Residues in the 1A rod domain segment and the linker L2 are required for stabilizing the A11 molecular alignment mode in keratin intermediate filaments.

Both analyses of x-ray diffraction patterns of well oriented specimens of trichocyte keratin intermediate filaments (IF) and in vitro cross-linking experiments on several types of IF have documented that there are three modes of alignment of pairs of antiparallel molecules in all IF: A11, A22 and A12, based on which parts of the major rod domain segments are overlapped. Here we have examined which residues may be important for stabilizing the A11 mode. Using the K5/K14 system, we have made point mutations of charged residues along the chains and examined the propensities of equimolar mixtures of wild type and mutant chains to reassemble using as criteria: the formation (or not) of IF in vitro or in vivo; and stabilities of one- and two-molecule assemblies. We identified that the conserved residue Arg10 of the 1A rod domain, and the conserved residues Glu4 and Glu6 of the linker L2, were essential for stability. Additionally, conserved residues Lys31 of 1A and Asp1 of 2A and non-conserved residues Asp/Asn9 of 1A, Asp/Asn3 of 2A, and Asp7 of L2 are important for stability. Notably, these groups of residues lie close to each other when two antiparallel molecules are aligned in the A11 mode, and are located toward the ends of the overlap region. Although other sets of residues might theoretically also contribute, we conclude that these residues in particular engage in favorable intermolecular ionic and/or H-bonding interactions and thereby may play a role in stabilizing the A11 mode of alignment in keratin IF.

To date, approximately 50 different genes encoding intermediate filament (IF) 1 chains exist in mammalian genomes. Based on differences in the organizations of their primary structures and genes, six different types of IF are now known (see Refs. 1-4 for reviews). The type I and type II keratins are the most numerous. In human, for example, each contains approximately 20 members, which are differentially expressed in various epithelial tissues. Each may be further divided into about 20 trichocyte keratin chains expressed almost exclu-sively in "hard" keratinizing tissues such as hair, and 20 cytokeratins. All keratin (as well as other) IF chains consist of a central rod domain composed of four ␣-helical segments (1A, 1B, 2A, and 2B) that possess a heptad repeat motif and are separated from one another by non-␣-helical linkers. The central rod domain is flanked on the head and tail by domains of differing size and chemical character. A large body of experimental evidence has now documented that the fundamental building block of all keratin IF is the heterodimer molecule, consisting of one type I and one type II chain (1)(2)(3)(4)(5)(6)(7). Although a number of important details remain to be resolved, this molecule is known to be stabilized in large part by the formation of a segmented ␣-helical coiled-coil by the appropriate parallel alignment of the central rod domain segments on the two chains. The next step is the formation of a pair of such molecules. Typically, this oligomer is the minimal IF structure that exists in solution, especially below the critical protein concentration required for assembly into macroscopic IF (ϳ40 g/ml). A number of biophysical, electron microscopic, and biochemical experiments have documented that the two molecules are aligned antiparallel and partly staggered in the A 11 or A 22 alignment mode, depending on whether the 1Aϩ1B or 2Aϩ2B rod domain segments overlap. Cross-linking data on cytokeratins and trichocyte keratins have revealed that both of these modes co-exist in solution, presumably in equilibrium with each other, as various experimental manipulations allow realignments (8 -10). Further, the cross-linking data have afforded quantitative estimates of the degree of overlap of the molecules. Thus, for the A 11 mode we have documented that the two molecules are displaced by approximately 112 amino acid residues with respect to each other. However, fundamental questions remain concerning the sequence features that specify and stabilize these alignment modes. In this study we have explored in K5/K14 IF which sequences may be involved in stabilizing the A 11 alignment mode. In this study, we have tested a current hypothesis that charged residues located along the rod domain segments may be important for molecular registration. Table I lists all possible charged residues that theoretically could be involved. By use of series of point substitutions of charged residues, we have identified several conserved residue positions that are important for stabilizing the A 11 alignment mode.

MATERIALS AND METHODS
Expression and Purification of K5 and K14 Chains-Full-length human K5 and K14 cDNAs were assembled into a pET11a vector and expressed in bacteria as described (12). Several mutant forms of both chains were generated by use of the QuickChange site-directed mutagenesis kit (Stratagene) (Table II). DNA sequencing was performed to confirm the mutations. Following induction, inclusion bodies were re-* The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom all correspondence should be addressed: NIAMS, Bldg. 6 covered, dissolved in SDS-PAGE buffer, and resolved in 3-mm-thick slab gels. The desired keratin bands were cut out, eluted into SDS gel buffer over night, and the solutions stored at Ϫ70°C. Protein concentrations were determined by amino acid analysis following acid hydrolysis.
In Vitro IF Assembly-Equimolar mixtures of either a wild type and/or mutant K5 and K14 chain were made from the stored SDS gel buffer solutions. The SDS was removed by ion-pair extraction (13) and the pelleted acetone-wet proteins redissolved (0.05 or 0.5 mg/ml) in a buffer of 9.5 M urea containing 50 mM Tris-HCl (pH 7.6), 5 mM Tris(2carboxyethyl)phosphine-HCl (TCEP) (Pierce), and 1 mM EDTA. For electron microscopy studies, IF were assembled by 1-h dialyses through solutions of decreasing urea solutions of 4, 2, and 1 M, and finally into assembly buffer of 10 mM Tris-HCl (pH 7.6), 1 mM EDTA and 5 mM TCEP (12). Final protein concentrations were 35-40 g/ml, which is below the critical concentration (C o ) for IF assembly (14), wherein mostly two-molecule assemblies formed, or 400 g/ml for optimal IF assembly. Particles were examined by electron microscopy following negative staining with 0.2-0.7% uranyl acetate over holey carbon film grids. Lengths of IF were measured (15) in fields of Ն400 m 2 . For IF assembly efficiency studies, protein mixtures in 9.5 M urea (40 l of Ϸ500 g/ml) were dialyzed directly into assembly buffer for 4 h. Solutions were then pelleted at 100,000 ϫ g for 30 min in an Airfuge (Beckman Instruments). Yields of protein in pellet were estimated by measuring the absorbance at 276 nm of the supernatant.
Transfection Experiments with K14-Green Fluorescent Protein (GFP) Plasmids-A construct encoding GFP coupled at the 5Ј end of the full-length coding sequence of wild type K14 was a generous gift of Dr. R. D. Goldman (Northwestern University Medical School, Chicago, IL). Point mutations were made in the plasmid as described above.
PtK2 (NBL-5) cells, epithelial-like rat kangaroo kidney cells, were obtained from ATCC (no. CCL-56). The cells were grown in 25-cm 2 tissue culture flasks and maintained in MEM (Eagle's minimal essential medium with nonessential amino acids, Earle's salts and reduced sodium bicarbonate at 0.85 g/liter) (Life Technologies, Inc.) with 10% fetal bovine serum. For cell passage, the cells were grown to near confluence, and the medium was aspirated, washed once with phosphate-buffered saline, and trypsinized for 20 s (0.25% trypsin; Life Technologies, Inc.). The trypsin solution was aspirated, and the cells   Table III. A, wild type; B, K5 wild type and K14 1A Arg 10 3 Leu; C, K5 wild type and K14 1A Arg 10  were left at room temperature for 3 min. Five milliliters of medium were pipetted over the cells to dislodge them from the flask, and transferred to a 15-ml conical tube. Following 5 min of 1000 rpm centrifugation to pellet the cells, the medium was aspirated, and the cells were resuspended in 2 ml of medium and counted.
For direct immunofluorescence studies, 3 ϫ 10 5 cells/ml were plated in 35-mm sterile tissue culture dishes, each containing a glass coverslip. After 24 h, the cells were transfected with 1 g of plasmid DNA and 3 g of Lipofectin as described by the manufacturer (Life Technologies, Inc.). After 4 h, the mix was aspirated and 1 ml of 15% glycerol in Keratinocyte-SFM (Life Technologies, Inc.) was applied for 3.5 min. The glycerol solution was replaced with 2 ml of fresh medium and the cells incubated at 37°C with 5% CO 2 for at least 24 h. The coverslips were washed in phosphate-buffered saline and mounted onto glass slides with Gel/Mount (Biomeda Corp.). Intracellular localization of GFP fusion proteins was determined by direct fluorescent microscopy.
Protein Chemistry Procedures-To examine molecular stabilities, equimolar mixtures of the desired K5/K14 chains (ϳ40 g/ml) were equilibrated by 2-h dialyses into urea solutions of the desired concentration in a buffer of 10 mM triethanolamine (pH 8.0). The proteins were cross-linked with 25 mM disulfosuccinimidyl tartrate (DST) for 1 h at 23°C, and terminated with 0.1 M NH 4 HCO 3 (final concentration) (16). Although significant random cross-linking also occurs, these conditions were used because the near quantitative modification of all lysines allows for less diffuse bands on 3.75-7.5% gradient PAGE gels.
To assess molecular alignments in the A 11 and A 22 modes, crosslinking with DST was performed using 0.4 mM reagent as described before (8,9). We used wild type and mutant proteins that had been equilibrated into assembly buffer at about 40 g/ml for 1 h. In this case, Ͻ10% of the lysine residues were chemically modified, except for several aligned residues that formed cross-links with yields of up to about 0.3 mol/mol. Following cleavage with CNBr and trypsin digestion, peptides were resolved by HPLC as before, except that a non-linear gradient over a 120-min time period was used. The positions of elution of the peptides cross-linked by DST corresponding to the A 11 and A 22 molecular alignment modes were similar to those published previously (9), although many were confirmed by sequencing for five Edman degradation cycles on a Porton LF-3000 sequencer. Semiquantitative estimates of molar yields of each were made based on peak heights of the integrated HPLC profiles.

RESULTS AND DISCUSSION
In this paper we have made a systematic analysis of those charged residues that, based on current structural information, are located in rod domain positions that could influence the specificity and stability of the A 11 alignment mode of a pair of antiparallel heterodimer molecules in K5/K14 IF. These encompass the segments 1A, 1B, 2A, and beginning of the 2B, as

FIG. 4. Stabilities of dimer (one-molecule) and tetramer (twomolecule) assemblies of wild type and/or mutant chains in concentrated urea solutions (as shown) following cross-linking.
These and all other data are summarized in Table IV. The compositions of the assembly reactions are as shown. T, D, and M, respectively, mark the position of migration of the tetramer (two-molecule), dimer (onemolecule), and single-chain species.
well as the linkers L1, L12, and L2. We found that 41 charged positions have been conserved in the type II keratin 5 (K5) (Fig.  1, upper row) and type I K14 ( Fig. 1, lower row) chains. Based on extant ideas (2-4), we have hypothesized in this study that some of these may influence molecular alignment stabilities. Indeed, using the known quantitative estimate of molecular spacing of the A 11 alignment mode for keratin IF (about Ϫ112 residues) (9), we document in Table I that most of the 41 conserved charged residue positions lie opposite to each other and so are well sited to theoretically form stabilizing ionic salt bond pairs and/or H-bonds. Nevertheless, we discharged all conserved charged residue positions (i.e. mutated them to a non-charged residue) (all mutations are listed in Table II). In addition, there are 59 residue positions in this set that are not conserved between the K5 and K14 chains, and 4 others that are oppositely charged; several of these are also theoretically good candidates to form stabilizing salt bonds ( Table I). Some of these residue positions were discharged as well. We then examined the facility with which equimolar mixtures of one mutant and one wild type chain could assemble into one-and two-molecule oligomers, as well as IF in vitro and in vivo.
Assembly of IF in Vitro and in Vivo-The initial criterion of assembly competence was formation of pelletable IF particles by use of a sedimentation assay in the Airfuge. Experience has shown that particles must be Ն750 kDa in size in order to pellet with high efficiency. 2 This corresponds to an oligomer of as many as 16 chains (8 molecules), i.e. it consists of a full-length half-width entity characteristic of an early stage of IF assembly (17). In almost all cases, however, we found empirically that assembly of mixtures of mutant and/or wild type chains either resulted in macroscopic IF (Ն0.5 m long), which were readily pelletable in the Airfuge and clearly visible by electron microscopy after negative staining, or no large IF particles were formed at all (Ͻ0.1 m long and Ͻ4 nm wide), which did not pellet in the Airfuge and required examination over holey carbon film grids to be visible.
Sixty-six combinations of K5/14 chains were examined in in vitro assays (Fig. 2, Table II 2, 3, 10). This displays the close proximity of the two conserved and Arg/Lys 10 residues of one molecule (blue lines, large blue dots) with the conserved acidic residues Glu 4 and Glu 6 in the L2 linker (red lines, large red dots) of the other molecule. In addition, large dots delineate the possible interactions between 1A Lys 31 and 2A Asp 1 . Smaller dots delineate possible interactions involving 2A Asp 3 and L2 Glu 7 . We hypothesize that these may form several intermolecular salt and/or H-bonds and thereby contribute essential specificity and stability to the A 11 alignment mode. The segments of the molecules are marked. alignment mode; closed circles indicate those denoting the A 22 mode. Semiquantitative information of each peak is listed in Table V. in high yield and appeared as native-type IF Ͼ1 m in length. However, several combinations did not, including three positions in the 1A rod domain segment (K14 Asp 9 3 Ala, Arg 10 (Table III).
In a related second set of experiments, nine of these mutations were introduced into the GFP-K14 construct and their propensities for assembly into keratin IF in vivo were examined after transfection into PtK2 cells (Fig. 3). These cells express predominantly the K6, K7, K16, and K17 keratin chains but have been shown previously to accommodate incorporation of transfected wild type or mutant K14 chains (19). Additionally, the efficacy of incorporation of transfected GFP-K14 constructs into cultured cells to explore keratin IF cytoskeletons is now established (20). 3 Four mutants (1A Arg 10 3 Leu (Fig. 3B), 1A Lys 31 3 Met (Fig. 3D), L2 Glu 6 3 Ala (Fig.  3G), and 2B Glu 106 3 Ala (Fig. 3H)), resulted in severely disrupted cytoskeletons in which most of the keratin IF had withdrawn to a perinuclear location, and there were bright spots of unassembled GFP-labeled protein. In five other cases, the keratin IF cytoskeletons were either unchanged (1A Lys 17 3 Met (Fig. 3C), 1B Lys 71 3 Ile (data not shown)), or mildly abnormal due to some apparent clumping and/or elongation of the keratin IF (1A Glu 22 3 Ala (data not shown), 1B Glu 56 3 Ala (Fig. 3E), and 1B Glu 84 3 Ala (Fig. 3F)).
Some of these data were expected and thus serve as controls. The 1A positions 9 and 10 have been shown previously to be sites for mutation in various keratinopathy diseases (18); in vivo and/or in vitro expression of proteins containing these mutations revealed limited or no IF assembly (21). Similarly, we have recently documented that the 2B residue positions 100, 104, and 106 are required to form stable molecules because they participate in coiled-coil trigger formation in IF (11).

Cross-linking Studies with DST in Urea Solutions to Assess
One-and Two-molecule Stabilities-The second criterion of assembly competence used in this study was the formation of stable one-and two-molecule assemblies. We have previously established (16) a method to assess the stabilities of single coiled-coil molecules and pairs of them by use of a graduated urea concentration titration assay coupled with cross-linking by DST. At protein concentrations below the critical concentration for IF assembly (ϳ40 g/ml) in assembly buffer in the absence of urea, the K5 and K14 chains form mostly twomolecule (and traces of one-, three-, and four-molecule) oligomers (8,9). These dissociate into single molecules at about 6.5 M urea (approximate concentration of half loss), and then the molecules dissociate to individual chains by about 9.5 M urea, as reported earlier by Wawersik et al. (22) for K5/14 keratin IF and for vimentin and ␣-internexin (16) (see Fig. 4A).
Mutants representing every single conserved charged residue position in either the K5 or K14 chain (from Fig. 1), and some nonconserved ones, were tested in this assay. Several observations are apparent (Table IV). First, for only the 2B rod domain positions 100, 104, and 106 were both the two-molecule and single-molecule entities unstable in even 1 M urea. This is expected from our earlier data, as these residues participate in the formation of a stabilizing coiled-coil trigger motif for IF (11). Second, in all other cases, the one-molecule species was essentially as stable as the wild type. However, third, there were several conserved charged residue positions that resulted in significantly destabilized two-molecule entities (ϳ4 M urea), including  a Axial positions are measured in terms of h cc , which corresponds to a 0.1485-nm rise of each residue in a coiled coil conformation. This list documents the oppositely charged residues that are located within Ϯ3 residues of each other (and thus theoretically could form a salt bond) when aligned using the known parameters of A 11 mode adduced for keratin IF. Asterisks identify those residue positions/pairs which were confirmed to be of importance in this study. two-molecule oligomer, known as A 11 , A 22 , and A 12 . In our hands, the A 12 mode exists only at high pH values and is not assembly-competent (14). However, a variety of chromatographic, ultracentrifugation, electron microscopic, solution birefringence, and cross-linking data have documented that twomolecule oligomer of a variety of mammalian IF exist in assembly-competent solutions as 60 -70-nm-long particles in which the two molecules are aligned in the A 11 and/or A 22 mode. Our previous cross-linking experiments have shown that the two must co-exist in solution, since we have been able to recover DST cross-linked peptides arising from links between antiparallel molecules aligned in both modes (8 -10). Therefore,  19 3 Leu 5Ј-TAC CTG GAC AAG GTG CtT GCT CTG GAG GAG GCC-3Ј X 1A Glu 22  we reasoned in the present experiments that the destabilization of the two-molecule oligomer in the several mutations identified above should be due to loss of one or both of these alignment modes. To check this, we performed additional larger scale cross-linking experiments with 0.4 mM DST. The proteins were then cleaved with CNBr and trypsin, and the resulting peptides were resolved by HPLC (Fig. 5), but using a broader and flatter gradient extending over 120 min versus 70 min previously (8). We found six common peaks in the one-and two-molecule species of wild type K5/14 arising from intramo-   (Fig. 5B). In the wild type two-molecule oligomer, there were an additional 15 peaks due to intermolecular links, of which 5 could be assigned to linkages between molecules aligned in the A 22 mode (Fig. 5C, closed circles), and 10 to linkages denoting the A 11 alignment mode (Fig. 5C, open circles). Semiquantitative data on the amounts of each were determined on the basis of peak areas (all data summarized in Table V). These experiments were repeated for seven mutant mixtures. As found previously (11), the Glu 106 3 Ala substitution in the 2B rod domain segment resulted in loss of the A 22 alignment mode, and resultant substantial loss of the A 11 mode. However, the 1A Arg 10 3 Leu (Fig. 5D) and Lys 31 3 Met, and L2 Glu 6 3 Ala single (Fig. 5E) or Glu 4 3 Ala/Glu 6 3 Ala double substitutions resulted in almost complete loss of the A 11 mode. Further, the yields of the cross-links denoting to the A 22 mode were generally increased over the wild type amounts (Table V). These data confirm that the A 11 and A 22 modes of molecular alignment in fact exist in equilibrium in solution and suggest that loss of the former by destabilization results in a net reduction of the stability of all tetramers, together with an accumulation of molecules into the latter. The Arg 10  Substitutions in Keratinopathy Diseases-Thus, we have presented three sets of data, which document that certain residues along the keratin IF chains are especially important for: successful IF formation in vitro and in vivo; the stability of the two-molecule hierarchical stage of IF assembly; and, in particular, for specifying and stabilizing the A 11 mode of alignment of two antiparallel molecules. Indeed, several of the residues identified here correspond to residue pairs documented in Table I that are theoretically good candidates to form stabilizing ionic salt bonds. Based on the known alignment parameters of two antiparallel molecules in the A 11 mode, these residues are likely to lie very close to each other in the A 11 mode (Fig. 6). Thus, the conserved Arg 10 position of the 1A rod domain segment is closely adjacent to the conserved set of two (and in type I IF chains, three) acidic residues in positions 4, 6, and 7 in the linker L2. Notably, discharging of any one of these residues severely compromised the A 11 alignment of the two-molecule hierarchical stage of IF structure. Asp 9 (often an isosteric Asn in many IF chains) is likewise adjacent to these residues in L2. In addition, we note from Fig. 6 that the conserved 1A residue Lys 31 lies near the conserved 2A residue Asp 1 and Asp 3 (K5 chain only); likewise, discharging of these residues resulted in impaired stability of the A 11 alignment mode.
The simplest explanation of these data is that the key residues identified in this study interact to afford essential stability. One possibility is that this stability is provided by the formation of a complex intermolecular network of salt bonds and/or H-bonds. However, we cannot formally exclude the possibility that head and/or tail domain sequences also cooperate in these stabilizing phenomena. In addition, it is to be expected that many other charged residues, in addition to the key ones identified here, may also contribute in important ways to the alignment of the A 11 mode. It is also possible that these residues may participate in higher orders of IF structure, in particular the lateral association of molecules in the A 12 alignment mode, and elongation of molecules by overlapping of the A CN alignment mode. The availability of the complete atomic structure of a single IF molecule should provide the opportunity to further explore these possibilities in model building studies. Finally, it is interesting to note that the key potential interactions identified here do not involve the 1B segment, which corresponds to the central region of the A 11 overlap. Instead, both ends appear to be crucial in making favorable intermolecular interactions. Nevertheless, we speculate that apolar interactions between residues in the antiparallel 1B segments could also play a role in stabilizing the A 11 alignment mode. Interestingly, substitution of the Arg 10 residue in the 1A rod domain segment of especially type I keratins often results in a very serious phenotype in a variety of keratinopathy diseases (recently reviewed in Ref. 18). The molecular basis of the consequence of this substitution on keratin IF structure has not heretofore been determined, although one report (23) suggested the problem occurred at a structural hierarchical level above the stability of a single molecule. Our present data indicate in a straightforward way that this substitution causes a serious problem at the level of the two molecule stage of IF assembly, in particular by destabilizing the A 11 alignment mode.