Hair follicle differentiation: expression, structure and evolutionary conservation of the hair type II keratin intermediate filament gene family

Basal cells of hair follicles and the epidermis have a high proliferation rate and undergo programs of continuous or cyclic renewal. Whereas the epidermis is a relatively simple stratified tissue consisting of three to four cell layers, hair is more complex, with at least 10 different cell types involved in an elegant architectural structure (for reviews see, Auber, 1950; Montagna and Parakkal, 1974; Swift, 1977; Powell and Rogers, 1990a). In the epidermis, the proliferative cells in the basal layer divide and differentiate as they move upwards, changing their pattern of keratin gene expression (Sun et al. 1984; Fuchs et al. 1989; Roop et al. 1989; O’Guin et al. 1990). In the hair follicle, the proliferative cells in the follicle bulb give rise to the cell types of the hair shaft, namely the hair cuticle, cortex and medulla, and to the inner root sheath, as distinctive differentiation programs are activated. A particularly complex differentiation pathway is established in the keratinocytes of the hair cortex in which some 50 or more keratin genes, belonging to several multigene families, are sequentially activated (Powell et al. 1992 and unpublished data). During terminal differentiation of the hair shaft keratinocytes, a filament network composed of two families of intermediate filament (IF) proteins is crosslinked with an interfilamentous matrix composed of several families of small proteins. Each protein seems to contain at least 15 cysteine residues (Crewther, 1976; Powell and Rogers, 1990a) and the extensive disulphide-bond cross-linking that occurs during terminal differentiation contributes to producing a tissue that is resistant to physical, chemical and biological agents. Striking protein patterns can be seen in many hairs by electron microscopy (Rogers, 1959) suggesting an orderly assembly of the filament network and the associated proteins. The hair keratin IF proteins appear to be the first differentiation-specific keratins produced when differentiation of the hair shaft keratinocytes commences.

The IF found in the cells of mature mammalian hairs and homy tissue are a special subset of keratin IF, comprising four pairs of cysteine-rich proteins (Crewther et al. 1980; Heid et al. 1986). All IF proteins are characterized by a common secondary structure consisting of a conserved central region of 310–340 amino acids, capable of adopting an α-helical coiled-coil conformation, flanked by end domains that are usually large and non-helical (for review see, Steinert and Roop, 1988). The combined protein and gene information indicate that the IF superfamily contains more than 40 components. Several genes have been sequenced and their common structures and sequence conservation suggest that they could have evolved from a primordial gene. The present IF gene superfamily is composed of at least 6 families and a salient feature is the diversity of expression of the individual genes (Sun et al. 1984; Fuchs et al. 1989; Roop et al. 1989; O’Guin et al. 1990). The keratin IF of epithelia form the largest group by far, comprising two families, each containing 10–15 proteins. The fundamental IF unit is composed of a heterodimer of one type I and one type II protein and approximately equal numbers of genes encode both types of proteins in the genome. Molecular and genetic data indicate that they are located in clusters and different chromosomal locations have been mapped for the epidermal type I IF and type II IF genes (Lessin et al. 1988; Romano et al. 1988; Rosenberg et al. 1988; Nadeau et al. 1989; Popescu et al. 1989). Separate gene clusters have been described for the hair keratin IF genes (Powell et al. 1986) but their chromosomal locations are not known. The expression of several type I and type II genes has been studied in keratinizing skin epithelia where the genes are expressed in type l/type II pairs and the expression of a particular pair is restricted to a stage in epithelial keratinocyte differentiation.

A number of immunocytochemical studies have described the expression of keratin IF proteins in hair follicles (French and Hewish, 1986; Lynch et al. 1986; Heid et al. 1988a, b). However, with one exception, each antibody appeared to recognize most of the components of either the hair type I or type II IF families. A single, specific antibody detected a minor type I IF component in human hair expressed in the upper region of the follicle bulb (Heid et al. 1988a, b). It is clear that the multiplicity of the keratin IF proteins and their sequence similarities has always hampered the generation of protein-specific antibodies and, in the hair follicle, the complexity of cell types which express IF proteins adds a further complicating dimension. A complementary approach, better suited to analysing the expression of keratin genes in the hair follicle in the absence of appropriate antibodies, is RNA in situ hybridization and we describe here the sequential expression of individual components of this gene family in hair keratinocyte differentiation and the first sequence of a hair type II keratin IF gene, KII-9.

We have adopted a new nomenclature in this paper. The present keratin IF nomenclature is limited in its capacity to incorporate new genes and the hair keratin gene nomenclature is cumbersome and would benefit from rationalization. We propose a simple modification of the existing nomenclature of Moll et al. (1982) which would create a flexible system that could readily incorporate new keratin IF genes and we have adopted this scheme in the present paper. The hair proteins belong to the IF protein superfamily and it is particularly important that a flexible, coherent nomenclature is adopted before any more hair IF gene sequences with diverse names are published. For example, published names for the wool type II keratin IF protein family are; a-keratin, a generic term including both type I and type II proteins (Fraser et al. 1972), low sulphur component 7a, 7b, 7c and 5 (Crewther, 1976) for the proteins, 09 for a cDNA (equivalent to protein 7c: Ward et al. 1982), and B and D for two of the genes (gene D is equivalent to the 09 cDNA: Powell et al. 1989). Each designation in the proposed system would be of the form, Kn-m.x where K denotes keratin, n is either I or II, denoting type I or type II, the m value is the current catalogue number for published genes or a new number for new genes and the x value, if necessary, would denote a variant of an existing gene. The current numbering system for the human keratin IF catalogue could be retained in each designation to preserve familiar associations and new genes or proteins would be awarded the next available ‘m’ number. Thus, as referred to in this report, K14, a type I IF keratin, becomes KI-14 (gene) or KI-14 (protein), K5, a type II keratin IF, becomes KII-5 (gene) or KII-5 (protein) and K6b becomes KII-6.2 (gene) or KII-6.2 (protein). The hair type II keratin IF genes described in this report have the following designations; KII-9 (gene B in Powell et al. 1989) and KII-10 (gene D in Powell et al. 1989; 09 cDNA in Ward et al. 1982; component 7c in Crewther et al. 1976). The genes encoding two partial cDNAs described here are KII-11 and KU-12 and the genes designated KII-13, KU-14 and KU-15 are A, C and E respectively in Powell et al. (1989).

For sequencing, appropriate DNA restriction fragments were either cloned into M13mpl8 or 19 vectors (Norrander et al. 1983) for direct sequencing, or deletion clones were generated by the DNAase I deletion method (Anderson, 1981). Singlestranded M13 template DNA was prepared by the method of Winter and Fields (1980). All sequencing was performed by the dideoxy chain termination method of Sanger et al. (1980) using either Bresatec (Adelaide, South Australia) or USB Sequenase sequencing kits and [α-³²P]dATP (3000 Ci/mmol: Bresatec). The DNA was sequenced in both directions. Double-stranded DNA fragments were labeled with ³²P by the oligolabeling method of Feinberg and Vogelstein (1983) using a Bresatec kit.

Primer extensions were performed with 10 μg sheep follicle RNA as described by Kuczek and Rogers (1987) using a 20-mer primer, 5′-GCAGGTCATGATCCTTCTGG-3’ specific for the 5′-noncoding region of the gene (see Fig. 3). RNAase protection analyses using 2 μg of sheep follicle RNA perprotection assay were performed as described by Kreig and Meiton (1987).

Southern transfers of DNA onto Zeta Probe membrane (BioRad) were performed by the alkali method as described by Reed and Mann (1985). Briefly, filter-bound DNA was prehybridized for at least 2 hours at 41°C in 47% formamide, 10% dextran sulphate, 3 ×SSPE, 1% SDS, 0.5% Blotto, 0.5 mg/rnl sonicated salmon sperm DNA, then hybridized overnight in the same solution with labeled probe. The final stringency of the posthybridization washes is given in the figure legends. Northern blots were performed as described by MacKinnon et al. (1990).

In situ hybridizations on paraformaldehyde-fixed and sectioned sheep wool follicle biopsies were performed as described by Powell and Rogers (1990b). RNA and cRNA probes labeled to high specific activity using [α-³⁵S]UTP (1,350 Ci/mmol; New England Nuclear, Boston, MA) were synthesized with either T7 or SP6 RNA polymerase by the method of Krieg and Melton (1987) using a kit obtained from Bresatec.

Probes for the various IF genes described below were subcloned into pGEM2 vectors (Promega Biotec).

The KII-9 3′-noncoding probe; a 196 bp fragment from the 3′-noncoding and flanking region of Kll-9 (nucleotides 62566451 in Fig. 3).

The KII-10 3′-noncoding probe; a 220 bp Pstl fragment from a type 11 IF cDNA clone, equivalent to KII-10 (Powell et al. 1986), including 18 bp of the C-terminal domain of KII-10 and 200 bp of the 3′-noncoding region.

The KII-11 3′-noncoding probe; a 120 bp Smal fragment.

The type I IF gene family probe; a 435 bp Pstl fragment encoding part of the a-helical region of a wool follicle type 1 IF cDNA clone (Powell et al. 1986; Wilson et al. 1988). A comparison of two sheep type I IF wool keratin genes over the probe region shows 92% nucleotide similarity.

The type II IF gene family probe; a 170 bp Pstl fragment encoding part of the α-helical region of a wool follicle type II IF cDNA clone (Powell et al. 1986). In comparing four sheep wool type II keratin IF genes for which sequences are available over this region there is at least 92% nucleotide similarity (KII-9 - this report; KII-10 cDNA clone - K Ward et al. unpublished data; KII-11 and KII-12 cDNAs - B. Powell, unpublished data).

One of two cosmids previously identified as containing sheep wool type II IF genes (Powell et al. 1986) has been extensively sequenced. Two genes were originally mapped in this cosmid using a partial cDNA clone as probe but fine mapping with exon probes from the sequenced KII-9 gene has revealed the 5′ end of a third type II IF gene in this cosmid (KII-14:Figs 1 and 2). An exon 1 probe detects three EcoRI fragments (Fig. 1); the 3.9 kb fragment containing the 5′ end of KII-14 was not previously detected with the short cDNA probe used to isolate and initially map this cosmid (Powell et al. 1986). Partial sequencing of KII-14 has confirmed that it is another related type II IF gene (B. Powell, unpublished data).

The complete KII-9 gene sequence is shown in Fig. 3. From the sequence, a mRNA of about 2 kb is predicted, encoding a basic protein of 506 amino acids with a M_r of 55 ×10³. Sequence comparisons indicate that the predicted protein belongs to the type II keratin IF protein class and it is very similar to the sequence of a wool type II keratin IF protein (KII-10: component 7c, Sparrow et al. 1989). Additionally, the sequence of the KII-9 gene reveals a similar pattern of exon/intron boundaries and protein domain arrangement to other type II keratin IF genes. The gene is split into 9 exons, and all the introns interrupt the coding region of the gene at exactly the same phase of the triplet codon as found in the human epidermal KII-5 gene (Lersch et al. 1989) and KII-6.2 gene (Tyner et al. 1985).

The KII-9 gene produces an abundant transcript in wool follicle RNA and northern analysis with a genespecific probe detects an mRNA of 2.3 kb in size (Fig. 4A). The start of transcription of the gene was determined by primer extension analysis. A genespecific 20-mer priming from the 5′ non-coding region produced strong extensions on sheep wool follicle RNA (Fig. 4B). There were two extension products, the major one 2 bp shorter than the other, indicating minor heterogeneity in the start site of the mRNA. 5′ noncoding regions of 63 bp and 65 bp are predicted. The 3′ end of the mRNA was identified by RNAase protection experiments (Fig. 4C). A gene fragment spanning the putative polyadenylation signal was used in an RNAase protection experiment with sheep follicle RNA. Two fragments of 225 bp and 230 bp were protected (Fig. 4C) indicating that the mRNA terminates 15–20 bp downstream of the AATAAA motif and has a 3′ non coding region of about 400 bp. Exclusive of the poly(A) tail, a mRNA of 2 kb is predicted.

Upstream of the transcription start site are CAAT box and TATA box-like sequences. A putative polyadenylation signal is found 378 bp downstream of the protein termination codon and about 90 bp 3′ to that is a 24 bp palindrome which could form a stem-loop structure. Interestingly, this palindrome appears to cause premature transcription termination by SP6 and T7 RNA polymerases in vitro (B. Powell, unpublished data).

The 506 amino acid protein predicted from the KII-9 gene is very similar to the 491 amino acid protein (KH-10; component 7c) isolated from sheep wool and sequenced by Sparrow et al. (1989). The N-terminal domain of KII-9 is four amino acids shorter than the domain in KII-10 and the C-terminal domain is 19 amino acids longer. In total there are 57 amino acid differences between the two proteins (see Figs 3 and 5) and the differences in the N- and C-terminal domains are clustered to the ends of those domains. Complete C-terminal domain sequences are now available for three hair type II IF proteins (Fig. 5). They are 69, 71 and 90 amino acid residues in length and each has an overall cysteine composition of 14%. Although the latter half of each domain shows considerable sequence variation the proteins terminate in a conserved dipeptide, a basic residue followed by cysteine.

Two partial cDNA clones (KII-11 and KII-12) have also been isolated from a wool follicle cDNA library using the wool KII-10 clone (B. Powell, unpublished data) and when the sequences for the predicted proteins are compared with the two complete sequences (KII-9 and KII-10) it is clear that they are a family of proteins (Fig. 5). The KII-11 cDNA encodes 153 amino acids of a type II IF protein, including part of the a-helical domain and all the C-terminal domain. The protein predicted from this clone is closely related to KII-9 and KII-10; for example, it shows only one amino acid difference, compared to 8 in the same region of the predicted KII-12 protein (Fig. 5). The KII-12 cDNA encodes 286 amino acids of the central a-helical domain of a type II IF protein and differs substantially from KII-9, KII-10 and KII-11 (Fig. 5).

Expression of this type II keratin IF gene family in hair follicle development was examined in a number of different hair types. Specific cRNA probes for three hair type II keratin IF genes and a general gene family probe were hybridized to follicle tissue sections. Our in situ data using these probes, in combination with genespecific probes, show that this gene family is expressed only in the hair shaft keratinocytes and not in other hair or epithelial cell lineages in the skin. The spatial pattern of expression found for the IF gene whose complete sequence is reported here is typical of the gene family although differential timing of expression of at least two other hair type II IF genes occurs (see below). A gene-specific 3’ noncoding region probe for KII-9 hybridized to the cortical cell keratinocytes of the wool follicle (Fig. 6). Hybridization grains were first detectable in the cells above the dermal papilla, in general within three cells distance above the basement membrane that separates the follicle bulb from the apex of the dermal papilla (Figs 6 and 9). The RNA hybridization signal was detectable until well up the hair shaft and examination of the stage of follicle development by multichrome staining (Auber, 1950) indicated that it persisted into the keratinization zone. No signals were observed in the outer root sheath or inner root sheath cells and no signals were observed in the medulla when medullated fibres were analysed (Fig. 7). Furthermore, no hybridization was detectable in any epidermal cells (data not shown).

The expression patterns obtained with three genespecific probes are shown in Figs 7, 8 and 9. The specificity of each probe was established by genomic Southern blots (data not shown). Cortical cell hybridization signals were observed for all probes. In situ hybridizations performed to serial follicle cross-sections from the upper bulb region with the three gene-specific probes suggest that expression of KII-11 is delayed relative to the others (Fig. 8). Immediately preceding the section in which KII-11 expression was examined and could not be detected (Fig. 8B,F), and therefore at a lower point in the follicle bulb, we detected expression of KII-10 (Fig. 8A,E). Comparison of KII-10 and KII-11 expression in the upper region of the follicle bulb in serial follicle longitudinal sections (Fig. 9E-H) clearly indicates that the initiation of KII-10 expression precedes that of KII-11. Serial sections probed with KII-9 and KII-10 show no difference in the start of their expression (Fig. 9A-D). Thus, it appears that the spatial and stage-specific expression patterns of KII-9 and KII-10 are probably identical, whereas the hybridization signals obtained with the KII-11 probe suggest that it is transcriptionally activated later.

General probes were constructed to examine the expression of the hair type I and type II IF gene families. Each probe encoded part of the a-helical domain and was highly conserved (⩾92% similarity) between the known wool IF gene sequences in each family. When serial longitudinal sections were probed a bilobed appearance of gene expression extended down around the apex of the dermal papilla and was more noticeable with the type II IF probe (Fig. 9I-L). In serial cross-sections through the upper part of the follicle bulb and dermal papilla expression of both type I and type II IF genes was visible and preferentially occurred on one side of the bulb (Fig. 10). This pattern may indicate that cells on one side of the follicle bulb differentiate first, or may simply reflect the normal deflected angle of the follicle bulb relative to the hair shaft in wool follicles, in which case it is likely that hair IF gene expression is simultaneously activated in cells equidistant from the dermal papilla (Fig. 9 and see Fig. 11).

Hair keratin gene transcription is first detected in cells around the apex of the dermal papilla and dramatically increases in cells above the apex (Fig. 9I-L). In contrast to the expression pattern shown by general hair IF gene probes, expression of KII-9 and KII-10 starts in cells about two to four cells above the dermal papilla (see Figs 6 and 9A-D) and transcription of KII-11 starts even later (Fig. 9G,H). To account for this differential hybridization between the general family probe and the gene-specific probe, another related gene must be transcriptionally activated first. As there appear to be only four wool type II IF proteins (see above) the gene encoding the KII-12 cDNA described in this report, a fourth member of the hair type II IF family, is a prime candidate as the first type II gene to be activated. The patterns of expression revealed with the gene-specific probes suggest that the genes of the hair type II keratin IF family are sequentially activated and coexpressed in the same cells (Fig. 11).

The evolutionary conservation of the hair type II keratin IF family was compared between genomes from the two more recent branches in mammalian evolution, the placental mammals and the marsupial mammals (Fig. 12). Three DNA probes, each encoding separate domains of the protein, were used. The N-terminal, a- helical and C-terminal probes detected multiple restriction fragments in all tracks, documenting the conservation of hair type II IF genes in mammalian evolution. The N-terminal domain probe (220 bp) encoding the first 64 amino acids of KII-9 detected eight sheep EcoRI fragments, four strongly and four weakly. Seven fragments were readily detected in human DNA but fewer and fainter fragments were detected in the more distantly related marsupial genomes, generally two to four fragments. The four strongly hybridizing sheep fragments correspond to the four type II IF genes, namely KII-9, KII-10, KII-13 and KII-14, located in the two cosmid clones previously isolated (Powell et al. 1986). The four fainter bands detected in the sheep genomic Southern blot (Fig. 12) could represent weakly homologous N-terminal domains from other type II IF genes, possibly belonging to the genes specifying the two related wool follicle type II IF cDNA clones, KII-11 and KII-12 (see Fig. 5).

A similarly complex pattern is seen with each of the other two domain probes (Fig. 12B,C). The increased complexity of the blot with the a-helical domain probe is likely explained by two features of this probe. The probe covers one of the most conserved regions of type II IF genes, which is the sequence encoding the 3’ end of the α-helical domain, and we expect it to detect type II IF genes expressed in other keratinizing epithelia. Secondly, the probe spans two intron locations found in all type II IF genes to date and could detect a number of EcoRI restriction fragments. Although all introns sequenced in this region appear to be small and for KII-9 and two other sheep type II IF genes (KII-13 in this cosmid and KII-11 in another sheep cosmid [Powell et al. 1986 and B. Powell; unpublished data]) a single EcoRI fragment is detected, the possibility remains that other type II IF genes could contain EcoRI sites within this region.

The C-terminal domain probe (324 bp) encoding 77 amino acids from the end of the KII-9 protein detected six sheep EcoRI fragments (Fig: 12C). The two strongest are derived from KII-9, reported here, and KII-10, in another sheep cosmid (Powell et al. 1986). In comparing equivalent sequences from the probe and the KII-11 cDNA there is only a 55% similarity over the C-terminal domain region covered by the probe and therefore the gene fragment specifying the KII-11 cDNA is likely to be one of those weakly hybridizing fragments. Once again, a number of fragments were readily detected in human DNA and weakly in the other mammalian DNAs examined.

The hybridization data indicate that at least two type II keratin IF subgroups exist within the sheep genome and are expressed in the wool follicle. In the other mammalian genomes examined, we do not see two distinct groups of hybridizing fragments and, with the exception of the human blots, at most five fragments were detected. Those sets of fragments may represent the genes most closely related to the sheep KII-9 probes in the other genomes and the subgroup represented by the weakly hybridizing fragments in the sheep DNA could be too different in those genomes to be detected under our hybridization conditions.

The formation of hairs involves the terminal differentiation of several cell types from a population of mitotically active cells in the follicle bulb. keratinocytes of the differentiating hair shaft express a number of keratin proteins (Crewther, 1976; Gillespie, 1983) and these are produced from several multigene families (for review see Powell and Rogers, 1990a) in specific sequential transcription patterns (Powell et al. 1992 and unpublished data). During this process the type II IF proteins interact with their type I counterparts to create a filament scaffold to which a number of families of small cysteine-containing proteins are believed to complex in the later stages of differentiation. We have characterized the expression of one of the major gene families in hair follicle differentiation, the hair type II keratin IF gene family.

Two families of hair keratin IF proteins are present in mammals (Marshall and Gillespie, 1977; Crewther et al. 1980; Heid et al. 1986; Powell and Rogers, 1986) and they appear to be subsets of epidermal keratin IF (for review see Powell and Rogers, 1990a). Their N- and C- terminal domains are immediately distinctive, with a high proportion of cysteine residues, and a mathematical analysis of keratin IF sequence similarities in the α- helical rod region confirms a subset classification (Conway and Parry, 1988). In each hair keratin family there are four predominant proteins and an additional minor component has been found in human and bovine hair but not in sheep wool (Heid et al. 1988a). Three hair type I keratin sequences have been published, a protein, a gene and a cDNA, two representing sheep wool proteins (Dowling et al. 1986; Wilson et al. 1988) and one representing a mouse hair protein (Bertolino et al. 1988). The complete gene sequence encoding a hair type II IF protein (KII-9) is reported here, and the partial protein sequences of two related proteins (KII-11 and KII-12) derived from cDNA clones, in combination with the sequenced wool KII-10 protein (component 7c; Sparrow et al. 1989) represent a family of four hair type II keratin proteins (Fig. 5). However, the hair type II keratin IF gene family can be divided into at least two subgroups, classified by different sequences encoding the protein N- and C-terminal domains. Within the subgroup typified by KII-9, two genes are expressed in the hair follicle, Kll-9 and KII-10, and there are at least two other genes, KII-13 and KII-14, which are not expressed in the follicle.

One of the first markers of hair keratin gene expression appears to be the transcription of the IF genes. We have compared the expression of the hair type I and type II keratin IF gene families and examined the expression of three type II IF genes in detail. Our in situ hybridization studies indicate that expression of some of the hair type I and type II keratin IF genes commences in follicle bulb cells located near the dermal papilla (Figs 9I-L and 10) and that differential expression of hair type II genes occurs during hair keratinocyte differentiation (Figs 8 and 9).

Two principles of gene expression have emerged from the epidermal keratin IF genes that have been studied; namely, coexpression of type I and type II genes and differential expression of genes during the movement of epidermal cells from the basal layer (Sun et al. 1984; Fuchs et al. 1989; Roop et al. 1989; O’Guin et al. 1990). In the epidermis the keratin genes expressed in the basal keratinocytes are down-regulated and new keratin genes are transcriptionally activated as the cells move upwards. The hair type II keratin genes are sequentially activated but do not appear to be down-regulated during follicle growth.

A number of immunocytochemical studies have shown the expression of keratin IF proteins in hair follicles (French and Hewish, 1986; Lynch et al. 1986; Heid et al. 1988a, b). With one exception, each antibody appeared to recognize most of the components of either the hair type I or type II IF families. The single, specific antibody produced in those studies detected a minor type I IF component in human hair (Heid et al. 1988a, b). The initial staining in the follicle bulb cells with the general antibodies was weak but soon increased in cells around and above the apex of the dermal papilla. Whereas some staining patterns (French and Hewish, 1986; Heid et al. 1988a, b) began in the middle of the follicle bulb other antibodies did not stain those lower bulb cells, but first showed staining in the upper bulb cells around the apex of the dermal papilla (Lynch et al. 1986). Differences in sensitivity may account for these variations or, alternatively, differences in antibody specificity. It is clear that the multiplicity of the keratin IF proteins and their sequence similarities has always hampered the generation of protein-specific antibodies and, in the hair follicle, the complexity of cell types that express IF proteins adds a further complicating dimension. RNA in situ hybridization, a complementary approach to immunolocalization, can be a more specific approach to analysing the expression of keratin genes in the hair follicle. Our data with general hair IF and gene-specific probes unequivocally demonstrate sequential expression of hair type II IF genes in follicle differentiation (Fig. 11).

The hair cells that express these keratin IF genes are differentiating as they move upward from the follicle bulb. The first cells in which hair keratin gene transcripts were detected were in the middle bulb region around the top of the dermal papilla (Fig. 10). Similar results were observed in rat hair follicles with a longer sheep wool type I keratin IF gene probe (Kopan and Fuchs, 1989). In relative distance from the top of the dermal papilla expression of Kll-9 and KII-10 starts two to four cells above it and KII-11 starts another two to four cells later. With the general antibodies staining was first detectable in cells at least two to three cell layers removed from the dermal papilla, although in a few instances very weak staining of some cells adjacent to those in contact with the basement membrane was seen (Heid et al. 1988a, b; Moll et al. 1988). A type I keratin IF protein present in low abundance in human hair was shown to be expressed at a late stage in follicle differentiation (Heid et al. 1988a) like the sheep KII-11 gene (Fig. 9G,H).

The hair keratin IF gene families are not expressed in the epidermis, and in the follicle their expression is restricted to the cells of the upper bulb and hair shaft keratinocytes. We have not seen any hybridization to the medulla with our hair IF gene probes (Fig. 7). Antibody studies of human and bovine hair follicles indicate that hair-related keratin IF are also expressed in the cuticle cells (Heid et al. 1988a). From our in situ data it is not clear whether the hair keratin IF genes are expressed in the hair shaft cuticle cells. Because the wool cuticle is composed of only a single layer of cells, we have not been able to distinguish between expression in adjacent cortical and cuticle cells with ³⁵S-labeled probes when there is expression in the cortex. Other ‘hard’ keratin structures, such as nail, hoof and horn have previously been shown to possess proteins with similar characterisitics to hair keratin IF (Marshall and Gillespie, 1977; Marshall, 1983; Baden and Kubilus, 1984). Hair-related proteins are found in the keratinising cells of hoof and nail (Lynch et al. 1986; Heid et al. 1988b; Moll et al. 1988), tongue (Heid et al. 1988b; Dhouailly et al. 1989) and possibly even in some thymic cells (Heid et al. 1988b). One hair-related protein expressed in the tongue appears to be a novel type I keratin IF, larger than any already identified (Dhouailly et al. 1989). We have recently found that KII-10 is expressed in sheep hoof and in the embryonic claw epithelium of transgenic mice (B. Powell, unpublished data). The two hair-like type II keratin IF genes in our cosmids that are not expressed in the hair follicle (KII-13 and KII-14, see above) are also candidates for expression in these epithelia. In one cosmid we sequenced the 5′ end of a novel hair-like type II IF gene which is not expressed in the hair follicle (KII-15’. Powell et al. 1989 and B. Powell, unpublished data). The predicted N-terminal domain of KII-15 contains five cysteine residues, compared to the 10 or more present in KII-9 and KII-10 and, interestingly, its predicted amino acid sequence looks like a hybrid between a hair and an epidermal type II protein, the human KII-6.2 protein (Tyner et al. 1985). Clearly there are hair or hair-related genes that are expressed in a diversity of epithelia and to positively identify the keratin IF genes expressed in these epithelia in situ hybridizations with gene-specific probes are required.

The keratin IF gene family is a multigene family whose genes are expressed in a variety of different epithelial cell types and under specific expression programs during the differentiation of those cells. The gene sequences and transcription factors that regulate these expression patterns are not precisely known, but rapid advances are being made towards discovering them. In hair keratinocyte differentiation the coexpression of hair type I and type II IF genes and the differential expression of the type II genes and several other families of genes encoding the hair matrix proteins (Powell et al. 1992 and unpublished data) point to the operation of complex regulatory mechanisms.

Inspection of the promoter region of the hair KII-9 gene and comparisons with a number of other hair keratin genes from sheep has revealed several conserved sequence motifs which could be involved in the regulation of hair keratin gene expression (Fig. 13 and see Powell et al. 1992). One interesting motif, a 9 bp nearly palindromic sequence, 5′-ClTTGAAGA-3′, is located 209 nucleotides upstream of the transcription start site of KII-9. This motif, denoted HK1, was initially identified in a survey of 15 hair keratin genes, being found in the promoter regions of six genes, including keratin IF and keratin IF-associated genes, and located between 180 and 240 bp upstream of the transcription start sites (Powell et al. 1992). It is well conserved, with only three nucleotide mismatches amongst the several copies. It is not present in the proximal promoter regions of any epidermal keratin IF genes but, intriguingly, the 5’ half of the HK1 motif is present in some putative regulatory elements in a recently compiled vertebrate gene survey (Locker and Buzard, 1990) and in the promoters of three human keratin IF genes (Johnson et al. 1985; Tyner et al. 1985; RayChaudhury et al. 1986).

To date, two motifs have been found in the promoters of epidermal keratin IF genes and shown to be involved in DNA-protein interactions (Leask et al. 1990; Snape et al. 1990). Both are similar to the AP2 consensus sequence and do not direct tissue-specific expression but seem to have a more general role in keratin gene transcription. The HK1 motifs in the promoters of the two expressed hair type II keratin genes {KU-9 and KII-10) are part of longer regions of DNA conservation which include API and KTF1 motifs (Snape et al. 1990), and both could be potential regulatory elements of these genes (Fig. 14). The KII-9 sequence is similar to the API motif, whereas the KII-10 sequence is more like the KTF1 motif.

In addition to the sequences discussed above there are potential binding sites for API and AP2 elsewhere in the hair KII-9 promoter. The API motif is palindromic and one of the slightly variant API motifs located in the first intron of KII-9 is part of a much longer palindromic sequence (Fig. 13). Interestingly, that sequence is conserved in three other sheep type II genes (KII-10, KII-13 and KII-14;, B. Powell, unpublished data). Both KII-9 API motifs differ in only the last nucleotide from the highly conserved API consensus sequence identified by Risse et al. (1989). Significantly, FOS protein has recently been demonstrated in rat hair follicles by immunohistochemisty (Fisher et al. 1991) and its expression pattern seems to parallel our in situ hybridization data with the hair keratin IF gene probes. With the finding of consensus API motifs in the promoter regions of the hair Kll-9 and KII-10 genes (Fig. 14 and Powell et al. 1992), these data suggest a role for API in hair keratin gene transcription.

In hair keratinocyte differentiation both coordinate and differential expression patterns of hair type II IF genes have been identified by in situ hybridization. It is now important to determine which of the conserved DNA motifs located in the promoters of these genes are involved in the regulation of hair keratin expression.

We thank Dr David Hayman for providing the marsupial DNAs, Dr Elizabeth Kuczek for the northern blot analysis, Elaine Batty for her skill in producing longitudinal follicle sections from sheep skin biopsies, Antonietta Nesci for help with the in situ hybridizations, Michael Calder for photographic assistance and Simon Bawden for comments on the manuscript. We also thank Dr. David Parry for his comments on our suggestions for a unified keratin IF protein and gene nomenclature. This work was supported by a grant from the Wool Research Trust Fund on the recommendation of the Australian Wool Corporation.