs or STATs to other signaling hubs, as well as the cell type-specific contributions of JAKs and STATs in health and disease. The striking phenotypical similarities between mouse models deficient for TYK2 or engineered to express kinase-inactive TYK2 and human patients carrying the respective germline mutations established TYK2 as a fundamental component in both innate and adaptive immunity. The (patho-)physiological and molecular pathway similarities of TYK2 in human and mice allow for highly informative comparative biomedical studies and efficient translation of basic molecular insights into clinical applications. The use of TYK2inibs in the treatment of immunological and inflammatory diseases is within reach [182] and is also attractive for malignancies with the involvement of hyperactivated TYK2. The role of TYK2 and GOF-mutated TYK2 upstream of oncogenic STAT3—and, less frequently, STAT1—is established, while, up to now, no mechanistic evidence for an oncogenic TYK2-STAT5 axis is given. Mouse models as genetic mimics of kinase-inhibited TYK2 exist [48,71,183,184] and are currently exploited to further dissect the kinase-dependent from the scaffolding functions of TYK2.

In a short-term perspective, work should concentrate on the use of refined TYK2 mouse models that allow studying the kinase-independent and cell type-specific functions, in order to fully in vivo assess TYK2inibs with respect to their benefits and unwanted side effects. Mouse models to study the consequences of aberrant high TYK2 and GOF-mutated TYK2 are underway (K. Wöss, T. Rülicke et al., unpublished). For pharmacological intervention with oncogenic TYK2, TYK2inibs with the highest possible selectivity are required, and efforts should focus on the further development and in vivo testing of these next-generation TYK2inibs.

In a long-term perspective, the further understanding of the TYK2 function requires the in-depth elucidation of the PTMs and the interactome of TYK2 under spatiotemporal conditions. Additionally, computational modelling and structure predictions (e.g., [185]) should complement the attempts to determine the holo-crystal structure of TYK2 and to use high-resolution imaging (e.g., [186]) to gain insight into the structural features of full-length wildtype and mutated TYK2, as well as its conformation bound to various cytokine receptors.

**Author Contributions:** K.W., S.M.-M., and M.M. designed the draft of the review and performed the literature search. K.W. and S.M.-M. compiled the table, and N.S. designed the figure. S.M.-M., M.M., N.S., and B.S. provided the final version of the manuscript.

**Funding:** This work was funded by the Austrian Science Fund FWF DK W1212, SFB F6101 and F6106, and DocFund DOC32-B28. We are thankful to Tanja Bulat for critically reading the manuscript.

**Conflicts of Interest:** The authors declare no conflicts of interest.

#### **References**


© 2019 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

#### *Article*
