Proc. Natl. Acad. Sci. USA Vol. 90, pp. 10754-10758, November 1993 Cell Biology Overexpression of human loricrin in transgenic mice produces a normal phenotype Kozo YONEDA AND PETER M. STEINERT Skin Biology Branch, National Institute of Arthritis and Musculoskeletal and Skin Diseases, National Institutes of Health, Bethesda, MD 20892 Communicated by Henry Metzger, August 25, 1993 (received for review July 7, 1993)

ABSTRACT The cornified cell envelope (CE) ofterminally novel glycine loop motif which is variable in sequence and differentiating stratified squamous epithelial cells is a complex highly flexible in likely conformation (16, 17). Since the same multiprotein assembly about 15 nm thick of which loricrin is a glycine loop motifexists on the intermediate filaments major component. We have produced transgenic mice bearing also expressed in terminally differentiated epithelial cells (16, the human loricrin transgene in order to study the role of 18-20), it was proposed that an interaction between the loricrin in CE assembly, structure, and function. By analyses glycine loop sequences of loricrin on the CE and on the of RNA and , we show that the human transgene is intracellular keratin intermediate filaments may stabilize expressed in mouse epithelial tissues in an appropriate devel- cellular structure (16). opmental manner but at an overall level about twice that of Attempts to test these hypotheses and to study the struc- endogenous mouse loricrin. Thus the 20-kbp construct used ture and function of loricrin have proven difficult for several contains all necessary regulatory elements. By immunogold reasons. First, loricrin is poorly expressed in established electron microscopy, all of the expressed protein is incorpo- epithelial cell culture systems (21). In our hands, transfec- rated into the CE. That no alternations were noted indicates tions of loricrin constructs into a wide variety of cells have that overproduction of the loricrin component of the CE does resulted in very little or no expression, apparently because it not affect the flexible structure or function of the epithelial is toxic (unpublished observations). Further, it has not been tissues. Furthermore, these data imply that loricrin may be the possible to isolate loricrin for direct biochemical experiments last protein to be deposited onto, and thus lines, the intracel- apparently because it is rapidly crosslinked after synthesis lular surface of the CE, where it may be accessible to interact (11). As an alternative, we have explored the expression of with the subjacent keratin intermediate-ifiament network. the human loricrin in transgenic mice. A 20-kbp genomic construct was found to direct the net overexpression of The cornified cell envelope (CE) is a 15-nm-thick layer of human loricrin that is assimilated without evident pathology. protein deposited on the intracellular surface of the plasma We interpret this to mean that it is a very late component of membrane of terminally differentiating stratified squamous CE assembly, so that the intracellular surface is layered with epithelia tissues, where its major role is to provide a barrier loricrin, an event that is consistent with the hypothesis for an against the environment (1, 2). The highly insoluble nature of interaction between loricrin and the keratin intermediate the CE can be attributed to crosslinking of its constituent filaments. by both disulfide bonds and NE-(--glutamyl)lysine isodipeptide bonds formed by the action of transglutami- MATERIALS AND METHODS nases. Likely CE proteins include (3, 4), cystatin A (5, 6), members of the class of small proline-rich proteins Construction of the Human Loricrin Transgene. A plasmid (7-9), (10), loricrin (11, 12), and possibly other containing the human loricrin gene was constructed by stan- proteins such as and keratin intermediate filaments dard methods in three steps. An existing human loricrin (13, 14). To date, only loricrin has been shown to be genomic clone which contained about 1.5 kbp of 5' and 9 kbp crosslinked directly to the CE by isodipeptide bonds (12). of3' flanking sequences (17) was cleaved at its unique BssHII Thus, major questions arise as to the relative contributions site in the coding sequence. By use of PCR the coding of these proteins to the structure, organization, and function sequence 3' to this site was modified by insertion of a 27-bp of the CE. It has proven difficult to ascertain the temporal fragment encoding the carboxyl-terminal part of the neu- order of expression, the abundance, and the function ofthese ropeptide substance P sequence, just prior to the termination proteins in the CE because the irreversibly crosslinked nature codon. The substance P nucleotide sequence (but not coding of the CE precludes dissection of the structure and because sequence) was modified by creation ofa unique Sal I site. The several of the proteins appear to be expressed concurrently engineered fragment was then reinserted into the genomic very late in terminal differentiation in epithelia such as the clone. Next, a 9-kbp genomic clone harboring sequence 5' to (1, 2). However, some structure-function data are the loricrin cap site was isolated by standard procedures (17) now emerging. Due to their elongated a-helical structures, and ligated to the assembly. The final construct of about 20 involucrin (15) and trichohyalin (10) have been suggested to kbp (Fig. 1) was assembled in the pGEM-3Z vector serve as "scaffold" proteins in the CE, to which other more (Promega) and contained in the following linear order, 5' to abundant proteins are later attached. Also, two types of data 3', about 9 kbp of 5' flanking sequence; the entire human suggest that loricrin is the major CE component: its mRNA loricrin gene consisting of exon I in 5' untranslated sequence, represents about 10% of the total in such tissues as the intron 1, and exon II modified with the substance P peptide; epidermis (10-13), and its amino acid composition is strik- and about 8.6 kbp of 3' noncoding sequence. Unique Sph I ingly similar to that of isolated epidermal CEs, especially andXba I sites at the 5' and 3' ends ofthe construct were used with respect to the high content ofglycine and serine (11-13). to excise the DNA from the plasmid vector. These glycine/serine-rich sequences are thought to adopt a Transgenic Mice. The construct was microinjected into fertilized mouse eggs [(C57BL6/J x SJLJ)F1, The Jackson as mice The publication costs of this article were defrayed in part by page charge Laboratory] essentially described (22). Transgenic payment. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. §1734 solely to indicate this fact. Abbreviation: CE, cell envelope. 10754 Downloaded by guest on October 4, 2021 Cell Biology: Yoneda and Steinert Proc. Natl. Acad. Sci. USA 90 (1993) 10755

8 kbp 200 bp 8 kbp ,Z.- K0~~ CAP ATG TAG MTAAA Mouse -> -1.6 ExonIl gI Exo

II SphI SphM sKI Dra I BssHU1 Sal I Dra I Kpn I Xba I Human -->. i- 1.3 FIG. 1. Structure of20-kbp construct. The central 2-kbp gene was gene flanked at both the 5' and the 3' end by about 9 kbp. The consists lJ ll1 of exon I in 5' untranslated sequence, a large intron, and exon II, which contains all of the coding sequence. The coding portion was Loricrin Substance P modified by insertion of the substance P neuropeptide sequence probe probe immediately prior to the termination codon. Diagnostic control FIG. 2. Northern blot of total epidermal RNA showing expres- sequences and restriction enzyme sites are shown. CAP, mRNA sion of endogenous mouse loricrin (1.6 kb) and transgenic hL.P (1.3 start; ATG, initiation codon; TAG, termination codon; AATAAA, kb) mRNAs. Wt, wild type; Non-Tg, nontransgenic littermates; polyadenylylation signal. TgTg, homozygous transgenic mice. Similar blots of RNA from epidermis, esophagus, and stomach were scanned in a phospho- were identified by PCR using two synthetic oligonucleotides: imaging device to estimate an -1.8-fold overexpression of the hL.P plus strand in the substrance P tag, 5'-GTCGACTGTCAGT- transcript. TCTTCGG-3', and minus strand in 3' noncoding sequence, 5'-CACCTTTCCGCCAATAGAGT-3'. PCRs employed a levels of structural homology in terms of their glycine loop commercial DNA amplification kit (Perkin-Elmer/Cetus) motifs, as well as , especially in their with an initial denaturation at 95°C (1.5 min) and 30 cycles of amino- and carboxyl-terminal glutamine-rich sequences. Ac- denaturation at 95°C (0.5 min), annealing at 54°C (1 min), and cordingly, we configured the substance P neuropeptide se- elongation at 72°C (1 min), with 1 ug of genomic DNA. quence just prior to the termination codon (Fig. 1) in order to Miscellaneous Procedures. RNA was isolated from heat- follow the expression of human loricrin (termed hL.P) in the separated epidermis or other tissues by extraction with mouse background at both the nucleic acid and protein levels. guanidine hydrochloride (23). Northern blots (12) were The 20-kbp construct used contained the entire human lori- probed with either the substance P oligonucleotide defined crin gene of 2 kbp flanked by about 9 kbp on each side (Fig. above (for specific reaction of hL.P transcripts only) or a 1). Following injection of this construct, three founder lines human loricrin cDNA coding probe (12) (which crossreacts were produced, which by PCR analysis and Southern blot with both the endogenous mouse and hL.P transcripts). RNA analysis using specific primers, had stably incorporated were done with DNase I-treated total PCR experiments five into the mouse genome (data not shown). cellular RNA (2 ng) of transgenic and nontransgenic animals about copies Such founders were then bred to produce heterozygous F1 (24). The two primers listed above generate a product of 390 and homozygous F2 transgenic colonies. Mice of each of bp only for the human transgenic construct, as shown by these lines were normal. electrophoresis in a 1% agarose gel (NuSieve, FMC). Two other primers-plus, 5'-TACCTGGCCGTGCAAGTAAG- Correct Tissue Expression of hL.P mRNA. Northern blot 3', and minus, 5'-ACACCTTGAGCGACTCAATG-3'-gave analyses of epidermal RNA from transgenic mice revealed mouse a product of 173 bp for the endogenous mouse loricrin only. the presence of transcripts of both the endogenous Total epidermal RNA from several homozygous transgenic (1.6 kb) and human (1.3 kb) loricrins, but the hL.P transcript mice was pooled and used for in vitro translation with the was not seen in nontransgenic littermates or wild-type mice New England Nuclear kit and [35S]-cysteine (11), followed by (Fig. 2). In a survey of tissues by RNA PCR (Fig. 3), hL.P immunoprecipitation with either an anti-loricrin antibody mRNA was expressed only in the stratified squamous epi- elicited against the carboxyl-terminal end (11, 12), which thelia of epidermis, palate, stomach, tongue and esophagus reacts with both mouse and human loricrins, or an anti- (see Fig. 6) or vagina (data not shown), but not in various substance P antibody (Wako Biochemicals, Osaka) (11, 12). tissues such as brain, heart, kidney, liver, lung and muscle, We found that incorporation into translated loricrins was exactly as reported for mouse loricrin (11). linear between 1 and 20 ,ug of total RNA; for quantitation Correct Temporal Expression of hL.P mRNA in the Epider- purposes, 10 ,ug was used. Autoradiograms were scanned on mis. Similar RNA PCR experiments were performed on a phospho-imaging device with IMAGEQUANT software ver- epidermal tissue samples recovered at various times of fetal sion 3.0 (Molecular Dynamics). Immunofluorescence (12) development. The hL.P mRNA appeared concurrently with

and immuno-electron microscopy procedures using protein .b were done with these two antibodies as described 011 \,- A-gold *. .z <\ ,.- ,-. e " ,z e , .p pp ..ell (24). 10 'e,\L\-6, ,O \". 0.,e.e qoe",q-,a 6\( p Epidermal Cell Culture. Epidermis from homozygous transgenic mice was harvested and cells were cultured in low-calcium (0.05 mM CaCl2) medium for 6 days (21). Some Human -) cultures were shifted to high-calcium (0.35 mM) at day 4 and cultured for 2 days, as described (21). RNA was then ex- tracted, probed by slot blotting as above, followed by quan- titation by phosphoimaging. Estimation ofhL.P Expression by Amino Acid Analysis. CEs were isolated from newborn mouse epidermis and their amino acid compositions were determined (12). Calculations of the relative abundances of loricrins based on their known amino acid compositions employed an established algorithm (13). FIG. 3. RNA PCR from selected tissues reveals expression of hL.P in only stratified squamous epithelial tissues oftransgenic mice. RESULTS AND DISCUSSION Arrows designate the 390-bp band of hL.P transcript and the 173-bp band of endogenous mouse loricrin. Wt and Non-Tg, epidermis from Construction of a Human Loricrin Transgene. Although wild-type mice and from nontransgenic littermates. Markers were human and mouse loricrins differ in size, they share high Msp I fragments of pBR322 DNA (GIBCO/BRL). Downloaded by guest on October 4, 2021 10756 Cell Biology: Yoneda and Steinert Proc. Natl. Acad Sci. USA 90 (1993)

'Cle Based on the known number of cysteines in mouse (11) and human loricrins, we calculated in k ell (12) multiple experiments wo .Cll0 4.p that the levels of mouse loricrin expression in wild-type and ,t0 & N(b N,. N,D Nq) -"A Nt Nq transgenic mice were similar, at 9.55 ± 0.08 and 9.68 ± 0.18

ir arbitrary molar units, respectively, whereas expression of hL.P mRNA was 14.9 ± 0.1 units, about 1.6 times higher. Human -* Calcium-Dependent Expression of hL.P mRNA. We next examined whether the calcium-responsive elements required A Mouse >

FIG. 4. RNA PCR shows that expression of hL.P begins at about embryonic day 16, in parallel with expression of endogenous mouse loricrin. the mouse loricrin at day 16 (Fig. 4), as expected for late epithelial differentiation products (25). High Levels ofhL.P mRNA. We used two methods to assess the relative amount of hL.P mRNA in tissues. In the first, Northern blots as in Fig. 2 were scanned in a phospho- imaging device. The relative hybridization efficiencies of the loricrin probe with the mouse loricrin and hL.P mRNAs were corrected by using standard loricrin probes in slot blotting experiments (10). In this way, we found that hL.P mRNA was 1.7-1.8 times more abundant than endogenous mouse loricrin mRNA in epidermis, esophagus, and stomach of homozy- gous transgenic mice. In a second approach, epidermal RNA from a litter of homozygous transgenic mice, as well as from nontransgenic and wild-type mice, was separately pooled and translated in vitro in the presence of [35S]cysteine with a commercial cell-free system. After immunoprecipitation with either the anti-loricrin or the anti-substance P antibodies, the products were resolved by polyacrylamide gel electrophore- sis and fluorography. Fig. 5 shows that the former antibody precipitated both 60-kDa (mouse loricrin) and 30-kDa hL.P products (lane 5), whereas the latter antibody precipitated only the hL.P (lane 6). These data therefore confirm that full-length hL.P mRNA is indeed expressed. The same flu- orograph was used to quantitate the bands of lanes 1 and 5. 1 2 3 4 5 6 kDa -97 -69 Mouse -> , -46 Human -- -30

LI L Wt m NH TgTg anti-Loricrin I II

anti-Substance P I FIG. 6. hL.P expression in tissues. (A) Hematoxylin- and eosin- FIG. 5. Quantitation of expression of hL.P mRNA. Total epider- stained section of homozygous transgenic skin showing normal mal cellular RNA was translated in the presence of [35S]-cysteine and appearance of the epidermis, dermis, and a developing hair follicle the protein products precipitated with either the anti-loricrin anti- bud. (B-D) Indirect immunofluorescence with either the general body (lanes 1, 3, and 5) or anti-substance P antibody (lanes 2, 4, 6) anti-loricrin antibody (B), the anti-substance P antibody (C), or both for SDS/7.5% PAGE. Bands were then quantitated in a phospho- (D), showing coincident expression of endogenous mouse and hL.P imaging device to estimate the relative amounts of mouse and hL.P loricrins in the granular layers of the epidermis. (E-F) hL.P expres- mRNA transcripts. RNA was from epidermis of wild-type mice (Wt sion in the inner (lumen) surfaces ofthe stratified squamous epithelial m, lanes 1 and 2) or homozygous transgenic mice (TgTg, lanes 5 and tissues ofesophagus (E) or stomach (F). (G) hL.P expression on both 6) or from normal human foreskins (NH, lanes 3 and 4). epidermal surfaces of the ear. (Bars = 50 ,Lm.) Downloaded by guest on October 4, 2021 CeU Biology: Yoneda and Steinert Proc. Natl. Acad. Sci. USA 90 (1993) 10757 for normal loricrin expression (21) were also retained on the from homozygous transgenic animals contained 46.8 ± 0.3% transgene. Total cellular RNA from epidermal glycine, which corresponds to a substantial net increase in grown in low (0.05 mM) or high [0.35 mM, known to be the total amount of loricrin. Since the RNA translation data optimal for expression of human loricrin (21)] calcium from of Fig. 5 show that amount of expressed mouse loricrin wild-type, nontransgenic littermates or homozygous trans- remains constant in transgenic animals, this increase in genic animals was extracted, slot blotted, examined with the glycine is attributable to expression ofthe hL.P (47.1%). With two loricrin probes, and quantitated by phosphoimaging. The the same algorithms, this corresponds to about 1.8 times level of hL.P or mouse loricrin mRNA expression (arbitrary more hL.P than endogenous mouse loricrin. units) in low calcium was bearly detectable (background) at Overexpression of hL.P Does Not Produce an Abnormal 10 ± 8. Expression of hL.P in high calcium was background Phenotype. In view of these data that indicate substantial of 8 t 10 in nontransgenic or wild-type animals but was 125 overexpression of the hL.P in the mouse background, we ± 4 in transgenic animals. This 10-fold increase indicates that surveyed the entire body ofnewborn homozygous transgenic the calcium-responsive elements for correct hL.P expression animals by both light (Fig. 6) and electron (Fig. 7) micros- are contained within the 20-kbp construct used. copy. We found no unusual morphology or evident pathology High Levels of hL.P Protein Synthesis as Determined by in any tissue, including the epidermis, where hL.P was Amino Acid Analysis. To date, it has not been possible to expressed most abundantly (Fig. 6A). Deposition of the isolate intact loricrin from any epithelial tissue, to determine endogenous mouse loricrin (Fig. 6B) or hL.P (Fig. 6 C-G) in its abundance directly, apparently because it is crosslinked trunk epidermis (Fig. 6 C and D), esophagus (Fig. 6E), shortly after synthesis (11), but indirect estimates have been stomach (Fig. 6F), or the two epidermal layers of both sides possible by use of amino acid analyses (13). Epidermal CEs of the ear (Fig. 6G) were entirely normal. We also used from wild-type or nontransgenic mice contain 45.1% glycine immunogold localization in the epidermis with the general (11), from which it is possible to calculate by least-squares loricrin and hL.P-specific antibodies. With both probes, fitting methods that mouse loricrin (55.1% glycine) consti- 93-95% of the gold particles were confined either to the tutes about 75% of the CE protein, and the several other granules of the inner living cell layers in the epidermis or to proteins cumulatively amount to about 25% of the CE (13). the cell periphery (CE) in the cornified cell layers. The few By amino acid analysis, we found that epidermal CEs isolated gold particles not associated with defined structures were

iL E se >P

TI

G

FIG. 7. Immunogold electron microscopy of transgenic mouse epidermis, using general anti-loricrin antibody (A-C) or anti-substance P antibody (D and E). Protein A-gold particles were restricted to either the loricrin (L) or filaggrin (F)-keratohyalin granules of the granular layers (G) of the epidermis (A, C, and E), or within 15 nm of the cell envelope at the cell periphery (CE) (A, B, and D). T, transition cells; SC, layer. [Bars = 1 Am (in A) or 0.5 ,um (B-E).] Downloaded by guest on October 4, 2021 10758 Cell Biology: Yoneda and Steinert Proc. Natl. Acad. Sci. USA 90 (1993) probably due to background (24). This means that virtually all 2. Polakowska, R. & Goldsmith, L. A. (1991) in Physiology and of the expressed endogenous mouse and hL.P loricrins are Molecular Biology of the Skin, ed. Goldsmith, L. A. (Oxford ultimately incorporated into the CE, as supported by amino Univ. Press, New York), pp. 168-201. acid analysis. Whereas all mouse loricrin gold particles were 3. Rice, R. H. & Green, H. (1979) Cell 18, 681-694. localized to the loricrin granules of the epidermis (Fig. 7C) 4. Yaffe, M. B., Murthy, S. & Eckert, R. L. (1993) J. Invest. (12, 24), in the case hL.P Dermatol. 100, 3-9. of the antibody, about 20% of the 5. Zettergren, J. G., Peterson, L. L. & Wuepper, K. (1984) Proc. gold particles were localized over the filaggrin granules (Fig. Natl. Acad. Sci. USA 81, 238-242. 7E). Normal human epidermis does not contain defined 6. Takahashi, M., Tezuka, T. & Katunuma, N. (1992) FEBS Lett. loricrin granules, but the human loricrin is deposited into 308, 79-82. compound granules together with profilaggrin (17). Thus the 7. Kartasova, T., van Miujen, G. N. P., van Pelt-Heerschap, H. hL.P contains all of the features that enable it to be assimi- & van de Putte, P. (1988) Mol. Cell. Biol. 8, 2204-2210. lated in the same way as for endogenous loricrin. 8. Hohl, D. & Beckendorf, C. (1992) Nature Genet. 1, 91. Concluding Remarks: Is Loricrin the Last Component 9. Marvin, K. W., George, M. D., Fujimoto, W., Saunders, Added During CE Assembly? The present experiments have N. A., Bernacki, S. H. & Jetten, A. M. (1992) Proc. Natl. documented at both the mRNA and protein levels that a Acad. Sci. USA 89, 11026-11030. 20-kbp construct directs the correct temporal and site- 10. Lee, S.-C., Kim, I.-G., Marekov, L. N., O'Keefe, E. J., Parry, specific expression of the human loricrin transgene in trans- D. A. D. & Steinert, P. M. (1993) J. Biol. Chem. 268, 12164- genic mice, at a net level 1.6-1.8 times greater than for 12176. 11. Mehrel, T., Hohl, D., Rothnagel, J. A., Longley, M. A., Bund- endogenous mouse loricrin. The nearly 3-fold increase in the man, D., Cheng, C., Lichti, U., Bisher, M. E., Steven, A. C., total amount of loricrin creates no detectable abnormality in Yuspa, S. H. & Roop, D. R. (1990) Cell 61, 1103-1112. any tissue or in the life of the mouse. Thus, events involved 12. Hohl, D., Mehrel, T., Lichti, U., Turner, M. L., Roop, D. R. in the orderly assembly of the CE have not been interferred & Steinert, P. M. (1991) J. Biol. Chem. 266, 6626-6636. with. In contrast, overexpression of involucrin in transgenic 13. Steven, A. C. & Steinert, P. M. (1992) J. Invest. Dermatol. 98, mice produces an abnormal phenotype in both the epidermis 559 (abstr.). and hair follicles, presumably because it is an early scaffold 14. Ming, M. E., Daryanani, H. A., Roberts, M. L., Haimowitz, protein essential for correct CE formation (26). Although our J. E. & Kvedar, J. C. (1993) J. Invest. Dermatol. 100, 578 new data do not directly address the question of temporal (abstr.). expression, according to in situ hybridization and immuno- 15. Yaffe, M. B., Beegen, H. & Eckert, R. L. (1992) J. Biol. Chem. 267, 12233-12238. logical studies, loricrin is expressed later during epidermal 16. Steinert, P. M., Mack, J. W., Korge, B. P., Gan, S.-Q., differentiation than involucrin, cystatin A, and the small Haynes, S. & Steven, A. C. (1991) Int. J. Biol. Macromol. 13, proline-rich proteins (6, 7, 17, 24). Thus, one possible expla- 130-139. nation of why overexpression of loricrin does not produce a 17. Yoneda, K., Hohl, D., McBride, 0. W., Wang, M., Cehrs, negative phenotype is that loricrin may be deposited over the K. U., Idler, W. W. & Steinert, P. M. (1992) J. Biol. Chem. preexisting scaffold at a very late or the last step of CE 267, 18060-18066. assembly. Accordingly, these conclusions provide support 18. Mack, J. W., Torchia, D. & Steinert, P. M. (1988) Biochemistry for the hypothesis (16) that the intracellular surface of the CE 27, 5418-5426. consists mostly of loricrin and that its main function is to 19. Korge, B. P., Gan, S.-Q., McBride, 0. W., Mischke, D. & Steinert, P. M. (1992) Proc. Natl. Acad. Sci. USA 89, 910-914. provide flexibility to the epidermis by interaction with the 20. Korge, B. P., Compton, J. G., Steinert, P. M. & Mischke, D. immediately subjacent network of keratin intermediate fila- (1992) J. Invest. Dermatol. 99, 697-702. ments in the terminally differentiated cells. 21. Hohl, D., Lichti, U., Breitkreutz, D., Steinert, P. M. & Roop, D. R. (1991) J. Invest. Dermatol. 96, 414-418. We thank Drs. Peter Hoppe (The Jackson Laboratory) and Eric 22. Hogan, B., Costantini, F. & Lacy, E. (1986) Manipulating the Lee (National Institute ofChild Health and Human Development) for Mouse Embryo: A Laboratory Manual (Cold Spring Harbor their assistance in the preparation and maintenance ofthe transgenic Lab. Press, Plainview, NY). mouse lines. We thank Dr. Lyuben Marekov for assistance with the 23. Frohman, M. A. (1990) in PCR Protocols: A Guide to Methods amino acid analyses and Dr. Ulrike Lichti for advice with the cell and Applications, eds. Innis, M. A., Gelfand, D. H., Sninski, culture methods. We are indebted to Drs. John Compton, In-Gyu J. J. & White, T. J. (Academic, New York), pp. 28-38. Kim, Bernhard Korge, Nelli Markova, and Alasdair Steven for their 24. Steven, A. C., Bisher, M. E., Roop, D. R. & Steinert, P. M. unstinting assistance and advice with many aspects ofthis work. Mr. (1990) J. Struct. Biol. 104, 150-162. John Birnbaum kindly assisted with the immunogold electron mi- 25. Dale, B. A., Holbrook, K. A., Hoff, M. & Sun, T.-T. (1983) J. crosopy. Cell Biol. 101, 1257-1269. 26. Crish, J. F., Zaim, T. M. & Eckert, R. L. (1993) Differentiation 1. Hohl, D. (1990) Dermatologica 180, 201-211. 53, 191-200. Downloaded by guest on October 4, 2021