Complementary Roles of Specific Cysteines in Keratin 14 toward the Assembly, Organization, and Dynamics of Intermediate Filaments in Skin Keratinocytes*

Background: The functional importance of inter-keratin disulfide bonding is underappreciated in epithelial biology. Results: Cysteines 4 and 40 in K14 participate in intermolecular disulfide bonding. Conclusion: Residues Cys-4, Cys-40, and Cys-367 in K14 and their dependent disulfide bonds regulate the assembly, organization, and dynamics of keratin filaments. Significance: Inter-keratin disulfide bonding plays an important role in regulating keratin intermediate filaments in skin keratinocytes. We recently showed that inter-keratin disulfide bonding plays an important role in the assembly, organization, and dynamics of keratin intermediate filaments in skin keratinocytes. In particular, cysteine 367 located in the central α-helical rod domain of keratin 14 is necessary for the formation of a stable perinuclear network of keratin filaments (with type II partner keratin 5) in skin keratinocytes analyzed by static and live cell imaging. Here, we show that two additional cysteine residues located in the non-helical head domain of K14, Cys-4 and Cys-40, also participate in inter-keratin disulfide bonding and tandemly play a key role complementary to that of Cys-367 in the assembly, organization, and dynamics of keratin filaments in skin keratinocytes in primary culture. Analysis of K14 variants with single or multiple substitutions of cysteine residues points to a spatial and temporal hierarchy in how Cys-4/Cys-40 and Cys-367 regulate keratin assembly in vitro and filament dynamics in live keratinocytes in culture. Our findings substantiate the importance and complexity of a novel determinant, namely inter-keratin disulfide bonding, for the regulation of several aspects of keratin filaments in surface epithelia.

We recently showed that inter-keratin disulfide bonding plays an important role in the assembly, organization, and dynamics of keratin intermediate filaments in skin keratinocytes. In particular, cysteine 367 located in the central ␣-helical rod domain of keratin 14 is necessary for the formation of a stable perinuclear network of keratin filaments (with type II partner keratin 5) in skin keratinocytes analyzed by static and live cell imaging. Here, we show that two additional cysteine residues located in the non-helical head domain of K14, Cys-4 and Cys-40, also participate in inter-keratin disulfide bonding and tandemly play a key role complementary to that of Cys-367 in the assembly, organization, and dynamics of keratin filaments in skin keratinocytes in primary culture. Analysis of K14 variants with single or multiple substitutions of cysteine residues points to a spatial and temporal hierarchy in how Cys-4/Cys-40 and Cys-367 regulate keratin assembly in vitro and filament dynamics in live keratinocytes in culture. Our findings substantiate the importance and complexity of a novel determinant, namely inter-keratin disulfide bonding, for the regulation of several aspects of keratin filaments in surface epithelia. 3 figure prominently in the cytoplasm of keratinocytes in epidermis and related epithelia (1), where they impact cytoarchitecture and are essential to the maintenance of physical integrity (2,3). Accordingly, disruption of the assembly or organization of keratin filaments (e.g. through mutation) results in cell and tissue fragility and underlies several inherited blistering diseases (2)(3)(4). Although individually rare, these diseases are often devastating, and there is unfortunately no effective treatment for disorders rooted in defective IFs (5,6). In part, the inability to therapeutically intervene reflects a limited understanding of the assembly and detailed structure of all types of intermediate filaments and of the mechanisms defining their intracellular organization and dynamics.

Ten-nanometer keratin intermediate filaments (IFs)
Although highly variable in size (40 -240 kDa) and primary structure, all IF proteins share a tripartite domain structure, consisting of a central ␣-helical (rod) domain of conserved length and substructure, flanked by non-helical N-terminal head and C-terminal tail domains of variable length and primary structure (7,8). IF assembly is driven largely by the central rod domain, given long-range heptad repeats mediating coiledcoil dimerization and higher order interactions, with variable contributions from the head and tail domains (depending of the IF system), notably with regard to post-translational modifications and interactions with associated proteins (9). In vivo, IF proteins assemble into 10-nm filaments either as obligate (e.g. keratins, neurofilaments) or facultative (e.g. vimentin, glial fibrillary acidic protein) heteropolymers (8). IF assembly appears complex and even polymorphic but can be readily recapitulated from purified proteins in vitro (10 -12). Further, live imaging studies over the last decade suggest that, in cultured epithelial cells, keratin IF assembly is initiated at the cell periphery and is completed as filament precursors become incorporated into a network that continuously flows toward the center of the cell, culminating in the formation of a robust perinuclear array of filaments (13)(14)(15). Despite this progress, much remains to be learned about the processes that preside over keratin assembly and network organization, in vitro as well as in vivo.
Disulfide bonding recently came to the fore as a novel and important mechanism that regulates the assembly, organization, and dynamics of keratin 14 (a type I keratin expressed in the progenitor basal layer of stratified epithelia)-containing intermediate filaments in skin keratinocytes (16,17). Particu-larly, residue Cys-367 in the central ␣-helical rod domain of human K14 mediates a trans-dimer homotypic disulfide bond in a crystal of the interacting K5/K14 rod domains as well as in epidermal keratinocytes in vivo (16). This cysteine and its associated disulfide bonds play a key role in the formation and/or maintenance of a perinuclear-concentrated network of keratin IFs and in the centripetal keratin filament cycle observed in epidermal keratinocytes in primary culture (17). The data in hand also show that, albeit important, Cys-367 does not account for all of the disulfide bonding-dependent properties of K14 in vitro and in vivo. In this study, we uncover complementary roles for two additional cysteine residues located in the non-helical head domain of K14, Cys-4 and Cys-40. As is the case for Cys-367, K14 Cys-4 and Cys-40 play key roles toward both the assembly and organization of keratin IFs. Further, we present evidence that all three cysteine residues act in concert to foster the execution of a normal keratin cycle in live cultures of epidermal keratinocytes.
Antibodies and Other Reagents-The antibodies used include anti-K5 (rabbit, 1:1000 for immunofluorescence) from Covance, anti-GFP (mouse, 1:1000) from Abcam, IRDye 800CW goat anti-mouse IgG (1:20,000) from LI-COR Biosciences, and Alexa Fluor 594 goat anti-rabbit IgG (1:1000) from Life Technologies, Inc. DAPI was from Life Technologies, tris(2-carboxyethyl) phosphine (TCEP) was from Thermo Scientific, and N-ethylmaleimide was from Sigma-Aldrich. Dulbecco's modified Eagle's medium (DMEM)/calcium-free/low glucose and DMEM/calcium-free/low glucose/without phenyl red, used in the preparation of flavin adenine dinucleotide (FAD) medium and FAD imaging medium, were from United States Biological. Restriction endonucleases were from New England Biolabs. All other chemicals were obtained from Sigma-Aldrich unless noted otherwise.
Cell Culture, Transfection, and Calcium Switch Assays-308 mouse skin keratinocytes (18) were used to identify cysteine residues participating in disulfide bonding in the K14 protein. Cells were cultured in DMEM/low glucose supplemented with 10% FBS, 100 units/ml penicillin, and 100 g/ml streptomycin (Life Technologies). Transfection of GFP-tagged K14 cysteine variants was performed using FuGENE HD (Promega) according to the manufacturer's instructions. At 24 h after transfection, cells were switched to Keratinocyte Basal Medium-2 without Ca 2ϩ (KBM-2, Lonza; so-called low calcium medium) for an additional 24 h and then switched to KBM-2 supplemented with 1.8 mM calcium and 10% FBS (high-calcium medium). Keratinocytes were harvested 4 h after high-calcium switch.
For immunoblotting, the Bradford protein assay (Bio-Rad) was used to determine protein concentration in cell lysates. Samples for gel electrophoresis were prepared in LDS sample buffer (Life Technologies) in the presence or absence of 20 mM TCEP. TCEP treatment was performed at room temperature for 1 h in order to fully reduce disulfide bonds. Equal amounts of cell lysates were resolved by SDS-PAGE and transferred to nitrocellulose membrane (0.45 m; Bio-Rad). Quantitative infrared immunoblotting analysis (LI-COR Biosciences) was performed using primary antibodies, followed by secondary antibodies conjugated to infrared fluorescent dye IRDye800 (1).
Krt14 Ϫ/Ϫ Mouse Keratinocytes in Primary Culture and Transfection Assays-All studies involving mice were reviewed and approved by the Johns Hopkins Institutional Animal Care and Use Committee. Freshly isolated keratinocytes from 1-or 2-day-old Krt14 Ϫ/Ϫ newborn mouse skin (19) were transfected by the pBK-CMV His-GFP-K14 cysteine variants using P1 primary cell 4D-Nucleofection X kit (Lonza). CnT57 (low calcium, 0.07 mM; CELLnTEC) and FAD media were used for keratin IF organization and live cell imaging studies, respectively (17). Transfected cells were fixed using prechilled methanol at 72 h post-transfection and processed for immunofluorescence staining to assess the organization of keratin IFs. Imaging of live keratinocytes was initiated at 48 h post-transfection. pShuttle-CMV-H2BmCherry was co-transfected in order to monitor the nucleus in transfected cells.
Immunofluorescence Microscopy-Keratinocytes were fixed by prechilled methanol; blocked in 10% normal goat serum, 0.1% Triton X-100, PBS for 1 h at room temperature; incubated in anti-K5 antibody solution for 1 h; washed in PBS; incubated in Alexa Fluor 594 goat anti-rabbit IgG solution; counterstained in DAPI; and mounted in FluorSave Reagent mounting medium (Calbiochem). Pictures were acquired as single focal planes using an Axio Observer.Z1 fluorescence microscope equipped with an ApoTome II attachment and AxioCam MRm camera (Carl Zeiss Microscopy). Focal planes were selected by focusing on the center of the nuclei. Pictures were taken using a 63ϫ plan-apochromat oil immersion objective, with the system controlled by the AxioVision version 4.8.2 software (Carl Zeiss Microscopy). Distinct modes of keratin filament network organization were analyzed by blindly counting transfected mouse keratinocytes and re-examined based on quantitative criteria for each of the three categories of keratin filament network organization (see Fig. S3 in Ref. 17).
Live Cell Imaging and Image Processing-Imaging of live keratinocytes and associated data processing were as described (17). A single-point laser scanning confocal microscope (LSM780-FCS, Carl Zeiss Microscopy), fitted with a 40ϫ Plan-Neofluar oil immersion objective, was used to record pictures. Cells in a closed culture dish (35 mm; MatTek Corp.) were kept in an environment chamber at 37°C with 5% CO 2 and 25°C humidity. Excitation was set at 488 nm for GFP and 561 nm for mCherry. Recording intervals were 5 min, color depth was 12 bit, and image resolution was 1024 ϫ 1024 pixels. Seven focal planes (optical slice thickness: 0.43 m/stack) were recorded for each time point, and the movie recording lasted 6 h. Maximum intensity projections were generated by compressing three-dimensional stacks of time series using the Zen software (Carl Zeiss Microscopy) and used for preparing figures and conversion into QuickTime movies (Apple). Reconstruction of 2.5-dimentional images from Z stacks was achieved using the Zen software. Measurements of filament length and of the fluorescence intensity of individual keratinocytes were performed using the ImageJ freeware (20). Tracking of nuclei was performed using the Imaris software (Bitplane).
Polymerization Efficiency and Morphological Analysis of Keratin Filament Assemblies-Physical states of the keratin polymers were analyzed using a high-speed sedimentation assay (150,000 ϫ g for 30 min) as described previously (22). Polymerization efficiency of filament assembly was estimated on SDS-PAGE through densitometry with the ImageJ freeware (20). The ultrastructure of assembled keratin filaments was examined by negative staining (1% uranyl acetate) and transmission electron microscopy using HU-12A instruments (Hitachi). Measurements of filament length and diameter were performed on recorded micrographs (80,000ϫ nominal magnification) using the AmtV602 capture engine software (Hitachi) and ImageJ freeware (20).
Statistical Analyses-Statistical analyses were performed using Microsoft Excel software 2011 (Microsoft Office) and Prism version 5 (GraphPad Software, Inc.). Student's t test analysis was used to compare two conditions or samples (see Figs. 1C, 5D, and 6 (A, C, and E)). One-way analysis of variance was used to compare three conditions or samples (see Figs. 2C and 3C). A statistical p value of Ͻ0.05 was considered significant.

Residues Cys-4 and Cys-40 in the K14 Head Domain Participate in Intermolecular Disulfide Bonding in Keratinocytes-
Human K14 contains five cysteine residues, with Cys-4, Cys-18, and Cys-40 located in its N-terminal head domain and Cys-367 and Cys-389 in its coiled-coil-forming rod domain (Fig. 1A). Previously (16), we showed that replacing Cys with Ala at position 367 (K14C367A variant) decreased but did not eliminate disulfide-bonded K14 species in transfected 308 mouse keratinocytes, suggesting that another cysteine(s) in the K14 head is also involved in intermolecular disulfide bonding (also see Ref. 17). We generated a series of variants by replacing cysteine residues with alanines at specific positions within the K14 protein (e.g. K14C4A/C367A, K14C18A/C367A, and K14C40A/C367A (Table 1)) in order to examine which cysteine(s) in the K14 head domain participate(s) in disulfide bonding and may play functional roles in keratin filaments. Upon expression in 308 mouse keratinocytes followed by immunoblotting analysis of total protein extracts, the GFP-K14C4A/C367A and GFP-K14C40A/ C367A variants result in a significant decrease in the yield of disulfide-bonded K14 species, as is the case for GFP-K14C367A, relative to control GFP-K14WT (Fig. 1, B and C). In addition, the triple cysteine variant GFP-K14C4A/C40A/ C367A yields no disulfide-bonded species of molecular masses greater than monomer size (same as the GFP-K14CF variant; Fig. 1C), suggesting that Cys-4, Cys-40, and Cys-367 in K14 are involved in intermolecular disulfide bonding in mouse skin keratinocytes.

Residues Cys-4 and Cys-40 in K14 Together Contribute to the Formation and/or Maintenance of a Perinuclear Concentrated
Network of Keratin Filaments-We recently showed that replacing Cys with Ala at position 367 (K14C367A) interferes with the formation and/or maintenance of perinuclear IF networks in transfected Krt14 Ϫ/Ϫ mouse keratinocytes in primary culture (17). Conversely, the presence of a single cysteine at position 367 in a completely cysteine-free K14 backbone (K14CF-C367) partially rescues this phenotype (17). To test whether K14 Cys-4 and Cys-40 also contribute to generating stable perinuclear IFs, we analyzed primary cultures of Krt14 Ϫ/Ϫ mouse keratinocytes transfected with either the GFP-K14C4A or GFP-K14C40A variant. Such transfected cells show a significant decrease in the frequency of perinuclear IF networks (seen in only 21 and 24% of cells expressing GFP-K14C4A and GFP-K14C40A, respectively; Fig. 2), similar to GFP-K14CF-expressing cells (Fig. 2C). Thus, the loss of a cysteine and/or disulfide bonding at either position 4 or 40 in K14 interferes with formation and/or maintenance of a perinuclear network, even in the presence of Cys-367. Conversely, the selective loss of Cys-367 in K14 also impairs the perinuclear organization of keratin IFs (even with Cys residues present at positions 4 and 40; see Ref. 17). Altogether, this suggests that Cys-4, Cys-40, and Cys-367 play important but potentially distinct roles in this regard.  Compared with GFP-K14WT-expressing cells, Krt14 Ϫ/Ϫ keratinocytes transfected with GFP-K14C18A or GFP-K14C389A show a very modest decrease in the frequency of perinuclear IF networks (ϳ15 and ϳ12% reduction, respectively, p Ͻ 0.01; Fig. 2, B and C), suggesting that K14 Cys-18 and Cys-389 may play a modest role in the genesis of the perinuclear network of keratin IFs although they are not apparently involved in intermolecular disulfide bonding (Fig. 1B). The properties of these two Cys variants have not been further analyzed in this study.
Additional variants were produced by site-directed mutagenesis at positions Cys-4 and Cys-40, either singly or in combination, in either the wild type (K14) or cysteine-free K14 backbone (Table 1), and their properties were assessed after transfection in Krt14 Ϫ/Ϫ mouse keratinocytes in primary culture. The main mode of keratin organization in either GFP-K14CF-C4-or GFP-K14CF-C40-expressing cells is pan-cytoplasmic (showing 60 and 62% frequencies, respectively; Fig. 3), whereas, as expected (17), the majority (62%) of GFP-K14WTexpressing cells show perinuclear concentrated networks (see Fig. 3, B and C). This outcome is similar to that observed in cells expressing the single replacement variants GFP-K14C4A or GFP-K14C40A (given 61 and 59%, respectively, of cells showing pan-cytoplasmic networks; Fig. 2B) and in cells expressing the double-replacement variant GFP-K14C4A/C40A (61% of cells; Fig. 3, B and C). By comparison, restoring both Cys-4 and   Cys-40 in the K14CF backbone leads to 51% of cells exhibiting a perinuclear concentrated array of keratin filaments (Fig. 3), similar to what is seen for GFP-K14CF-C367-expressing cells (49% of cells; Fig. 3, B and C). Together, these findings suggest that the presence of both Cys-4 and Cys-40 in the K14 head domain is necessary and sufficient to foster the formation and/or maintenance of a perinuclear network of keratin filaments in the absence of Cys-367. Because both sets of cysteines, Cys-4 and Cys-40, in the K14 head and Cys-367 in the K14 rod can each foster the formation of a perinuclear network of keratin filaments, we next tested whether they function synergistically or separately to this end. A new variant was generated by restoring these three cysteines in the K14CF backbone (K14CF-C4/C40/C367) and then analyzed by transfection in Krt14 Ϫ/Ϫ mouse keratinocytes. Relative to GFP-K14CF-C4/C40 and GFP-K14CF-C367 (51 and 49% frequencies, respectively), expression of the triple-cysteine-restoring variant GFP-K14CF-C4/C40/C367 showed no further improvement in terms of yielding perinuclear concentrated keratin networks (47% of cells; Fig. 3). This outcome suggests that Cys-4 and Cys-40 in the K14 head, on the one hand, and Cys-367 in the K14 rod, on the other, may function at different stages during the genesis of a perinuclear concentrated IF network. We also find that expression of cysteine-restoring variants, including the triple variant GFP-K14CF-C4/C40/C367, does not completely replicate the ability of GFP-K14WT to rescue perinuclear IF organization in Krt14 Ϫ/Ϫ mouse keratinocytes (47% versus 62%; p Ͻ 0.01; Fig. 3, B and C), providing additional evidence that the other cysteines in K14 (Cys-18 and Cys-389) make a contribution to this phenomenon.

Formation of K14CF-C4/C40-mediated Inter-disulfide Bonds Plays an Important Role in the Assembly of Keratin Filaments in
Vitro-We recently reported that the K14CF-C367 variant (having a single Cys residue at position 367) yielded very aberrant IFs having abnormally large and variable diameters (8 -27 nm) when co-assembled with K5 in vitro (17). Given this, we next tested whether the formation of Cys-4 and Cys-40-dependent disulfide bonds plays a role in 10-nm filament assembly in vitro. Under standard polymerization buffer conditions (in the presence of a reducing agent) and with K5 as the type II pairing partner, the K14CF-C4/C40 variant shows a high assembly efficiency (ϳ98%) (Fig. 4, A and B) and forms normal 10-nm filaments, as seen by electron microscopy (data not shown). Under an oxidation-prone environment, however, the K5/K14CF-C4/ C40 pairing yields a subnormal assembly efficiency (ϳ30% reduction) and filaments that are only short to intermediate in length (0.2-1-m range) but display a normal and fairly uniform diameter (9 -11 nm) (Fig. 4). Unlike the case for Cys-367, therefore, formation of disulfide bonds involving Cys-4 and/or Cys-40 does not interfere with either the control of subunit number per filament or their compaction or maturation during assembly in vitro.
Restoring Cys-4, Cys-40, and Cys-367 in the K14CF backbone (K14CF-C4/C40/C367 variant) also yields subnormal assembly efficiency along with filaments that are quite short after co-polymerization with K5 in vitro under oxidizing buffer conditions (Fig. 4). This is largely similar to the properties of K5/K14WT assemblies (Fig. 4, A and B; also see Ref. 17). This outcome suggests that the inhibition of filament elongation under oxidizing conditions is due to formation of disulfide bonds predominantly mediated by Cys-4, Cys-40, and Cys-367 in K14. Remarkably, these K5/K14CF-C4/C40/C367 filaments display a normal diameter (9 -12 nm; see Fig. 4C). Thus, restoring Cys-4 and Cys-40 in addition to Cys-367 in the K14CF backbone results in the nearly complete normalization of the diameter in the resulting filaments, suggesting that the intermolecular disulfide bonds mediated by Cys-4 and Cys-40 act in a dominant fashion over those mediated by Cys-367 to regulate the number of subunits per filament cross-section or to foster filament compaction and/or maturation in vitro.

K14 Cys-4 and Cys-40 Foster a Normal Keratin Filament Cycle in Live Mouse
Keratinocytes-We reported that, upon its transfection in Krt14 Ϫ/Ϫ mouse keratinocytes, GFP-K14WT behaves in a manner consistent with the keratin cycle originally described by Leube and colleagues (13)(14)(15). In brief, short filaments first form at the cell periphery, elongate and integrate into the keratin IF network as they start moving toward the centrally located nucleus, and finally become part of the perinuclear concentrated network (17). Loss of all K14-bound cysteines and/or disulfide bonds or even the single loss of Cys-367 impairs this normal keratin filament cycle and disrupts perinuclear IF organization in live keratinocytes. This said, restoration of a single cysteine residue at position 367 in the K14CF backbone incompletely rescues the keratin filament cycle and perinuclear IF organization (45% rescue efficiency, as opposed to 59% when testing K14WT; see Ref. 17). Such findings provide further support for the notion that cysteine(s) other than Cys-367 in K14 participate in these cellular processes.
To examine this possibility, we transfected Krt14 Ϫ/Ϫ mouse keratinocytes with GFP-K14CF-C4/C40 and subjected them to live cell recordings. Two types of outcome were observed, irrespective of transgene expression levels (Fig. 6A). Most transfected cells (ϳ80%) exhibit an impaired keratin filament cycle and fail to form a perinuclear concentrated IF network (supplemental Movie 1). However, a small population of cells (ϳ20%) displays normal keratin filament cycling and formation of a perinuclear filament network (supplemental Movie 2 and Fig. 5,  A and B), similar to GFP-K14WT-transfected cells (supplemental Movie 3). Comparing non-rescued keratinocytes expressing either GFP-K14CF-C367 or GFP-K14CF-C4/C40, two differences stood out: (i) the occurrence of long (Cys-367) versus short (Cys-4/Cys-40) filaments at the cell periphery (a location where keratins are more likely to be seen as single filaments) (13,15) and (ii) filaments showing uniform (Cys-367) versus spotted (Cys-4/Cys-40) fluorescence intensity in the cytoplasm (Fig. 5, C and D). Accordingly, although both sets of cysteines foster the establishment of a normal keratin filament cycle, it appears that they function at different stages, with Cys-4 and Cys-40 involved early on (e.g. during nucleation and/or elongation) and Cys-367 playing a more defining role at later stages (e.g. when the perinuclear network of filaments is being formed; see "Discussion"). Besides, we observed that 85% of GFP-K14CF-C4/C40-expressing keratinocytes fail to exhibit the aberrant cell motility behavior characteristic of Krt14 Ϫ/Ϫ keratinocytes (Fig. 6, B and E; note that 90% rescue in this behavior was reported for GFP-K14CF-C367 in Ref. 17), suggesting that residues Cys-4, Cys-40, and Cys-367 in K14 contribute to regulating keratinocyte motile properties.
Next, we expressed the GFP-K14CF-C4/C40/C367 variant in Krt14 Ϫ/Ϫ mouse keratinocytes and performed live cell imaging to assess the combined contributions of Cys-4, Cys-40, and Cys-367 to the keratin filament cycle. Greater than 70% of cells expressing GFP-K14CF-C4/C40/C367 display a normal keratin cycle, culminating in the formation of a perinuclear concentrated network, whereas only ϳ30% of cells exhibit an impaired keratin cycle and failure to form a perinuclear keratin network (data not shown). Analysis of fluorescence intensity suggests that this difference in outcome is not rooted in transgene expression levels in individual cells (Fig. 6C). Whereas this rescue efficiency is superior to that seen for the GFP-K14CF-C367 (45%) (17) and GFP-K14CF-C4/C40 variants (20%) (this study), these findings indicate that residues Cys-4, Cys-40, and Cys-367 in K14 are necessary but not sufficient to promote a completely normal keratin cycle and formation of a stable perinuclear network. Also of interest, the majority of Krt14 Ϫ/Ϫ keratinocytes expressing GFP-K14CF-C4/C40/C367 exhibit parallel arrays of prominent keratin filament bundles apposed against the nucleus (supplemen-talMovie4 and Fig. 5E). Many factors could account for enhanced filament bundling in that region of the cell, including but not limited to altered disulfide bonding between keratin filaments and/or keratins and interacting proteins and/or reduced turnover of keratin filaments in the perinuclear region. Besides, 92% of GFP-K14CF-C4/C40/C367-expressing Krt14 Ϫ/Ϫ keratinocytes show a marked normalization of the abnormal cell motility exhibited by Krt14 Ϫ/Ϫ keratinocytes (Fig. 6, D and E), consistent with observations made when analyzing GFP-K14CF-C367-and GFP-K14CF-C4/C40-expressing cells (17) (this study) and further supporting the notion that cysteines 4, 40, and 367 in K14 contribute to regulating aspects of keratinocyte motility.

Discussion
Human K14 contains five cysteine residues that are perfectly conserved in its mammalian orthologs and in the highly homologous human K16 (17). Additionally, select cysteines (e.g. Cys-367) are conserved in other type I keratin proteins expressed in skin epithelia (e.g. K10 and K17). Building on two previous reports (16,17), we show here that three of the five cysteine residues occurring in K14 are extensively involved in interkeratin disulfide bonding, in a manner that is readily detectable when using a combination of transfection of reporter constructs in cultured keratinocytes and immunoblotting analyses of protein extracts. These three residues are Cys-4 and Cys-40 (this study), located in the head domain of K14, and Cys-367 (17), located within a stutter segment in the 2B portion of the central ␣-helical rod domain (16). All three cysteines play significant roles (i) during 10-nm filament assembly in vitro; (ii) in fostering the formation of a perinuclear network of keratin filaments in cultured skin keratinocytes; iii) during the centripetal keratin filament cycle that typifies epithelial cells in culture. Further, our findings highlight clear differences in the contributions of Cys-4 and Cys-40, on the one hand, and Cys-367, on the other, to these processes and hint at the existence of a spatial and temporal hierarchy with regard to the involvement of these residues and their disulfide-bonding capabilities.
A popular model in the field conveys that, as IF proteins polymerize, short 100-nm-long (unit-length) filaments (ULFs) first form and then anneal to form long but wide (16-nm) filaments, which then undergo collapse or compaction in their cross-sectional axis to give rise to mature 10-nm filaments (8). If this model indeed applies to keratin assemblies in their native context, then inter-keratin (inter-subunit?) disulfide bonding may function as a significant regulatory step during keratin assembly. The combination of in vitro and transfection-based findings amassed to date lead us to propose a model, schematized in Fig. 7, that also takes into account differences in redox potential across the cell (23,24). We propose that K14 Cys-4 and Cys-40 make a significant contribution to the regulation of filament elongation, specifically the annealing of ULFs, and this selectively takes place at sites of new filament formation. In epithelial cells in culture, such sites preferentially occur at the cell periphery (Fig. 7) (see Ref. 25), where the redox environment is reportedly oxidizing (23). Formation of Cys-4-and Cys-40-dependent disulfide bonds would inhibit the annealing of ULFs but also protect against their lateral growth beyond ϳ10 nm, thus contributing to define the number and/or sites of filaments to be further elongated (26). Movement of ULFs toward the center of the cell (e.g. in the context of the keratin cycle) would entail exposure to a more reducing environment (27), thereby reducing Cys-4-and Cys-40-dependent disulfide bonds and fostering the annealing of ULFs and the formation of long 10-nm filaments. As cycling filaments reach the more oxidizing environment prevailing in the perinuclear region (23,28), Cys-367-dependent disulfide bonding, in particular, would play a significant role in the regulation of filament-filament interactions and dynamics (Fig. 7) (16, 17). The model should also account for the fact that Cys-367 is also influential during  the annealing of ULFs, although less so than Cys-4/Cys-40 (17) (this study). This model is obviously largely conceptual at this stage. Major differences in the contribution of disulfide bonding-forming cysteines are to be expected, depending, for instance, on the type of cell, the density and distribution of cysteine residues in the IF proteins involved, and the prevailing biological context. Much experimentation along with better insight into the high-resolution structure of keratin filaments is needed to test these ideas. Formation of intramolecular disulfide bonds in K14, either in vitro or in vivo, cannot be ruled out, given the limits of the methods and assays used in our study. Formation of intramolecular disulfide bonding facilitates the correct folding and contributes to the stability of proteins (29,30). Intramolecular disulfide bonding is unlikely for Cys-389, given that its side chain faces inward in the coiled-coiled interface between K5 and K14 (16). The finding that expression of K14C389A has a modest but consistent impact on perinuclear keratin networks thus serves as a reminder that some of the cysteine-to-alanine substitutions introduced may have an impact on the local structure of the K14 protein. Likewise, the rationale for the similarly modest impact of the K14C18A variant on the genesis of a perinuclear network awaits investigation. Also of significance, we cannot account for contributions by other type I keratins, K16 and K17 in particular, to the behaviors observed in keratinocytes transfected with various K14 cysteine variants.
Two glaring questions arise when the findings reported here are combined with previous studies (16,17). What are the mechanisms responsible for disulfide bond formation in keratinocytes? And what is their functional importance in vivo? Mechanistically, disulfide bonds could form enzymatically or non-enzymatically, and there is evidence that both can occur in the cytoplasm of keratinocytes. Sulfhydryl oxidases that are involved in catalyzing the formation of disulfide bonds in the late differentiating layers of epidermis (31, 32) also occur in the basal layer of mouse epidermis (both mRNA and protein products were detected in mouse skin keratinocytes in primary culture). 4 In addition, changes in redox potential and/or intracellular oxidative stress were recently implicated in terminal differentiation of epidermal keratinocytes, providing another likely mechanism for the formation of keratin disulfide bonding (33)(34)(35). Regarding the issue of physiological relevance, our findings show that the disulfide bonding-dependent formation of a K14-containing perinuclear network of keratin filaments is important for stably positioning the nucleus in keratinocytes in culture (this study) (17) and also impacts the size and shape of the nucleus (16). Whether any of this plays out in the epidermis in situ and the associated significance are unknown. Whereas there is no report of disease-causing mutation affecting any of the K14 cysteine residues (4, 6), a G61C variant in the head domain of K8 predisposes to liver injury (36,37), also the R125C mutation in K14 (38) is a recurring allele in K14 (4, 6, 39) and other type I keratins and is associated with severe disease presentations. To best address this issue, select K14 cysteine variants (e.g. C4A, C40A, and C367A alone or in combination) should be knocked in at the Krt14 locus in the mouse genome in order to ascertain the functional relevance of the findings reported so far.
Author Contributions-X. F. and P. A. C. designed the research. X. F. performed the experiments. X. F. and P. A. C. analyzed data and wrote the paper.

Cell periphery Perinuclear region Cytoplasm
C4 and C40 and their dependent disulfide bonds are involved at early stages of the keratin cycle (e.g. nucleation and elongation) to regulate filament elongation, fostering the formation of short filaments and regulating lateral contacts between subunits C367 and its dependent disulfide bonds are involved at later stages of the keratin cycle to regulate filament organization (possibly filament-filament interactions) and dynamics (genesis and/or maintenance of perinuclear IF networks) FIGURE 7. Model suggesting roles for K14-dependent disulfide bonding toward the organization and dynamics of keratin IFs in skin keratinocytes. We propose that Cys-4 and Cys-40 and their dependent disulfide bonding make a significant contribution to regulating the early stages of filament assembly (e.g. nucleation and elongation), whereas Cys-367 and its dependent disulfide bonding play a significant role in the genesis and/or maintenance of perinuclear IF networks as part of the keratin filament cycle (15). Pink and blue shades represent regions of relatively oxidizing and reducing environment within the cell, respectively, based on evidence from the literature (23). This schematic is inspired and adapted from Ref. 15.