Proc. Natl. Acad. Sci. USA Vol. 95, pp. 2067–2072, March 1998 Biochemistry

A highly conserved lysine residue on the head domain of type II is essential for the attachment of intermediate filaments to the cornified cell envelope through isopeptide crosslinking by transglutaminases

ELEONORA CANDI*, EDIT TARCSA*, JOHN J. DIGIOVANNA*†,JOHN G. COMPTON*, PETER M. ELIAS‡, LYUBEN N. MAREKOV*, AND PETER M. STEINERT*§

*Laboratory of Biology, National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health, Bethesda, MD 20892-2752; and ‡Department of Dermatology, Veterans Administration Medical Center, San Francisco, CA 94121

Communicated by Laszlo Lorand, Northwestern University Medical School, Chicago, IL, December 12, 1997 (received for review July 14, 1997)

ABSTRACT We have addressed the question of how keratin nocyte becomes a terminally differentiated corneocyte, the struc- intermediate filaments are associated with the cell envelope at tural integrity of desmosomes decreases and, indeed, most the periphery of cornified epidermal cells. Many peptides from epitopes of such as are no longer detectable human epidermal cell envelopes containing isopeptide crosslinks as a result of degradation or masking (11). Thus, the question inserted by transglutaminases in vivo have been characterized. A arises as to how the KIF is mechanically integrated major subset involves the type II keratin chains , 2e, 5, with the CE after the major cellular remodeling events involved or 6 crosslinked to several partners through a lysine in terminal differentiation. residue located in a conserved region of the V1 subdomain of We have characterized the protein composition of CEs their head domains. This sequence specificity was confirmed in obtained from foreskin epidermal corneocytes and have shown in vitro crosslinking experiments. Previously the causative mu- that a variety of structural proteins are crosslinked together by tation in a family with diffuse nonepidermolytic palmar-plantar the action of transglutaminases (TGases) to form the CE (refs. keratoderma was shown to be the loss in one allele of the same 12–15; P.M.S. and L.N.M., unpublished data). TGases form lysine residue of the keratin 1 chain. Ultrastructural studies of N␧-(␥-glutamyl)lysine crosslinks between protein-bound Gln affected palm have revealed abnormalities in the and Lys residues, which result in stable, permanent macromo- organization of keratin filaments subjacent to the cell envelope lecular assemblies (1–3). The major proteins are loricrin and in the shape of the cornified cells. Together, these data admixed with lesser amounts of small proline-rich (SPR) suggest a mechanism for the coordination of cornified cell proteins and elafin (14). These proteins constitute about 80% structure by permanent covalent attachment of the keratin of CE protein and occupy the most intracellular aspect of the cytoskeleton to the cell envelope by trans- CE, that is, that which corresponds to the later stages of CE glutaminase crosslinking. Furthermore, these studies identify assembly. After saponification of some of the protein-bound the essential role of a conserved lysine residue on the head from the extracellular surface of the CE, we iden- domains of type II keratins in the supramolecular organization tified by immunogold electron microscopy (15) and sequencing of keratin filaments in cells. (P.M.S. and L.N.M., unpublished data) certain proteins pre- sumably involved in the earlier stages of CE assembly, includ- The terminal differentiation program of stratified squamous ing and the carboxyl termini of desmoplakin and epithelia includes the formation of a cell envelope (CE) envoplakin. In addition, we identified many peptides involving structure on the intracellular surface of the plasma membrane, the keratin 1, 2e, 5, or 6 chains, crosslinked by isopeptide bonds which it eventually replaces (1–3). In the case of human to a variety of these proteins, both in the untreated and epidermis, the CE consists of an Ϸ15-nm-thick layer of insol- saponified CE samples. Similar peptides have been isolated uble crosslinked protein with an Ϸ5-nm-thick layer of attached and characterized from immature foreskin (3–7). This insoluble CE is vitally important in barrier (15), as well as from cultured keratinocytes induced to function for the tissue and organism (1–7). The bulk of the terminally differentiate in submerged serum-free cultures terminally differentiated epidermal corneocyte consists of a (P.M.S. and L.N.M., unpublished data). In this paper, we cytoskeletal keratin intermediate filament (KIF)– have characterized these keratin-containing crosslinked pep- complex that is retained within this CE structure (4). tides and report that KIF are crosslinked by isopeptide bonds In nucleated epidermal cells, the KIF cytoskeleton is mechan- to the CE primarily through a single Lys residue located on ically integrated with the cell periphery primarily at desmosomal the head domain of the type II keratin chains. The impor- junctions, which in turn provide structural continuity throughout tance of this anchorage of the KIF cytoskeleton to the CE is the tissue (8–11). Although the exact details remain unclear, the demonstrated in a family with Unna–Thost disease, autoso- connection of KIF to desmosomes involves a large number of mal dominant diffuse nonepidermolytic palmar-plantar ker- proteins, including the major intracellular desmosomal protein atoderma (NEPPK), in which a mutation involving loss of desmoplakin (8–11). However, the terminal differentiation pro- this essential lysine residue in one allele (16) results in gram in the epidermis also involves major changes in morphology, structural changes in the epidermis. These data identify a the formation of a -enriched barrier, loss of most ‘‘house- critical function of the end domains of the epidermal keratins keeping’’ components, including the cytoskeletal connections to in the supramolecular coordination of cell structure. the cell periphery, and controlled cell death (4). As the kerati- Abbreviations: KIF, keratin intermediate filaments; NEPPK, nonepi- The publication costs of this article were defrayed in part by page charge dermolytic palmar-plantar keratoderma; SPR, small proline-rich (class of proteins); TGase, transglutaminase; CE, cell envelope. payment. This article must therefore be hereby marked ‘‘advertisement’’ in †Present address: Division of Dermatopharmacology, Department of accordance with 18 U.S.C. §1734 solely to indicate this fact. Dermatology, Brown University, Providence, RI 02903. 0027-8424͞98͞952067-6$0.00͞0 §To whom reprint requests should be addressed. e-mail: pemast@ PNAS is available online at http:͞͞www.pnas.org. helix.nih.gov.

2067 Downloaded by guest on October 1, 2021 2068 Biochemistry: Candi et al. Proc. Natl. Acad. Sci. USA 95 (1998)

MATERIALS AND METHODS it were dissolved and mixed in a buffer of 8 M urea͞50 mM ⅐ ͞ ͞ Sequencing of CE Peptides. Experimental details for the Tris HCl, pH 7.6 1 mM EDTA 1 mM DTT and dialyzed into isolation, sequencing, sorting, and characterization of peptides KIF assembly buffer as above. KIF were examined by electron containing isopeptide crosslinks have been described from: microscopy after negative staining with 0.7% uranyl acetate. foreskin cornified tissue (14); immature foreskin epidermis In Vitro Transglutaminase Crosslinking Assays. A synthetic and saponified cornified tissue (15); and from 3-day and 7-day peptide of sequence Val-Ser-Ser-Gln-Gln-Val-Thr-Gln-Ser- cultured normal human epidermal keratinocytes grown in Cys-Ala, corresponding to residues 210–223 of human loricrin serum-free submerged cultures in the presence of high (1.2 (22), was synthesized, partially labeled by acetylation of its 3 mM) Ca2ϩ medium (P.M.S. and L.N.M., unpublished data). In amino terminus with H-labeled acetic anhydride (23), and this paper, only those crosslinks involving keratin chains are purified by HPLC, with a resultant specific activity of 83 ͞ discussed. dpm pmol. Activated guinea pig epidermal TGase 3 enzyme Expression of Wild-Type and Mutated Forms of (24) was used, and its specific activity was measured by and 14 Chains and Assembly of KIF in Vitro. Full-length incorporation of 3H-labeled putrescine into succinylated ca- human keratin 5 and 14 cDNAs were constructed by PCR sein (17). Analytical crosslinking reactions contained (in 100 amplification by using DNA from a human library ␮l reaction volume) 0.4 nmol of KIF containing the keratin 1 as template and confirmed by DNA sequencing. These were and 10 (1͞10) or 5͞14 chains, a 0.5- to 5-fold molar excess of assembled into the pET11a vector for protein expression in labeled loricrin peptide, in a buffer of 0.1 M Tris-acetate, pH ͞ ͞ bacteria (17). Two mutant forms of the keratin 5 chain (Lys-71 8.3 1mMDTT5 mM CaCl2, and sufficient enzyme to 3 Ile and Leu-174 3 Pro) were generated by using the incorporate 1 pmol͞min of putrescine into casein (10–20 pmol QuikChange site-directed mutagenesis kit (Stratagene). The of enzyme in different experiments). This was determined in keratins were expressed in very high yields. Bacterial pellets control assays to be sufficient enzyme to drive the crosslinking were dissolved in SDS gel buffer, and the proteins were reactions to completion. After1hat37°C, reactions were resolved on preparative PAGE gels. The desired keratin bands terminated by addition of EDTA (to 15 mM). Aliquots of 25 were cut out, eluted into SDS gel buffer overnight, freed of ␮l were spotted onto 3-mm filter papers, washed in trichloro- bound SDS by ion-pair solvent extraction (18), and finally acetic acid, and counted (17, 20). In preparative-scale exper- dissolved into buffer. We were unable to express the iments, 8 nmol of keratin 1͞10 KIF was mixed with 15 nmol of human keratin 1 chain, presumably because of toxicity of its the 3H-labeled loricrin peptide, reacted as above, and termi- glycine-loop sequences (19, 20). nated. The reactions were digested to completion within 3 h Keratin 1 and 10 proteins were extracted from human with trypsin (Sigma, bovine, sequencing grade, 1% wt͞wt). foreskin epidermis (21) and dialyzed through solutions of The tryptic digests were resolved by HPLC on a reverse-phase decreasing urea concentration into assembly buffer of 10 mM ultrasphere ODS column (4.5 ϫ 250 mm) at a flow rate of 1 Tris⅐HCl (pH 7.6) and 1 mM DTT to give a final protein ml͞min with a gradient of 0–100% acetonitrile containing concentration of 0.5–1 mg͞ml (Ϸ4–8 nmol͞ml of keratin 1 and 0.08% trifluoroacetic acid, collected into 0.5-min fractions, of 10 heterodimer) (21). Equimolar amounts of the expressed which half was counted for radioactivity. Aliquots of selected wild-type chain and either the wild-type keratin 5 fractions were covalently attached to a support (Sequelon-AA, chain, or the Lys-71 3 Ile or Leu-174 3 Pro mutant forms of Millipore) for sequencing (14, 15).

Table 1. Occurrences of proteins directly crosslinked to keratins CE experiment Number of number peptides Cystatin ␣ Elafin Envoplakin Involucrin Loricrin SPRs 1 and 3 Keratin 1 16 6 (70) 2 8 1 (15) 3 (50) 1 (40) 3 (250) 3 31 1 (10) 1 (10) 19 (240) 1 (50) 4 (600) 5 (190) 4 7 1 (10) 4 (80) 1 (110) 1 (25) 5 19 1 (15) 1 (15) 2 (25) 9 (140) 4 (220) 2 (40) Keratin 2e 11 1 (5) 2 4 2 (20) 1 (20) 1 (50) 3 3 1 (5) 1 (10) 1 (50) 41 1 (10) 52 2 (40) Keratin 5 (or 6) 10 2 12 7 (160) 4 (60) 1 (10) 3 33 1 (10) 2 (20) 22,1 (340) 1 (10) 5 (60) 1 (10) 4 67 1 (5) 1 (10) 34,3 (650) 21,3 (270) 1 (5) 2 (10) 1 (10) 5 55 2 (20) 3 (50) 18 (250) 14,3 (240) 8,1 (110) 5 (40) 1 (5) 11 1 (10) 22 2 (20) 31 1 (20) 40 5 3 1 (10) 2 (30) The numbers in parenthesis are yields in pmol. CE experiment numbers are: 1, foreskin (14); 2, saponified foreskin stratum corneum (15); 3, ‘‘immature’’ foreskin epidermis (15); 4, 3-day NHEK cultures (P.M.S. and L.N.M., unpublished data); 5, 7-day NHEK cultures (P.M.S. and L.N.M., unpublished data). In most cases, the short sequences of the keratin branch did not permit distinction between the keratin 5 or 6 chains. The second number listed defines known keratin 6 sequences; e.g., 22,1 (340) means that of a total of 23 peptides (totaling 340 pmol), 22 involved the keratin 5 or 6 chains and 1 was confirmed from keratin 6. Downloaded by guest on October 1, 2021 Biochemistry: Candi et al. Proc. Natl. Acad. Sci. USA 95 (1998) 2069

Table 2. Sequences involved in direct crosslinks with keratin chains Keratin chains yield, pmol Chain Sequence 12e5͞66 Keratin chain Keratin 1 (Lys-73) NLGGSKSISISVAR Keratin 2e (Lys-69) GLGGTKSISISVA Keratin 5 (Lys-71) NLGGSKRISISTR Keratin 6 (Lys-68) GLGGSKRISI Crosslinked partner Cystatin ␣ (Gln-45) TQVVAGTNY 25 35 Elafin (Gln-2) AQEPVKGPV 40 80 Envoplakin* (Gln-1970) AQLLQDESSFEKDL 235 25 600 40 Envoplakin (Gln-1973) AQLLQDESSFEKDL 80 460 30 Involucrin (Gln-88͞525) HLEQQEGQLK 120 5 110 Involucrin (Gln-288) YLEQQEGQLK 170 20 130 20 Involucrin (Gln-328) HLEQQEGQLEQL 20 5 110 5 Involucrin (Gln-436) HLEEQEGQLK 40 Involucrin (Gln-445) HLEQQQGQLEV 90 15 Involucrin (Gln-446) HLEQQQGQLEV 50 10 Involucrin (Gln-455) PEQQVGQPKNL 40 65 10 Involucrin (Gln-456) PEQQVGQPKNL 10 Loricrin (Gln-6) QKKQPTPQPPV 40 10 Loricrin (Gln-153) SGQAVQCQSY 70 10 Loricrin (Gln-215) YVSSQQVTQTSCA 570 105 65 Loricrin (Gln-216) YVSSQQVTQTSCA 360 20 20 Loricrin (Gln-219) SSQQVTQTSCA 180 20 Loricrin (Gln-303) CHQTQQKQA 30 Loricrin (Gln-305) HQTQQKQA 20 40 Loricrin (Gln-308) HQTQQKQTW 10 10 SPR1 (Gln-5) QQQKQPC 10 SPR1 or 3 (Gln-19) QQQQV 10 SPR1 or 3 (Gln-83) QQKTKQK 90 10 SPR1 or 3 (Gln-87) QQKTKQK 170 15 SPR3 (Gln-5) QQKQTF 10 SPR2 (Gln-6) QQQQCKQPCQPPPV 5 SPR2 (Gln-9) QQQQCKQPC 5 Single-letter code is used for amino acids. Residues involved in crosslinks are shown in bold. *Note that three envoplakin variants have been described (15), but the data are combined here because the same Gln residues were used.

Electron Microscopy Procedures. Punch biopsies (ca. 3mm keratin chains (Table 2). This Lys residue is present only in the diameter) were surgically removed from the previously un- keratins expressed in stratified squamous epithelia that form treated sides of the palms of one member of a previously CE structures (16). Only seven peptides involved either lysines described (16) family with diffuse NEPPK and from an age- 9 or 32 located in the E1 subdomain of the keratin 10 chain. and sex-matched donor with no known palmar-plantar involve- Analysis of the data shows that most crosslinks involved ment, fixed in glutaraldehyde, and washed by established keratins 5 and͞or 6 with the early CE protein components procedures (25). Samples were then postfixed with OsO4 and involucrin and envoplakin (ref. 15; P.M.S. and L.N.M., unpub- processed for electron microscopy (7, 25). lished data). Although the molar yield of each was relatively low, the total yield was about 60% of all keratin-containing crosslinks. RESULTS AND DISCUSSION Another 40% of the total yield involved the later CE protein Keratin Chains Are Crosslinked by Isopeptide Bonds in components elafin, SPRs, and loricrin, with some of the highest Vivo to Many CE Proteins Through a Common, Highly yields recovered for peptides having crosslinks between keratin 1 Conserved Lysine Residue. We have reported immunogold and three favored donor Gln residues of loricrin, Gln-215, localization and protein sequencing data that suggests that the Gln-216, and Gln-219 (20). The keratin 5 (and keratin 6 from CEs assembly of the CE of foreskin epidermis occurs in at least two made in cell cultures) chains are typically expressed in basal discrete steps (refs. 13–15; P.M.S. and L.N.M., unpublished epidermal cells, although they are retained into the differentiat- data). These involve first, the accretion at or near desmosomes ing layers (26), where keratins 1 and 2e are expressed in larger of early proteins such as involucrin, envoplakin (and variants), amounts (26, 27). and cystatin ␣, which together form a scaffold for the second The Conserved Lys-71 of the Keratin 5 Chain Is Not Essential step: the accumulation of late ‘‘reinforcement’’ proteins in- for Efficient KIF Assembly in Vitro. We expressed in bacteria the cluding elafin, SPRs, and large amounts of loricrin. During the wild-type keratin 5 and 14 chains, as well as the Lys-71 3 Ile and course of characterization of peptides from several epidermal Leu-174 3 Pro-substituted forms of keratin 5. After isolation CE preparations (refs. 14 and 15; P.M.S. and L.N.M., unpub- from a preparative SDS͞polyacrylamide gel, equimolar mixtures lished data), we identified 256 that included a keratin chain of keratin 14 and either the wild-type (Fig. 1a) or the Lys-71 3 directly crosslinked to one of these proteins (Table 1). Of Ile-substituted (Fig. 1b) keratin 5 chains assembled in vitro into Ͼ ␮ these, 249 (97%) involved a type II keratin, either keratins 1, long (Lav 20 m) KIF. This Ile substitution is comparable to 2e, 5, or 6. Significantly, all 249 peptides involved a crosslink that of the keratin 1 chain (Lys-73 3 Ile) seen in disease (16). formed through a Lys residue in a conserved sequence region These data suggest this Lys residue is not required for KIF of the head domain, encompassing the V1 subdomain, of the assembly in vitro. However, a negative control substitution of Downloaded by guest on October 1, 2021 2070 Biochemistry: Candi et al. Proc. Natl. Acad. Sci. USA 95 (1998)

FIG. 1. The conserved Lys residue 71 of the keratin 5 chain is not essential for KIF assembly in vitro. KIF were assembled from equimo- FIG. 3. Recovery by HPLC of tryptic peptides containing crosslinks lar mixtures of: wild-type keratins 5 and 14 (a); the Lys-71 3 introduced by the TGase 3 enzyme in vitro between keratin 1͞10 KIF and Ile-substituted form of keratin 5 and wild-type keratin 14 (b); and the a synthetic loricrin peptide. (Lower, Inset) The labeled peptides shown Leu-174 3 Pro-substituted form of keratin 5 and wild-type keratin 14 were isolated for quantitation and sequencing (Table 3). (c). (Bar ϭ 100 nm.) further explored in vitro the specificity of crosslinking of KIF by Leu-174 3 Pro, corresponding to position 7 of the 1A rod domain TGases. In initial experiments, we found that the TGase 3 enzyme of the keratin 5 chain, failed to assemble (Fig. 1c), as expected, ͞ ͞ because it interferes with KIF structure (28). did not use keratin 1 10 or 5 14 KIF as a complete substrate in Crosslinking of KIF and a Synthetic Peptide in Vitro with the vitro, because Ͻ2% of KIF chains became oligomerized as judged TGase 3 Enzyme Involves the Same Conserved Lys Residue. We on SDS͞PAGE gels (data not shown). We made and labeled a synthetic peptide corresponding to sequences of human loricrin that efficiently donate Gln residues for TGase crosslinking both in vivo and in vitro (14, 20). Only about 1 mol of peptide was incorporated per mol into KIF assembled from wild-type keratins 5͞14 (Fig. 2, solid bars) or wild-type keratins 1͞10 KIF (not shown) with peptide:KIF ratios Յ1:3. At similar ratios, only trace amounts were incorporated into KIF assembled from the wild-

Table 3. Labeled peptides from in vitro crosslinking of keratin 1͞10 KIF Yield, Peptide mol Sequence Origin 1 0.02 VSSQQVTQTSCA Gln-216 : YSSSSKQFSSSR Keratin 10 Lys-9 2A 0.55 VSSQQVTQTSCA Gln-215 : SLVNLGGSKSISISVAR Keratin 1 Lys-73 2B 0.48 VSSQQVTQTSCA Gln-219 : SLVNLGGKSISISVAR Keratin 1 Lys-73 3 0.05 VSSQQVTQTSCA Gln-215 : ISSSKGSLGGGF Keratin 10 Lys-32 FIG. 2. Stoichiometry of crosslinking of a synthetic loricrin peptide to KIF in vitro. Wild-type keratin 5͞14 KIF (solid bars) or with KIF Single-letter code of amino acids is used. Molar amounts of assembled from the keratin 5 chain with Lys-71 3 Ile substitution and crosslinked peptides were determined from the recovered amount of wild-type keratin 14 (open bars) were mixed with the synthetic peptide in labeled peptide of known specific activity, because in the Edman various molar ratios as shown, crosslinked with the TGase 3 enzyme, and degradation sequencing reactions, each peak from Fig. 3 also con- assayed. tained other tryptic peptides from the keratin chains. Downloaded by guest on October 1, 2021 Biochemistry: Candi et al. Proc. Natl. Acad. Sci. USA 95 (1998) 2071

FIG. 5. Model showing two different modes of association of KIF with the foreskin epidermal CE. At left is the possible organization of KIF near the vicinity of a desmosomal remnant including the cyto- plasmic domain of desmoplakin with envoplakin and involucrin. At right is shown the possible association of KIF with the intracellular surface of the CE, which is enriched in loricrin and SPRs. This model is not drawn to scale, because the molecular dimensions of no CE structural proteins are known. Note that although few direct data are available, it is to be expected that foreskin and palmar-plantar epidermal CEs will be generally similar, especially because they both contain high amounts of loricrin and have similar masses per unit area (30); however, they are different from other internal stratified squa- mous epithelia, which contain little or no loricrin.

as of the epidermis of the palms and soles and, to a mild degree, other limited body sites. In view of the foregoing data on the importance of this Lys residue in crosslinking, we performed ultrastructural studies on skin samples. We could not detect any abnormality in the organization of KIF bundles in FIG. 4. Ultrastructural analysis of palmar epidermis from a normal basal or spinous keratinocytes (not shown). However, in granular individual (a and d) and an affected patient with diffuse NEPPK disease cells where assembly of the CE is initiated, KIF bundles appeared (b, c, and e). In the patient, the basal and lower spinous layers appeared to detach and retract abnormally from beneath desmosomes and normal (data not shown). However, in cell layers extending from the along much of the cell periphery, often forming microclefts or upper spinous to granular layers, where CE assembly is initiated, sepa- b rations of KIF from desmosomes and the cell periphery occur vacuolar inclusions (Fig. 4 ). This feature is not present in palmar in the patient, leading to microclefts, which are marked by asterisks in b. epidermis from a normal individual (Fig. 4a). Moreover, the In addition, at the interface between the uppermost granular and first boundary between the uppermost granular layer and the first cornified cell, a highly convoluted interface is evident in the patient (c), cornified cell layer was irregular in shape with deep interdigita- which is absent in normal epidermis (d). At higher resolution, the KIF tions between the two layers (Fig. 4c), which was also not evident appear withdrawn from the desmosomal and interdesomomal regions in normal palmar epidermis (Fig. 4d). At high magnification, the (asterisk) in many places in the patient (e), unlike in normal (d). [Bars ϭ ␮ ␮ ␮ retraction of the KIF from the cell periphery at and between 2.5 m(a and b), 1 m(c), 0.5 m(d and e)]. SC, stratum corneum; G, desmosomal sites is evident (Fig. 4e), which is not present in granular layer; KH, keratohyalin granule. normal epidermis (Fig. 4d). The outer cornified layers appeared type keratin 14 and Lys-71 3 Ile-substituted keratin 5 chains (Fig. similar to the control (not shown). Epidermis from other unin- 2, open bars). However, at higher molar ratios, Ͼ1 mol of volved body sites appeared normal (not shown). peptide͞mol of KIF was incorporated into both wild-type and These data suggest that the loss of the Lys-73 residue in about 50% of the keratin 1 chains expressed in the cells could cause substituted KIF (Fig. 2). discoordination between the KIF cytoskeleton and periphery of In a preparative crosslinking experiment with keratin 1͞10 granular cells, apparently resulting in a distortion in the shape of KIF, followed by tryptic digestion, three labeled tryptic peptide the keratinocyte. The increased surface area of the corneocytes peaks were identified by HPLC (Fig. 3 Lower) and isolated. By and their abnormal shape may prevent efficient distribution of amino acid sequencing (Table 3), we could identify four novel lamellar body lipid material, which in turn results in reduced peptides, although mixed with unlabeled tryptic peptides. From barrier formation, and thus a hyperkeratotic response (5). the incorporated radioactivity, 94% involved Lys-73 of the keratin The Role of the Conserved Lysine Residue: A Model for the 1 chain. Small amounts were also incorporated into Lys-9 and Coordination of KIF and the CE. There are more than 50 Lys Lys-32 of the keratin 10 chain, as also seen in the in vivo data. residues in the keratin 1 and 10 (or 5 and 14) chains, yet our Together, these data confirm the specificity of the conserved Lys data reveal that only one of these, located within a highly residue of the type II keratin chains for TGase crosslinking. conserved region of the V1 subdomain of the head domain of Significant Morphological Changes in Human Epidermis Are type II chains, is utilized with high specificity for crosslinking Due to the Loss of the Conserved Lys Residue in a Family with to many protein partners by TGases in vivo and in vitro.In Diffuse NEPPK. Previously, one case of this disease was reported addition, this residue is not essential for KIF assembly in vitro to be caused by a mutation in the KRT1 of one allele that (Fig. 1). That this residue may not be essential for normal KIF resulted in a substitution of the Lys residue by Ile in the keratin assembly in vivo comes from the ultrastructural data of Fig. 4: 1 chain (16), which is the same residue identified in the above in there was no abnormal KIF clumping or cell lysis typically seen vivo and in vitro crosslinking studies. This disease was manifested in other keratin disorders because of mutations in rod domain Downloaded by guest on October 1, 2021 2072 Biochemistry: Candi et al. Proc. Natl. Acad. Sci. USA 95 (1998)

sequences that are known to be essential for KIF assembly (28, implicated in in vitro assays in the association of KIF with 29). Therefore, this residue position is likely to be more desmoplakin (33, 34). important for the function or supramolecular organization of We thank the patient and normal volunteer for cooperation with this the KIF in cells. study. Specifically, our new data suggest that this Lys residue plays an essential role in the normal crosslinking of KIF to the cell 1. Hohl, D. (1990) Dermatologica 180, 201–211. periphery and desmosomes in the formation of the CE. We 2. Greenberg, C. S., Birckbichler, P. J. & Rice, R. H. (1991) FASEB propose that in this way the KIF cytoskeleton of cornified or other J. 5, 3071–3077. terminally differentiated stratified squamous epithelial cells is 3. Reichert, U., Michel, S. & Schmidt, R. (1993) in Molecular structurally attached permanently to the CE of the cornified cell. Biology of the Skin, eds. Darmon, M. & Blumenberg, M. (Aca- demic, New York), pp. 107–150. Our data suggest that this may occur in two types of interaction 4. Holbrook, K. A. & Wolff, K. (1993) in Dermatology in General complexes (Fig. 5). The first is at desmosomal anchorage sites Medicine, eds. Fitzpatrick, T. B., Eisen, A. Z., Wolff, K., Freed- where KIF containing the keratin 5 chain (and keratin 6 in berg, I. M. & Austen, K. F. (McGraw–Hill, New York), pp. palmar-plantar epidermis and in cultured cells) and later, the 97–145. keratin 1 chain, interact with envoplakin, involucrin, desmo- 5. Elias, P. M. (1996) J. Dermatol. 23, 756–758. , and probably other proteins. The data in Table 1 imply 6. Downing, D. T., Stewart, M. E., Wertz, P. W. & Strauss, J. S. these may represent up to 60% (molar basis) of all keratin- (1993) in Dermatology in General Medicine, eds. Fitzpatrick, T. B., containing crosslinks. A second or concurrent interaction com- Eisen, A. Z., Wolff, K., Freedberg, I. M. & Austen, K. F. plex is where KIF cytoskeleton containing predominantly the (McGraw–Hill, New York), pp. 210–221. keratin 1, 2e, and 10 chains in the cornified cells contacts the 7. Swartzendruber, D. C., Wertz, P. W., Madison, K. C. & Downing, D. T. (1987) J. Invest. Dermatol. 88, 709–713. intracellular surface of the CE, which is predicted to be composed 8. Schwarz, M. A., Owaruibe, K., Kartenbeck, J. & Franke, W. W. mostly of loricrin and SPRs (14) (Fig. 5). Our data from Table 1 (1994) Annu. Rev. Cell Biol. 6, 461–491. imply these constitute at least 40% of the total. 9. Collins, J. E. & Garrod, D. R. (1994) Molecular Biology of Existing data allow an estimate of the degree of crosslinking Desmosomes and (R. G. Landes Company, of KIF to the CE. These calculations assume a model in which Austin, TX). KIF of 15 nm in diameter containing 16 heterodimer mole- 10. Garrod, D. R., Chidgey, M. & North, A. (1996) Curr. Opin. Cell cules͞46 nm (29) line the intracellular surface of the CE (ref. Biol. 8, 670–678. 4; see also Fig. 4d) and assume that: (i) keratin-containing 11. Green, K. J. & Jones, J. C. R. (1996) FASEB J. 10, 871–881. crosslinks represent about 0.1% of the total isopeptide 12. Steven, A. C. & Steinert, P. M. (1994) J. Cell Sci. 107, 693–700. crosslink in foreskin epidermis (14, 15); (ii) there are 89 nmol 13. Steinert, P. M. (1995) Cell Death Differ. 2, 23–31. of crosslinks͞mg of foreskin CE proteins (14, 15, 22); and (iii) 14. Steinert, P. M. & Marekov, L. N. (1995) J. Biol. Chem. 270, 17702–17711. CEs from a variety of murine tissues including the footpad, 15. Steinert, P. M. & Marekov, L. N. (1997) J. Biol. Chem. 272, which is analogous to human palmar-plantar epidermis, are ϭ ϫ 2021–2030. remarkably consistent in thickness of about 7.2 kDa [ 7.2 16. Kimonis, V., DiGiovanna, J. J., Yang, J.-M., Doyle, S. Z., Bale, 3 2 10 atomic mass units (amu)]͞nm (30). Thus, there are [(89 ϫ S. J. & Compton, J. G. (1994) J. Invest. Dermatol. 103, 764–769. Ϫ Ϫ 10 9) ϫ (1 ϫ 10 3) ϫ (6 ϫ 1023)] keratin crosslinks͞6 ϫ 1020 17. Kim, S.-Y., Kim, I.-G, Chung, S.-I. & Steinert, P. M. (1994) atomic mass units (amu) of CE mass ϭ 1.1 ϫ 108 amu͞keratin J. Biol. Chem. 269, 27979–27986. crosslink ϭ one keratin crosslink͞1,500 nm2 of CE surface ϭ 18. Konigsberg, W. H. & Henderson, L. (1983) Methods Enzymol. 91, one crosslink͞100 nm of KIF length, or about 0.5 crosslink͞ 254–259. unit molecule length of KIF. This means that 2–3% of available 19. Steinert, P. M. (1991) J. Struct. Biol. 107, 175–188. keratin 1 Lys-73 residues participate in crosslinking. Accord- 20. Candi, E., Melino, G., Mei, G., Tarcsa, E., Marekov, L. N. & J. Biol. Chem. 270, ingly, irrespective of the exact molecular model of association Steinert, P. M. (1995) 26382–26390. 21. Chipev, C. C., Korge, B. P., Markova, N. G., Bale, S. J., with the CE, the data indicate that KIF are frequently attached DiGiovanna, J. J., Compton, J. G. & Steinert, P. M. (1992) Cell to the CE, providing a tight, permanent anchor. Therefore, it 70, 821–828. could be reasonably expected that reductions in crosslinking 22. Hohl, D., Lichti, U., Turner, M. L., Roop, D. R. & Steinert, P. M. caused by a mutation may have significant consequences for (1991) J. Biol. Chem. 266, 6626–6636. the structural integrity of the cornified cell. 23. Plaue, S. & Briand, J. P. (1988) in Laboratory Techniques in A question thus arises as to why pathology in the case of Biochemistry and Molecular Biology, eds. Burdon, R. H. & van NEPPK disease affects primarily sites of thickened epidermis of Knippenberg, P. H. (Elsevier, Amsterdam), pp. 41–94. the palms and soles. One possible explanation lies with the 24. Kim, H.-C., Lewis, M. S., Gorman, J. J., Park, S.-C., Girard, J. E., abundance of total crosslinks in CE samples from different body Folk, J. E. & Chung, S.-I. (1990) J. Biol. Chem. 265, 21971–21978. sites, which range from 89 nmol͞mg in foreskin epidermis (3, 14, 25. Elias, P. M. (1983) J. Invest. Dermatol. 80, 44–49. 26. Sun, T.-T., Eichner, R., Cooper, D., Schermer, A., Nelson, W. G. & 20), to 75 nmol͞mg in human trunk epidermis (31), but 45 ͞ Weiss, R. A. (1984) in The Cancer Cell: The Transformed Phenotype, nmol mg in palmar callus epidermis (32). It is possible that the eds. Levine A., Topp, W. & Vande Woude, G. (Cold Spring Harbor loss of 50% of the potential crosslinks through the keratin 1 chain Lab. Press, Plainview, NY), pp. 167–176. may reduce the amount necessary for normal structure and 27. Collin, C., Moll, R., Kubicka, S., Ouhayoun, J. P. & Franke, function below a critical threshold level in the palms and soles. W. W. (1992) Exp. Cell Res. 202, 132–141. In summary, our data reveal a mechanism by which the KIF 28. Fuchs, E. (1996) Annu. Rev. Genet. 30, 197–231. cytoskeleton is structurally integrated with the cell periphery 29. Steinert, P. M., Marekov, L. N., Fraser, R. D. B. & Parry, D. A. D. in terminally differentiated stratified squamous epithelia such (1993) J. Mol. Biol. 230, 436–452. as the epidermis. This system utilizes TGase crosslinking to 30. Jarnik, M., Simon, M. N. & Steven, A. C. (1997) J. Invest. build a permanent, stable structure essential for normal epi- Dermatol. 108, 603 (abstr.). 31. Martinet, N., Beninati, S., Nigra, T. P. & Folk, J. E. (1990) dermal structure and function. In addition, the present iden- Biochem. J. 271, 305–308. tification of in vivo crosslinks, coupled with this case of diffuse 32. Rice, R. H. & Green, H. (1977) Cell 11, 417–422. NEPPK, has provided valuable new information of the role of 33. Kouklis, P. D., Hutton, E. & Fuchs, E. (1994) J. Cell Biol. 127, the highly conserved Lys residue on head domain of type II 1049–1060. keratin chains in the supramolecular organization and function 34. Meng, J.-J., Bornslaeger, E. A., Green, K. J., Steinert, P. M. & Ip, of KIF. Also, sequences near this lysine residue have been W. (1997) J. Biol. Chem. 272, 21495–21503. Downloaded by guest on October 1, 2021