Keratin filaments arise from the copolymerization of type I and II sequences, and form a pancytoplasmic network that provides vital mechanical support to epithelial cells. Keratins 5 and 14 are expressed as a pair in basal cells of stratified epithelia, where they occur as bundled arrays of filaments. In vitro, bundles of K5–K14 filaments can be induced in the absence of cross-linkers, and exhibit enhanced resistance to mechanical strain. This property is not exhibited by copolymers of K5 and tailless K14, in which the nonhelical tail domain has been removed, or copolymers of K5 and K19, a type I keratin featuring a short tail domain. The purified K14 tail domain binds keratin filaments in vitro with specificity (kD ∼2 μM). When transiently expressed in cultured cells, the K14 tail domain associates with endogenous keratin filaments. Utilization of the K14 tail domain as a bait in a yeast two-hybrid screen pulls out type I keratin sequences from a skin cDNA library. These data suggest that the tail domain of K14 contributes to the ability of K5–K14 filaments to self-organize into large bundles showing enhanced mechanical resilience in vitro.
Keratins (40–70 kD) are the major structural proteins of epithelial cells, where they occur as intermediate-sized filaments in the cytoplasm. The >49 functional keratin-encoding genes in the human genome (Hesse et al., 2001) occur as two major types, I and II. Keratin filaments are built from lateral and longitudinal interactions involving heterodimers, and most keratin genes are regulated in a pairwise and differentiation-related fashion within epithelia. A major function of keratin intermediate filaments (IFs)* is to endow epithelial cells with mechanical resilience. Mutations in keratin proteins result in fragility states whereby the cell type(s) affected ruptures when subjected to trauma (Fuchs and Cleveland, 1998). Bundles represent the preferential form of keratin filaments in complex epithelia. Whether they are examined in situ or in primary culture, basal keratinocytes of the skin exhibit loosely packed bundles of keratin filaments in their cytoplasm. However, there is no known cross-linking activity that is keratin filament–specific in the cytoplasm of these and many other epithelial cells in which keratin occurs as bundled arrays (Coulombe et al., 2000).
In vitro, keratin filaments can be induced to form large bundles after minor changes to the buffer conditions, in the absence of exogenous proteins. As expected, this significantly enhances the mechanical resilience of keratin gels (Ma et al., 2001). Along with other data (Coulombe et al., 2000), this suggests that keratin IFs can modulate their organization and mechanical properties though self-interaction. Here we provide evidence that the nonhelical tail domain of K14, the predominant type I keratin in basal cells of skin epithelia, binds keratin filaments in vitro and ex vivo, and contributes to the remarkable properties of keratin filaments.
Results And Discussion
The nonhelical tail domain contributes to the mechanical properties of cross-linked keratin gels
Polymerization of K5–K14 in low ionic strength buffer adjusted at pH 7.0, or at pH 7.4 with 10–20 mM NaCl added, results in the formation of large bundles of filaments that can be seen using DIC light microscopy. Rheological assays show that suspensions of bundled keratin filaments maintain their elasticity and solid-like character even when subjected to large deformations. These data (Ma et al., 2001) imply that stable filament–filament interactions, which depend upon cis-acting elements, occur under these conditions.
There is evidence suggesting that the COOH-terminal tail domain of cytoplasmic IF proteins may influence filament–filament interactions (Herrmann and Aebi, 1998). Moreover, the tail domain of K14 is not required for efficient copolymerization with K5 (Coulombe et al., 1990; Wilson et al., 1992). We compared the properties of wt K5–K14 and K5–K14ΔT assemblies at pH 7.0 to assess the role of K14's tail domain toward bundling. Whereas wt K5–K14 displays the texture of a gel under this condition, the K5–K14ΔT polymer (Fig. 1 A) flows like a liquid. In a high-speed pelleting assay designed to measure polymerization efficiency, >95% of both the wt K5–K14 and K5–K14ΔT protein pools are retrieved in the pellet (Fig. 1 B). In a low-speed pelletting assay devised to assess the formation of cross-linked networks (Pollard and Cooper, 1982), nearly all the wt K5–K14 proteins end up in the pellet, compared with only ∼20% for K5–K14ΔT. DIC light microscopy confirms the presence of extensive bundling in the wt K5–K14 sample, whereas few bundles can be seen for K5–K14ΔT (unpublished data).
In rheological studies, the wt K5–K14 and K5–K14ΔT assemblies produced at pH 7.0 both display properties typical of viscoelastic solids when subjected to small deformations (Fig. 1 C). K5–K14ΔT displays a slightly larger phase shift compared with wt K5–K14 (12° vs. 6°, respectively), indicating subtle differences in the solid-like character of these samples when in the linear regime. However, significant differences are seen when these samples are subjected to larger deformations. Whereas wt K5–K14 maintains its elasticity up to deformations exceeding 300% and shows a “bump” characteristic of strain hardening before softening, K5–K14ΔT begins to soften at 10% strain and shows comparatively less strain hardening before yielding (Fig. 1 C). Strain hardening can be captured in real time through a faster-than-linear increase of the shear stress σ(t) when it is monitored continuously while a fixed oscillatory deformation γ (of amplitude 100%) is applied to the sample (Ma et al., 2001). In pH 7.0 assembly buffer, wt K5–K14 displays significant strain hardening (Fig. 1 D). In contrast, the K5–K14ΔT polymer shows little of this behavior (Fig. 1 D).
The K14ΔT protein we used has a 12-amino-acid-long epitope tag at its COOH terminus (Albers and Fuchs, 1987). To rule out that its rheological properties stem in part from this short tag, we repeated these studies with wt K5–K19 polymer (Fradette et al., 1998). K19 features a short tail domain (13 amino acids) and is coexpressed with K5 in a subset of progenitor basal cells in the skin (Stasiak et al., 1989). Rheological studies show that K5–K19 behaves like a weak viscoelastic solid when subjected to small deformations at pH 7.0. However, as is the case for K5–K14ΔT (Fig. 1 B), the K5–K19 polymer rapidly softens when subjected to progressively larger deformations. At 200% strain amplitude, for instance, the wt K5–K14 and K5–K19 assembled at 0.2 mg/ml (4 μm) exhibit elastic moduli G′ of 38 ± 7 and 4.2 ± 0.2 dynes/cm2, respectively. Interestingly, immunoelectron microscopy studies involving human skin epithelia showed that K5–K19-rich basal cells exhibit a looser keratin network than K5–K14-rich basal cells (Dr. Lucie Germain, personal communication). The data we report here suggest that although the nonhelical tail domain of K14 is dispensable for 10-nm filament assembly, it contributes to the intrinsic potential of K5–K14 for interaction with self, which enhances its ability to withstand large deformations.
The purified K14 tail domain binds to keratin filaments in vitro
To assess whether the short nonhelical tail domain of K14 (∼50 residues) can interact directly with keratin filaments, we added an NH2-terminal histidine (His) tag to facilitate its purification. Purified His-T14 migrates with a mass of ∼7 kD on SDS-PAGE, reacts with antibodies directed against the His tag and the K14 COOH terminus (Fig. 2 A), and behaves as a monomer with an extended shape in solution (unpublished data). His-T14 cosediments with wt K5–K14 filaments produced under standard buffer conditions at pH 7.4 (Fig. 2, B and B′, kD ∼2 μM). In contrast, His-NPT (control for His tag) and insulin (size control) do not cosediment (Fig. 2 B), establishing specificity. His-T14 also cosediments with K8–K18 filaments and K5–K14ΔT filaments (Fig. 2 C), but not with vimentin filaments (Fig. 2 C) or F-actin (unpublished data). Therefore, the tail domain of K14 binds directly to determinant(s) shared between K5–K14 and K8–K18 filaments.
The outcome of the binding studies predicts that an excess of purified His-T14 should inhibit the gelation of keratin filaments induced by lower pH or added salt. We tested His-T14:keratin stoichiometries ranging from 1:8 to 3:1 (Fig. 2 B′). Bundle formation was scored through a low-speed centrifugation assay (10,000 × g, 20 min). In the absence of His-T14 and at pH 7.0, virtually all of the K5–K14 pool is retrieved in the pellet fraction (Fig. 2 D). Adding His-T14 at low stoichiometry (i.e., 1:4) caused a redistribution of wt K5–K14 to the supernatant (Fig. 2 D). This effect was not seen at the higher concentrations of His-T14 tested (Fig. 2 D). The reasons for this observation are unclear. These studies show that addition of the His-T14 can partially inhibit the gelation of wt K5–K14 induced by low pH under some conditions.
The K14 tail domain associates with keratin filaments in epithelial cells in culture
We performed transient transfection assays with constructs in which the K14 tail domain is fused to a myc epitope tag. In PtK2 cells, an epithelial cell line that displays keratin and vimentin IF networks, Myc-T14 is detected in the cytoplasm, where it occurs as a filamentous pattern, as well as in the nucleus, where it does not exhibit any pattern (Fig. 3, A and B). Double-immunofluorescence staining shows that in the cytoplasm, the signal for myc-T14 colocalizes with K8–K18 filaments to a significant extent (Fig. 3, A′, A″, B′, and B″), but not with vimentin IFs (unpublished data). The best instances of colocalization involved bundles of keratin filaments whose shape and organization are reminiscent of stress fibers. Dual immunostaining of untransfected PtK2 cells for F-actin and K8–K18 shows that the two types of filaments are closely apposed in a significant fraction of cells, indicating that this phenomenon is not a function of myc-T14 expression (unpublished data). Myc-T14 may thus show a greater affinity for a specific subset of keratin filaments, or weakly promote the formation of straight K8–K18 bundles. Alternatively, this observation may simply stem from the sensitivity of the assay. Transfected myc-PTE1, used as a control, exhibits a punctate cytoplasmic pattern that is independent from keratin filaments (Fig. 3, C–C″). Similar results were obtained in mouse 308 cells, a K5–K14-expressing skin keratinocyte line (unpublished data).
We next performed triple transfection assays in BHK-21 fibroblasts, a cell line that does not contain any keratin filaments. In cells coexpressing K5, K14, and Myc-T14, there is a significant spatial correspondence between the signals for myc and for K14 (Fig.3, D–D″). Again, a significant fraction of the myc signal is localized to the nucleus (Fig. 3 A). In contrast, the transfected myc-PTE1 localizes both to the nucleus and cytoplasm, but is clearly distinct from K5–K14 filaments (Fig. 3, E–E″). We conclude that the nonhelical tail domain of K14 can specifically bind keratin filaments in the cytoplasm of various cell lines in culture.
The K14 tail domain interacts with type I keratin sequences in a yeast two-hybrid screen
In an effort to isolate tail domain binding proteins, the COOH termini of both human K14 and K16 were used as baits in a yeast two-hybrid screen involving a human cDNA library made from a squamous cell carcinoma of the leg. Among 2,000,000 cfus screened with the K14 tail domain, 121 scored positive for both growth on His-minus medium and lacZ expression. Among the 40 clones sequenced, 32 (80%) corresponded to type I keratin sequences, and an additional 2 clones corresponded to a keratin-related sequence. The 5′ boundary of all the interacting type I keratin clones is given in Fig. 4 A. These findings can be interpreted in two ways. First, the K14 tail domain binds multiple sites along type I keratin sequences (Fig. 4 B). Some of these sites would be shared among them (i.e., linker L12), and others potentially unique (e.g., the head domain of K14). Second, the tail domain binds a restricted number of sites, perhaps a single one, located within the COOH-terminal half of the rod domain of many type I sequences (Fig. 4, A and B). In this case, the heterogeneity observed at the 5′ end would reflect the manner with which the reverse transcriptase operated while making the library. Of note, all type I sequences contained at least coil 2 from the rod, and no type II keratin clones were pulled out. Among 3,500,000 cfus screened with the K16 tail domain, 581 scored positive. Among 143 clones that were sequenced, 132 encoded elongation factor 1 γ, and none encoded a keratin. The complete results of this screen will be published elsewhere. Meanwhile, these studies provide genetic evidence that K14's tail domain, but not K16's, can bind sequence determinant(s) shared by several type I keratins. These findings are interesting given the functional differences observed between K14 and K16 in the context of a protein replacement study in transgenic mouse skin (Paladini and Coulombe, 1999; Wawersik and Coulombe, 2000). These differences are, in part, due to the COOH-terminal 105 amino acids in these two homologous type I keratins. We have rheological evidence that suspensions of K5–K16 filaments are mechanically weaker than K5–K14 ones when prepared and tested under bundling-promoting conditions (unpublished data).
Defining a role for the nonhelical tail domain of keratins
The properties of the nonhelical tail domain of cytoplasmic IF proteins have been studied through mutagenesis. The consensus that emerged is that the tail domain contributes to the lateral growth and stability of filaments, and influences filament–filament interactions (Herrmann and Aebi, 1998). The tail domain of many type I keratins is not required for complete copolymerization with a type II keratin (Albers and Fuchs, 1987; Coulombe et al., 1990; Hatzfeld and Weber, 1990; Lu and Lane, 1990; Bader et al., 1991; Wilson et al., 1992). Limited proteolysis studies in vitro (Herrmann and Aebi, 1998) and accessibility to kinases in vivo (Omary et al., 1998) suggest that the tail domain of IF proteins is exposed at the filament surface. Extensive bundling of K5–K14 filaments in vitro can be promoted by small changes in the assembly buffer conditions (Ma et al., 2001). Here we show that this depends, in part, on the presence of the tail domain of K14. We provide biochemical and genetic evidence that the isolated tail domain behaves like a keratin binding protein, with an affinity comparable to that of most F-actin cross-linkers (Matsudaira, 1994). We propose a model (Fig. 4 C) in which the nonhelical tail domain of K14 (and possibly other keratins) contributes to the ability of keratin filaments to self-organize into bundles. Such interactions are postulated to assist in keratin organization and function in vivo. In support of this model, a frameshift mutation resulting in the loss of the COOH-terminal 21 residues in K1 was recently implicated in a failure of keratin filaments to bundle in differentiating epidermal keratinocytes, giving rise to a rare variant of epidermolytic hyperkeratosis (Sprecher et al., 2001).
The binding site(s) for the K14 tail domain on the keratin polymer is displayed by K5–K14, K5–K14ΔT, and K8–K18 filaments, and may be shared by several type I keratins (as suggested by the yeast two-hybrid screen). Electrostatic forces may well play a role in these interactions, as filament–filament contacts can be modulated by changes in pH and in salt conditions. Filaggrin, a known keratin cross-linker, binds keratin filaments via electrostatic interactions (Mack et al., 1993). Likewise, the 50-residue-long IF binding motif contained within the tail domain of plectin is rich in basic residues (Nikolic et al., 1996). Given the estimate of ∼700 keratin monomers per μm of filament length (Herrmann and Aebi, 1998), only a subset of polymer-bound tail domains need be exposed and involved in these interactions to account for a filament cross-bridging effect (Fig. 4 C). Additional studies are needed to define the binding sites and mechanism(s) involved, determine whether this applies to other keratin polymers, and assess the importance of this phenomenon for the mechanical scaffolding function of keratin polymers in vivo.
Materials And Methods
Construction of fusion cDNAs
Preparation of His fusion proteins.
For expression in Escherichia coli, the human K14 tail coding sequence (Marchuk et al., 1984) was subcloned into a bacterial expression vector, pT7–6 × His (Geisbrecht et al., 1999). The NH2 terminus of the resulting fusion protein is MGSSHHHHHHSSGLVPRGSHMASGQSTM427S, where the tag is underlined and 427S corresponds to residue 427 in K14. His-NPT, used as a control, features the neomycin phosphotransferase sequence (GenBank/EMBL/DDB accession no. AF335419; MW 29.2 kD).
Preparation of Myc fusion proteins.
For expression in mammalian cells, the K14 tail sequence was transferred to vector pcDNA3-Nmyc (Jones et al., 1999), which has a CMV promoter and a polyadenylation signal. The NH2 terminus of the resulting fusion proteins is MAEQKLISEEDLLGSGSTM427S, where the tag is underlined and 427S correspond to the first residue of K14's tail domain. Myc-PTE1, which codes for a myc fusion to acyl-CoA thioesterase (Jones et al., 1999), was used as a control.
Production, expression, and purification of recombinant proteins
Plasmids pET-K5, pET-K14 (Coulombe and Fuchs, 1990), pET-K14ΔT (Coulombe et al., 1990), and pET-K19 (Fradette et al., 1998) were used to generate recombinant human proteins. For His-tagged proteins, the plasmids pET-His-T14 and pET-His-NPT were transformed into E. coli strain BL21 (DE3), and protein purification was done using a nickel column (Novagen). Routine electrophoretic assays were used to assess protein size, purity, and identity.
Filament assembly and analysis
Keratin IFs (0.5 or 1 mg/ml) were reconstituted by serial dialysis starting from type I–type II heterotypic complexes (Ma et al., 2001) using the following three buffers at room temperature: (a) 9 M urea, 25 mM Tris-HCl, pH 7.4, 25 mM β-ME for 4 h; (b) 2 M urea, 5 mM Tris-HCl, pH 7.4, 5 mM β-ME, for 1 h; and (c) 5 mM Tris-HCl, 5 mM β-ME, for >12 h (overnight). The pH of the last buffer was adjusted at values of 7.4 or 7.0, depending on the experiment. The physical state of the polymer and polymerization efficiency were assessed as described (Ma et al., 2001). Filament morphology was examined by negative staining (1% uranyl acetate) and electron microscopy (Philips CM120, Germany). Sampling was restricted to regions of the sample where individual filaments could be seen. Larger polymer structures (i.e., bundles) were examined using DIC light microscopy (Eclipse; Nikon). Bundling was assessed by subjecting the assemblies to low-speed centrifugation (8,200 × g for 20 min) (Pollard and Cooper, 1982) followed by SDS-PAGE. Human recombinant vimentin was assembled as described (Herrmann and Aebi, 1998). Purified actin and protocols were provided by Enrique de la Cruz (Yale University, New Haven, CT). For copelletting assays, 50 μl of assembled keratin (10 μM), vimentin (10 μM), and actin filaments (5 μM) were mixed with either His-T14, His-NPT (Geisbrecht et al., 1999), insulin (Sigma-Aldrich), or vinculin's tail domain, a gift from Dr. Susan Craig (Johns Hopkins University, Baltimore, MD). After 2 h, the mixtures were centrifuged at 150,000 × g for 30 min, and the pellet and supernatant fractions were analyzed by SDS-PAGE (Ma et al., 2001).
All measurements (Coulombe et al., 2000) were obtained using a strain-controlled 50-mm cone-and-plate ARES 100 rheometer (Rheometrics, Inc.) as described (Ma et al., 2001). The linear equilibrium values of the elastic modulus G′ (ω) and loss (viscous) modulus G″(ω) of the suspensions were measured by setting oscillatory strain amplitude at γ = 1% and sweeping from low to high frequency ω. Strain-dependent viscoelastic moduli were measured by subjecting polymers to three cycles of deformation of increasing amplitude at 1 rad/sec−1; G′ and G″ were computed from the maximum magnitude of the measured stress. To assess strain hardening, we conducted assays in which the stress was continuously monitored as eight cycles of oscillatory shear deformation were applied (Ma et al., 2001). These studies were repeated three times using independent protein preparations.
Yeast two-hybrid screen
PCR products corresponding to the human K14 tail domain, starting at 425His (Marchuk et al., 1984), and the K16 tail domain, starting at 426Ser (Paladini et al., 1995), were fused in frame to the GAL4 DNA binding domain in the yeast expression vector pAS2–1 (ClONTECH Laboratories, Inc.). For library screening, the Saccaromyces cerevisiae strain Y190 was transformed with AS-T14 or AS-T16. Single colonies were selected after growth in synthetic minimal carbon source (2% sucrose) medium lacking Trp, grown, and retransformed with a human skin cDNA library fused in GAL4 activator domain in pACT2 by a PEG-based method. This library was constructed from poly(A) mRNAs extracted from a squamous cell carcinoma including adjacent normal tissue. mRNA priming for reverse transcription was done using random hexamers, and products of 0.5–2.0-kb size were used to make the library. Transformed yeast cells were plated on SC medium lacking Trp, Leu, and His in the presence of 25 mM 3-amino-1, 2, 4-triazole, incubated at 30°C for 7 d. Growing colonies (His+) were subjected to β-galactosidase assay. Plasmids harboring cDNA were recovered from positive colonies and sequenced. As negative controls, AS-T14 and AS-T16 were tested for their interaction with K5 or K6 protein.
Cell lines, transfections, and immunological reagents
Cell lines were maintained as recommended by their sources: PtK2 (kidney epithelium of the rat kangaroo; American Type Culture Collection); 308 (mouse epidermal keratinocytes), a gift of Dr. Stuart Yuspa (National Cancer Institute, Bethesda, MD); and BHK-21 (baby hamster kidney fibroblasts; American Type Culture Collection). Transient transfections were performed using polyethylenimine (Sigma-Aldrich) as described (Boussif et al., 1995). At 24 or 48 h posttransfection, cells were fixed with 3.0% paraformaldehyde for 15 min, extracted with 0.01% Triton-X in PBS buffer for 5 min, and processed for indirect immunofluorescence. We used rabbit polyclonal antisera against K18 and K14, obtained from E. Fuchs (University of Chicago, Chicago, IL); mouse monoclonals against vimentin (V9; Sigma-Aldrich); the myc epitope (American Type Culture Collection) K14, a gift from Dr. Irene Leigh (Imperial Cancer Research Fund, London, UK), and K8–K18, a gift from Dr. Bishr Omary (Stanford university, Stanford, CA); and chicken polyclonal against K14 (Wawersik and Coulombe, 2000). For immunofluorescence microscopy, rhodamine- or FITC-conjugated goat anti–mouse or anti–rabbit secondary antibodies were used (Kirkegaard and Perry Labs). For Western blotting, we used enhanced chemiluminescence (Amersham Pharmacia Biotech) or alkaline phosphatase (Bio-Rad Laboratories) detection.
We thank Ms. Kelsie Bernot and Dr. Brian Geisbrecht for their generous help, and Drs. Steve Gould and Susan Craig (Johns Hopkins University), Stuart Yuspa, Enrique De La Cruz, Elaine Fuchs, Irene Leigh, and M. Bishr Omary for providing reagents.
These studies were supported by National Institutes of Health grants AR44232 and AR42047 to P.A. Coulombe, and National Science Foundation Grants CTS9812624 and DB19729358 to D. Wirtz.
O. Bousquet and L. Ma contributed equally to this work.
Abbreviations used in this paper: His, histidine; IF, intermediate filament.