Intermediate filament protein keratin 8 (K8) binds to heat shock protein 70 (Hsp70) and p38 MAPK, and is phosphorylated at Ser74 by p38α (MAPK14, hereafter p38). However, a p38 binding site on K8 and the molecular mechanism of K8-p38 interaction related to Hsp70 are unknown. Here, we identify a p38 docking site on K8 (Arg148/149 and Leu159/161) that is highly conserved in other intermediate filaments. A docking-deficient K8 mutation caused increased p38-Hsp70 interaction and enhanced p38 nuclear localization, indicating that the p38 dissociated from mutant K8 makes a complex with Hsp70, which is known as a potential chaperone for p38 nuclear translocation. Comparison of p38 MAPK binding with keratin variants associated with liver disease showed that the K18 I150V variant dramatically reduced binding with p38, which is similar to the effect of the p38 docking-deficient mutation on K8. Because the p38 docking site on K8 (Arg148/149 and Leu159/161) and the K18 Ile150 residue are closely localized in the parallel K8/K18 heterodimer, the K18 I150V mutation might interfere with K8-p38 interaction. These findings show that keratins, functioning as cytoplasmic anchors for p38, modulate p38 nuclear localization and thereby might affect a number of p38-mediated signal transduction pathways.
Keratins (Ks) are the primary intermediate filament proteins present in epithelial cells (Eriksson et al., 2009; Omary et al., 2009). They are subdivided into acidic type I (K9-K28; K31-K40) and basic type II (K1-K8; K71-K86) keratins, which form obligatory type I/type II keratin heteropolymers (Jacob et al., 2018; Schweizer et al., 2006). Keratins are expressed in both tissue-specific and cell type-specific manners. For example, simple epithelia such as liver express K8 and K18, whereas stratified epithelia such as epidermis express K5 and K14 (Gonzales and Fuchs, 2017; Omary et al., 2009). Keratins play a particularly important role in providing mechanical support for cells and enhancing their structural integrity (Coulombe and Omary, 2002; Kumar et al., 2015). In addition, they also play an important role in nonmechanical functions such as intracellular cell signaling by modulating cell growth, apoptosis and cancer progression (Pan et al., 2013).
Keratins are modulated by post-translational modifications such as phosphorylation, which is known to regulate the assembly and disassembly of keratin filaments as well as keratin solubility and subcellular localization (Loschke et al., 2015; Omary et al., 2006; Snider and Omary, 2014). Mitogen-activated protein kinase (MAPK) pathways are significant signal transduction modules that regulate cell proliferation, differentiation and apoptosis (Kyriakis and Avruch, 2012; Nishida and Gotoh, 1993). MAPKs are activated by phosphorylation of their conserved Thr-X-Tyr dual phosphorylation motif in response to a wide variety of cellular stressors (Raingeaud et al., 1995). Previous studies have shown that the K8 Ser74 and Ser432 sites, matching a pSer/Thr-Pro consensus motif for the MAPK target sequence, are phosphorylated by proline-directed MAPKs such as extracellular signal-regulated kinases (ERKs) and c-Jun N-terminal kinases (JNKs) (He et al., 2002; Ku et al., 2002a). In addition, p38α (MAPK14, hereafter p38) associates with K8/K18 and phosphorylates K8 Ser74, thereby significantly affecting keratin filament reorganization (Ku et al., 2002a). Nevertheless, the interaction regions in K8 that are required for specific and efficient phosphorylation by p38 have not been reported and the functions of these interactions are not fully understood.
MAPKs bind to their substrates via different sites. The interaction between the catalytic cleft of the kinase and the phosphorylation site of the substrate is a key contributor to kinase-substrate recognition (Ubersax and Ferrell, 2007). Moreover, the additional docking motif increases the overall specificity and efficiency of kinase-substrate interaction (Ubersax and Ferrell, 2007). Two distinct docking domains, D domains and DEF domains (docking sites for ERK, FXFP), are well established in MAPK substrates (Cargnello and Roux, 2011; Ubersax and Ferrell, 2007). The D domains are characterized by one or two basic residues, a short spacer and a hydrophobic groove (R/K)1-2-X2-6-φ-X-φ, where φ is a hydrophobic residue (Cargnello and Roux, 2011; Tanoue and Nishida, 2003). The DEF domain, which consists of a Phe/Tyr-X-Phe/Tyr-Pro motif, has been identified in ERK and other MAPK substrates (Fantz et al., 2001; Kyriakis and Avruch, 2012). The basic and hydrophobic residues of the D domain in MAPK substrates bind to the acidic and hydrophobic area of MAPKs called the common docking (CD) domain (Tanoue et al., 2000). Multiple domains on MAPKs and their substrates cooperatively contribute to enhancing the specificity and efficiency of signal transduction (Kyriakis and Avruch, 2012).
The translocation of MAPKs into the nucleus is important for physiological function in response to extracellular stimulation. The p38 MAPKs reside in both the cytosol and the nucleus, and phosphorylated p38 translocates into the nucleus in response to environmental stress (Wood et al., 2009). Although the p38 kinase does not contain the canonical nuclear localization signal (NLS), the kinase translocates into the nucleus by interaction with importin-7 and importin-9 (Maik-Rachline et al., 2018) or through binding with NLS-containing heat shock protein 70 (Hsp70) (Gong et al., 2012). Notably, Hsp70 is known to shuttle between the cytoplasm and nucleus, functioning as a chaperone responsible for the translocation of client proteins such as importin α/β, Smad3 and p38 (Kose et al., 2005; Snider and Omary, 2014; Zhou et al., 2010). Interestingly, under stress, K8-Hsp70 interaction is increased (Liao et al., 1995) whereas K8-p38 interaction is reduced (Lee et al., 2013). Although Hsp70 is a potential chaperone for p38 nuclear translocation in response to stress (Gong et al., 2012), the underlying mechanism of the nucleocytoplasmic translocation of p38 related to K8 and Hsp70 remains largely unknown.
In this study, we identified a docking site (148RRQLETLGQEKLKL161) in K8 that plays a crucial role in its binding to p38. Mutations in these residues effectively reduced the ability of K8 to associate with p38 and, at the same time, increased p38-Hsp70 binding, which led to p38 nuclear localization. Furthermore, the liver disease-associated keratin variant K18 I150V significantly decreased its association with p38 while enhancing p38 nuclear localization. Although K18 has neither a p38-specific docking site nor a site phosphorylated by p38, the K18 I150V mutation interferes with K8-p38 interaction, probably because of the spatial proximity between p38 docking sites on K8 (Arg148/149 and Leu159/161) and the K18 Ile150 residue in the parallel K8/K18 heterodimer. Moreover, the expression of p38-dependent target genes such as inducible nitric oxide synthase (iNOS), vascular cell adhesion molecule (VCAM) and vascular endothelial growth factor C (VEGFC) tends to be more increased in cells transfected with the K18 I150V mutant than in controls. Taken together, these results indicate that the interaction of K8 and p38 is crucial in determining the efficiency of p38-Hsp70 complex formation, which affects p38 nuclear localization and thereby p38-dependent signal transduction pathways.
Interaction of K8/K18 with p38 kinase is phosphorylation dependent
A previous study has shown that p38 kinase directly binds to K8 and phosphorylates K8 at Ser74 (Ku et al., 2002a). Based on an overlay assay with purified p38 kinase, the study clearly demonstrates that p38 specifically interacts with K8 but not K18 (Ku et al., 2002a). The binding of p38 and K8 is reduced by treatment with okadaic acid (OA), known as a protein phosphatase inhibitor, which implies that their interaction is inhibited by phosphorylation (Lee et al., 2013). To address whether phosphorylation of the K8 Ser74 residue affects the interaction between K8 and p38 kinase, BHK cells were transfected with wild-type (WT) K8 or K8 mutants containing substitutions of Ser74 with alanine (S74A) or aspartic acid (S74D) to block or mimic phosphorylation, respectively. Immunofluorescence staining showed that the transfection efficiencies were approximately similar in cells transfected with different plasmids (Fig. S1). Co-immunoprecipitation experiments showed that K8 S74A enhanced the interaction between K8 and p38, whereas K8 S74D greatly reduced this interaction, suggesting that K8 Ser74 phosphorylation interferes with K8-p38 binding (Fig. 1A). The increased interaction between p38 and K8 S74A as compared with K8 WT was observed under treatment of OA (a protein phosphatase inhibitor) (Fig. S2A), whereas p38-K8 WT interaction was almost equivalent to p38-K8 S74A interaction under treatment of protein phosphatase 2 (Fig. S2B). This indicates that p38 binds more favorably to dephosphorylated K8/K18. Taken together, these results show that the interaction between K8/K18 and p38 is phosphorylation dependent.
Dual phosphorylation at the tripeptide motif Thr-X-Tyr located in the activation loop of p38 kinase is crucial for p38 to perform its physiological functions (Raingeaud et al., 1995). To explore whether the dual phosphorylation in p38 has any effect on K8-p38 association, we compared K8 binding with p38 WT and p38 AF, a kinase-dead dominant negative form of p38 that is mutated at the dual phosphorylation sites. Our results showed that, compared with p38 WT, p38 AF enhanced the interaction with K8/K18 (Fig. 1B), suggesting that p38 phosphorylation inhibits p38 interaction with K8/K18. This result (Fig. 1B) agrees with a previous study demonstrating p38 kinase release from K8/K18 under stress, such as treatment with anisomycin, which causes the phosphorylation/activation of p38 (Lee et al., 2013). Indeed, Raf kinase is also released from K8/K18 upon Raf hyperphosphorylation/activation during oxidative stress (Ku et al., 2004). Taken together, our findings indicate that K8/K18 binds and sequesters p38 under basal conditions, whereas stress-induced phosphorylation/activation of p38 causes the dissociation of p38 from K8/K18, which allows the released p38 to be involved in signal transduction under stress conditions.
K8 has a docking site for interaction with p38 kinase
In addition to the pSer/Thr-Pro consensus motif on the MAPK substrate, MAPKs also interact with other additional docking domains on their substrates, including the D and DEF domains, to increase the efficiency and specificity of signal transduction (Cargnello and Roux, 2011). The D domain is well characterized by one or two positively charged amino acids surrounded by hydrophobic residues, (R/K)1-2-X2-6-φ-X-φ, where φ is a hydrophobic residue (Tanoue and Nishida, 2003). In our study, we found six clusters of positively charged amino acids as potential D domains for p38 in K8 and seven clusters in K18 (Table 1). However, neither K8 nor K18 contains a potential DEF domain consisting of a Phe/Tyr-X-Phe/Tyr-Pro motif.
To test the potential D domains in K8/K18, we constructed keratin mutants in which the positively charged amino acids arginine and lysine (R, K) were replaced by negatively charged glutamic acid (E) (Table 1) and compared the efficiency of binding to p38 kinase. Each keratin mutant was analyzed for the effect of mutation on p38 binding and Ser74 phosphorylation. The interaction between p38 and K8 did not completely correlate with the phosphorylation status of K8 Ser74 (Fig. 2A). For example, keratin mutations such as K185, R186E or R301/302E, K304E positively correlated with Ser74 phosphorylation but not with p38 binding. This suggests the involvement of other kinases in Ser74 phosphorylation. Interestingly, K8 R148/149E mutation showed a dramatic effect on both Ser74 phosphorylation and p38 binding (Fig. 2A). The co-immunoprecipitation assay showed that K8 R148/149E mutation markedly decreased keratin-p38 interaction, whereas other mutations had minimal effects (Fig. 2A,B). Hence, we focused on the R148/149E mutant. Arg148/149 residues lie downstream of the phosphoacceptor site (Ser74) in K8. Notably, the disrupted interaction between K8 R148/149E and p38 resulted in a reduction in K8 Ser74 phosphorylation (Fig. 2A), demonstrating that the positively charged amino acids in the docking site contribute to binding and phosphorylation efficiency.
The D domain is well characterized by one or two positively charged amino acids surrounded by hydrophobic residues, (R/K)1-2-X2-6-φ-X-φ, where φ is a hydrophobic residue (Tanoue and Nishida, 2003); therefore, to help identify the p38 docking site on K8, the leucine (L)159/161 hydrophobic amino acids near to K8 Arg148/149 residues were mutated to alanine (A). Mutations of the positively charged and/or hydrophobic residues in K8 148RRQLETLGQEKLKL161 caused a large reduction in its interaction with p38 and diminished K8 Ser74 phosphorylation (Fig. 2C). Notably, quadruple mutation (R148/149E, L159/161A) reduced K8 Ser74 phosphorylation more dramatically (Fig. 2C), indicating that the docking interactions achieved using both two basic amino acids and two hydrophobic amino acids are necessary for efficient phosphorylation of K8 Ser74. The p38 docking motif was localized in a rod domain of K8 and the motif was highly conserved in other keratins and intermediate filament proteins (Table 2).
The newly identified D domain of K8 is particularly important for interaction with p38, but not with ERKs and JNKs (Fig. S3), although ERKs and JNKs phosphorylate K8 Ser74 (He et al., 2002; Ku et al., 2002a). In addition, we examined the impact of the quadruple mutation on keratin filament organization and keratin solubility. The immunostained filament patterns of K8 WT and the R148/149E, L159/161A mutant were indistinguishable (Fig. S4A). Serial cell fractionation demonstrated that the solubility of the quadruple mutant was similar to the solubility of K8 WT (Fig. S4B). The mutation had no effect on either filament organization or keratin solubility.
The CD domain in p38 kinase interacts with the docking site in K8
Several regions have been reported in the MAPK family that are involved in docking interactions with their substrates (Rubinfeld et al., 1999; Tanoue et al., 2000). The negatively charged residues of MAPKs establish crucial electrostatic interactions with positively charged residues of the docking domains of their substrates (Tanoue et al., 2000). The hydrophobic residues that have been identified in MAPKs play a role in establishing hydrophobic interactions with the hydrophobic residues of the substrate D domains (Enslen and Davis, 2001). These CD domains are located on the opposite side of the catalytic region of MAPKs (Cargnello and Roux, 2011; Tanoue et al., 2001). Mutational analyses with p38 have identified the negatively charged amino acids Asp313, Asp315 and Asp316 in p38 as acting as the conserved docking motif (Tanoue et al., 2000).
To examine the role of the CD domain in the p38-K8 association, we performed a co-immunoprecipitation assay by co-transfection of K8/K18 WT with p38 WT or a mutant (CDm) in which Asp313, Asp315 and Asp316 were replaced by asparagine. As shown in Fig. 3A, the interaction of p38 CDm with K8/K18 was significantly decreased compared with p38 WT, indicating the significance of these three acidic amino acids of p38 in docking interactions with K8. We observed that the level of K8 Ser74 phosphorylation was similar in cells transfected with either p38 WT or CDm, which is probably the result of K8 Ser74 phosphorylation by other MAPKs. Together, the negatively charged amino acids in the CD domain of p38 and the positively charged/hydrophobic amino acids in the docking site of K8 contribute to regulation of their interaction (Fig. 3B). The Tyr311 and His312 residues of p38 are identified as a potential docking site that interacts with the hydrophobic residues on p38 substrates, and its position is adjacent to the negatively charged residues in the CD domain (Xu et al., 2001). These two residues of p38 probably interact with the hydrophobic residues (Leu159 and Leu161) of K8 in the D domain (Fig. 3B).
K8 modulates formation of p38-Hsp70 complexes and Hsp70-mediated nuclear translocation of p38
Because p38 and Hsp70 are known to associate with K8 (Ku et al., 2002a; Liao et al., 1995) and Hsp70 is a potential chaperone for p38 nuclear translocation under stress (Gong et al., 2012), we examined whether K8 could affect the formation of p38-Hsp70 complexes and thereby be involved in p38 nuclear translocation. Co-immunoprecipitation experiments revealed that overexpression of K8, but not K18, reduced p38-Hsp70 association (Fig. 4A), whereas overexpression of the p38 docking-deficient K8 mutant caused an increase in their interaction (Fig. 4B), suggesting that K8 regulates p38-Hsp70 association by sequestering p38 from Hsp70. The p38 MAPK is reported to shuttle between the nucleus and the cytoplasm, and phosphorylated p38 kinase translocates into the nucleus in response to stress conditions (Brand et al., 2002; Wood et al., 2009). Interestingly, a previous study has shown that Hsp70 interacts with p38 and serves as a potential chaperone for nuclear translocation of p38 (Gong et al., 2012). Given that activated p38 is released from K8/K18 under stress conditions (Lee et al., 2013) and that a kinase-inactive mutant of p38 (p38 AF), with mutated dual phosphorylation sites T180A and Y182F in the activation loop (Han et al., 1994), exhibits increased interaction with K8/K18 (Fig. 1B), we hypothesize that K8/K18 interferes with p38 nuclear translocation by sequestering p38 from Hsp70 under basal conditions, and that phosphorylated/activated p38 kinase is released from K8/K18 in response to stress and forms a p38-Hsp70 complex that translocates into the nucleus.
To test whether K8 influences the nuclear localization of p38 by modulating p38-Hsp70 complex formation, the localization of p38 in both cytoplasm and nucleus was examined using biochemical subcellular fractionation analysis. BHK cells were co-transfected with either K8 WT or a docking-deficient mutant along with p38 and Hsp70 and then treated with the phosphatase inhibitor OA, which enabled detection of the phosphorylated/activated p38. After OA treatment, the increased amount of p38 in the nucleus and the increased levels of phosphorylated p38 in total lysates were observed independently of transfected K8 constructs. However, the phosphorylated p38 was more prominently accumulated in the nucleus of cells transfected with the p38 docking-deficient K8 mutant compared with K8 WT (Fig. 4C). These data clearly show that the disrupted docking interaction between K8 and p38 promotes the nuclear distribution of phosphorylated p38. However, we observed no significant difference in Hsp70 distribution in cells transfected with K8 WT or the p38 docking-deficient K8 mutant (Fig. 4C). It is likely that there was a technical difficulty in detecting subtle differences in Hsp70 distribution in the subcellular fractionated lysates because Hsp70 is a relatively abundant protein in cells and binds to many different proteins for protein folding. The purity of cell fractionation was assessed and the results are shown in Fig. S5. We compared the p38 nuclear distribution in HepG2 cells treated with K8-specific siRNA or a negative control sequence and observed a reduction in K8 expression in cells treated with K8-specific siRNA (Fig. S6A). However, in an immunostaining assay under tested conditions, we detected no significant difference in p38 nuclear localization in treated and untreated cells (Fig. S6B).
K18 I150V interferes with p38-K8 interaction, which leads to enhanced p38-Hsp70 formation, resulting in nuclear translocation of p38
Mutations in the K8/K18 genes have been shown to be related to liver disease (Ku et al., 2016; Omary, 2017; Omary et al., 2004; Toivola et al., 2015). The mechanisms by which the identified keratin mutations cause a predisposition to liver damage are not fully understood. Based on previously described K8/K18 mutations associated with liver disease (Ku et al., 2005), we examined whether these K8/K18 mutations have an effect on K8/K18 interaction with p38. Co-transfection and co-immunoprecipitation assays showed that the K18 I150V variant, as compared with the other tested variants, markedly inhibited the interaction between p38 and K8/K18 (Fig. 5A,B). Similar to the effect of the p38 docking-deficient K8 mutant, the co-transfected K8 WT was heteromeric with the K18 I150V variant, which resulted in less efficient phosphorylation at K8 Ser74 (Fig. 5A). Interestingly, inhibition of K8 Ser74 phosphorylation in the K8 G62C mutant (Fig. 5B) was observed as described in a previous study (Ku and Omary, 2006), whereas the level of K8 Ser74 phosphorylation in other K8 mutants showed no statistically significant change (Fig. 5B).
We then investigated whether K18 I150V has an effect on regulating the subcellular localization of p38 by increasing the physical interaction between p38 and Hsp70. Co-immunoprecipitation analysis showed that the K18 I150V mutation caused a reduced association of K8/K18 with p38 kinase along with increased p38-Hsp70 interaction (Fig. 6A). Furthermore, cell fraction analysis demonstrated that K18 I150V mutation enhanced the nuclear localization of p38 kinase (Fig. 6B). The purity of cell fractionation was assessed and the results are shown in Fig. S7. We then examine the expression of eight known p38-dependent target genes, including the inflammatory genes, using real-time quantitative PCR (RT-qPCR). Expression of target genes such as those encoding iNOS, VCAM and VEGFC was inclined to be more increased in cells transfected with the K18 I150V mutant than with K18 WT (Fig. 6C). These findings indicate that the interaction status between K8 and p38 is crucial in determining the efficiency of p38-Hsp70 complex formation, which could affect the subcellular localization of p38 and thereby the expression of p38-dependent target genes (Fig. 7). Taken together, our results suggest that K8, functioning as a cytoplasmic anchor for protein kinases, regulates the nuclear localization of p38 and thus might affect a number of signal transduction pathways.
A role of cytoskeletal proteins in subcellular localization of proteins
Numerous studies have shown the linkage between the dynamic rearrangements of cytoskeletal proteins and the expression of correlated genes. For example, monomeric G-actin in cytoplasm binds to myocardin-related transcription factors (MRTFs) and retains MRTF in the cytoplasm, whereas incorporation of G-actin into filamentous actin (F-actin) filament releases MRTFs to enter the nucleus and interact with serum response factor (SRF) (Olson and Nordheim, 2010). This interaction causes the expression of SRF-target genes. The actin cytoskeleton is also involved in localization of two related transcription coactivators, Yes-associated protein (YAP) and transcriptional co-activator with PDZ-binding motif (TAZ), both known as central effectors of the Hippo signaling pathway (Aragona et al., 2013).
The role of cytoskeletons in the subcellular localization of interacting proteins is somewhat controversial. Depending on cytoskeletal proteins and their associated proteins, their interactions can promote the nuclear localization or cytoplasmic retention of the complexes. Monomeric G-actin transports p53 towards the nucleus through direct protein-protein interaction (Saha et al., 2016). Suppression of microtubule dynamics by TN16, which perturbs microtubule assembly dynamics, increases the association of NF-κB with microtubules and facilitates nuclear translocation of NF-κB in a dynein-dependent manner (Rai et al., 2015). In contrast, microtubules bind directly to the myc-interacting zinc finger protein (MIZ-1) and contribute to its sequestration in the cytoplasm (Ziegelbauer et al., 2001); furthermore, the intermediate filament proteins K17 and K18 bind to the adaptor protein 14-3-3 and cause the retention of 14-3-3 proteins in the cytoplasm (Kim et al., 2006; Ku et al., 2002b). Although the cytoskeletal proteins clearly contribute to subcellular localization of interacting proteins, the molecular mechanisms, including identification of specific binding sites on the cytoskeletal proteins that interact with their partner proteins, are not fully understood.
Characteristics of the interaction between p38 and K8/K18
Intermediate filament proteins associate with various protein kinases, which phosphorylate them and then cause filament reorganization (Omary et al., 2006; Snider and Omary, 2014). For example, MAPKs phosphorylate lamins and cause disassembly of the nuclear lamina during mitosis (Barascu et al., 2012; Nigg, 1992; Peter et al., 1992). Keratins, along with other intermediate filament proteins, become phosphorylated and reorganized in response to stress signals (Omary et al., 2006; Snider and Omary, 2014). p38 MAPK, one of the stress-activated protein kinases, binds to K8/K18 heteropolymers through direct association with K8 and phosphorylates K8 Ser74 residue, which leads to keratin filament reorganization (Ku et al., 2002a). Interestingly, p38 phosphorylates K5 Thr150 and K6 Thr145 in the LLS/TPL motif, which is a highly conserved motif in K8 (72LLSPL) of type II keratins (Toivola et al., 2002). However, to date, the specific binding sites on intermediate filament proteins where interaction with protein kinases takes place have not been reported.
The findings presented here are the first evidence that the highly conserved docking domains in keratins and other intermediate filaments mediate their interaction with specific protein kinases. Our data clearly show that (1) the interaction of K8/K18 with p38 kinase is phosphorylation dependent [i.e. stress-induced phosphorylation/activation of p38 causes the dissociation of p38 from K8/K18 (Fig. 1)], (2) K8 has a specific docking site for interaction with p38 kinase (148RRQLETLGQEKLKL161) (Fig. 2) and (3) the p38 docking motif on K8 is highly conserved in other keratins and intermediate filament proteins (Table 2). The docking site on K8 is located in the rod domain distinct from the phosphoacceptor residue K8 Ser74, which is located in the head domain. However, the head domain in an assembled filament folds back on the rod domain and covers a coil 1A of rod domain in K8/K18 and other intermediate filaments (Herrmann and Aebi, 2016; Premchandar et al., 2015). Thus, the phosphoacceptor residue K8 Ser74 that nominally is located close to the N-terminus is probably very close to the p38 docking motif in the rod domain of K8, because of the folding of the head domain. Interestingly, the spacer residue between the basic residues and the hydrophobic groove in keratins is (R/K)1-2-X8-10-φ-X-φ, which differs from the conserved motif of D domains in other proteins, (R/K)1-2-X2-6-φ-X-φ. The longer spacer (X8-10) in keratins probably arises because of their helical structure. As the rod domain of keratins forms an α-helix with 3.6 amino acid residues per turn (Eisenberg, 2003), the distance between the basic residues and a hydrophobic groove might be closer than it seems; hence, the distance in keratins and other proteins could have high similarity.
In addition, our study demonstrates the molecular basis for selective interaction of K8 with p38, but not with other MAPKs such as ERK1/2 and JNK1/2 (Fig. S3). Although ERK, JNK and p38 all phosphorylate K8 Ser74 and/or Ser432 (Ku et al., 2002a; Omary et al., 2006), it seems that they do not share the same D domain in their interaction with K8, implying that the D domain has specificity for p38 MAPK but not for other kinases. In searching for potential D domains for p38 in K8, we found that there are six sequences in K8 and seven sequences in K18 that resemble the docking sequences of p38 (Table 1). Although the newly identified K8 D domain for p38 is not the D domain of other MAPKs, the remaining sites are putative D domains for other MAPKs.
p38 docking interaction with K8 and its subcellular localization
It has been shown that docking interactions could regulate the nucleocytoplasmic localization of MAPKs and MAPK-associated proteins. For example, the N-terminal residue of MEK encompassing the ERK-binding site can regulate the subcellular distribution of ERK (Fukuda et al., 1997). The p38 MAPK is distributed in both the cytosol and the nucleus, and phosphorylated p38 translocates into the nucleus in response to environmental stress (Wood et al., 2009). Although the underlying mechanism of the nucleocytoplasmic translocation of p38 remains largely unknown, processes dependent on β-like importins (Maik-Rachline et al., 2018) and mechanisms mediated by NLS-containing protein Hsp70 (Gong et al., 2012) have been reported to be involved in p38 nuclear translocation. In addition, a previous study has demonstrated that the microtubule depolymerizing reagent nocodazole can interfere with the nuclear translocation of p38, indicating the role of cytoskeletal proteins in p38 transport (Gong et al., 2010). However, a p38 docking site on cytoskeletal proteins, including tubulin, has not been identified. Here, we have identified a p38 docking site on K8 (Fig. 2) and demonstrated the role played by keratins in Hsp70-mediated p38 nuclear translocation (Fig. 7). Under basal conditions, K8 as a cytoplasmic anchor binds and sequesters p38 kinase; however, in response to stress, phosphorylated/activated p38 phosphorylates K8 at Ser74, which expedites p38 separation from keratin complexes. The released active form of p38 forms a complex with NLS-containing Hsp70, a potential chaperone for p38 translocation into the nucleus, which contains the substrates of p38 that are mediated in signaling pathways under stress conditions.
Previous studies have demonstrated that keratins are responsible for the localization of other proteins. K17 and K18 interact with the adaptor protein 14-3-3 and their interaction then causes the retention of 14-3-3 proteins in the cytoplasm (Kim et al., 2006; Ku et al., 2002b). K17 also regulates cancer cell-cycle progression and tumor growth by promoting nuclear export of tumor suppressor p27KIP1 (Escobar-Hoyos et al., 2015). Moreover, K19 is required for nuclear import of transcription factor early growth response-1 (Egr1) through increased interaction between Egr1 and importin-7 (Ju et al., 2013). Our results in this study are consistent with evidence that keratins influence the localization of other proteins.
Disrupted docking interaction between p38 and a liver disease-associated keratin variant
Although the K8 and K18 variants are associated with liver diseases (Ku et al., 2006, 2016; Omary et al., 2004, 2009), the molecular mechanisms by which the identified keratin variants cause a predisposition to liver damage are not fully understood. In this study, we demonstrated that one of the tested variants, K18 I150V, significantly decreased interaction with p38 and, at the same time, increased the formation of p38-Hsp70 complex involved in p38 nuclear localization. This resembles the effect of p38 docking-deficient mutations on K8. Given the structural organization of keratins K8 and K18, which intertwine to form heterodimers in which the helices run in parallel from the N- to the C-terminal (Herrmann et al., 2009), the observed results are probably because the location of the K18 I150V mutation is close to the location of the K8 docking site (Arg148/149 and Leu159/161) in heterodimers. Interestingly, the K18 I150 residue is localized in the conserved potential site of p38 docking, as shown in Table 2.
It is important to note that these molecular consequences could provide evidence for the impact of disease-causing mutations on the outcome of liver disease. Interestingly, the results of RT-qPCR demonstrated that the expression of tested inflammation-related genes tended to be enhanced in cells transfected with the K18 I150V mutant compared with control K18 WT (Fig. 6C). Notably, the increased levels of iNOS, VCAM and VEGFC are statistically significant (Fig. 6C). The iNOS family of enzymes causes hepatotoxic effects such as hemorrhagic shock, ischemia/reperfusion injury and endotoxemia (García-Monzón et al., 2000; Hierholzer et al., 1998; Nussler and Billiar, 1993). Nitric oxide generated by iNOS is closely implicated in the development of various types of liver disease such as hepatic fibrosis and cirrhosis (Iwakiri and Kim, 2015). Both VCAM and VEGFC have been shown to regulate chronic vascular inflammation in various disease models (Hsu et al., 2019; Kong et al., 2018). In summary, our results suggest that keratins, functioning as cytoplasmic anchors for protein kinases, regulate the nuclear localization of p38 and thereby could affect a number of signal transduction pathways and the expression of correlated genes.
MATERIALS AND METHODS
Antibodies and reagents
The antibodies used were as follows: L2A1 mouse monoclonal antibody (Ku et al., 2004); anti-K8/K18 rabbit polyclonal antibody 8592 (Ku et al., 2000), which recognizes human K18; rabbit anti-human/K18 antibody 4668 (Ku et al., 2000); anti-p38 MAPK, phosphorylated p38 MAPK (Thr180/Tyr182), lamin A/C, p44/42 MAPK, and JNK2 antibodies (Cell Signaling Technology, Danvers, MA); GFP, Hsp70 and JNK1/3 antibodies (Santa Cruz Biotechnology, Dallas, Texas); cytokeratin 8 (TS1) and actin antibodies (Thermo Scientific, Fremont, CA); anti-DDK antibody (OriGene, Rockville, MD); and α-tubulin antibody (Sigma-Aldrich, St. Louis, MO).
Various mutations in K8, K18 and the α-isoform of p38 (MAPK14) were constructed using PCR-based mutagenesis. The mutants were generated using the QuikChange II XL site-directed mutagenesis kit (Agilent Technologies, Santa Clara, CA) with designed corresponding pairs of oligonucleotide primers. Each mutation was verified by DNA sequencing.
Cell culture and transfection
BHK-21 cells obtained from American Type Culture Collection (Rockville, MD) were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal bovine serum (FBS), 1% penicillin/streptomycin and 0.1% methotrexate in the cell incubator (10% CO2, 37°C). Cells were transfected with plasmid DNA using jetPRIME reagent (PolyPlus, Berkeley, CA), according to the manufacturer's instructions. For analysis of cell fractionation, transfected cells were treated with OA (an inhibitor of protein phosphatases 1 and 2A) for 2 h (1 µg/ml).
Co-immunoprecipitation and immunoblot analysis
The cultured cells were collected and lysed in 1 ml of 0.5% Empigen (Sigma-Aldrich) in PBS (pH 7.4) containing 10 mM EDTA, 5 Mm sodium pyrophosphate and 50 mM sodium fluoride for 5 h at 4°C. After centrifugation at 14,000 rpm for 20 min, the supernatants were used for immunoprecipitation with specific antibodies. The immunoprecipitates were resolved by SDS-PAGE and then subjected to either Coomassie Blue staining or immunoblotting. For immunoblotting, proteins were transferred to polyvinylidene difluoride membranes (Millipore, Billerica, MA) and detected with the corresponding antibodies. Specific proteins were then visualized and quantified using enhanced chemiluminescence.
Cells were resuspended in buffer A (10 mM HEPES, 10 mM KCl, 0.1 mM EDTA, 0.1 mM EGTA, 1 mM DTT and protease inhibitor cocktail) and kept on ice for 10 min. IGEPAL CA-630 reagent (Sigma-Aldrich) was added to give 0.5% (v/v). The cytosolic fraction was isolated by centrifugation (800 g, 10 min at 4°C) and the remaining pellet washed three times with lysis buffer A and centrifuged again (800 g, 10 min at 4°C). The pellets were solubilized in buffer B (20 mM HEPES, 0.4 mM NaCl, 1 mM EDTA, 1 mM EGTA, 1 mM DTT and protease inhibitor cocktail) followed by incubation on ice for 30 min and centrifugation (16,000 g, 20 min at 4°C) to obtain the nuclear fraction. Fraction markers α-tubulin (cytosolic) or lamin A/C (nuclear) were used to verify the purity of each fraction.
RNA purification and real-time qPCR analysis
RNA purification and cDNA synthesis were performed as described previously (Roh et al., 2018). In brief, total RNA was isolated using TRIzol reagent (Ambion, Foster City, CA) and then cDNA was synthesized using a RevertAid First-Strand cDNA Synthesis Kit (ThermoFisher Scientific, Fremont, CA). RT-qPCR was conducted using SensiFAST Sybr No-Rox Mix (Bioline, London, UK). Quantification and relative gene expression analyses were performed using a CFX384 Real-Time PCR system (Bio-Rad, Berkeley, CA). Primers for RT-qPCR are described in Table S1.
Densitometry was prepared using ImageJ/FIJI software and the graph data are presented as the mean±s.d. Statistical analysis was performed using Student's t-test from three independent experiments and data analyzed using GraphPad Prism software. A P-value of less than 0.05 was considered statistically significant.
The authors thank Hayan Lee for editing of the manuscript.
Conceptualization: S.-Y.L., N.-O.K.; Methodology: S.-Y.L., Y.L., H.-N.Y., S.K.; Validation: S.-Y.L., Y.L., N.-O.K.; Formal analysis: S.-Y.L., N.-O.K.; Investigation: S.-Y.L., Y.L., H.-N.Y., S.K.; Writing - original draft: S.-Y.L.; Writing - review & editing: N.-O.K.; Visualization: H.-N.Y., S.K.; Supervision: N.-O.K.; Project administration: N.-O.K.; Funding acquisition: N.-O.K.
This work was supported by the Korean Ministry of Education, Science, and Technology (grants 2016R1A2B4012808 and 2018R1D1A1A02086060 to N.-O.K.) and the Yonsei University Research Fund (2018-22-0072 to N.-O.K.).
The authors declare no competing or financial interests.