Nerve Growth Factor Binds to Normal Human Keratinocytes through High and Low Affinity Receptors and Stimulates Their Growth by a Novel Autocrine Loop*

Normal human keratinocytes synthesize and secrete biologically active nerve growth factor (NGF) in a growth J. Biol. Chem. 266, 21718-21722). Here we show that the same human keratinocytes bind NGF via low and high affinity receptors. In parallel with the course of NGF synthesis, the expression of low affinity NGF receptor ( ~ 7 5 ~ ~ ~ ‘ ) decreases when a confluent, differentiated, and fully stratified epithelium is obtained. In skin sections, is present in basal keratinocytes and absent from suprabasal, terminally differentiated cells. The trkA protooncogene product (p140trkA), a component of the NGF receptor, is not expressed by keratinocytes. Instead, keratino- cytes express a new member of the trk family (that we termed trkE), which generates 3.9-kilobase tran- scripts. Keratinocyte-derived NGF plays a key role in the autocrine epidermal cell proliferation. This has

(1991)), responsible for the maintenance, development, and differentiation of several neural crest-derived cell types Barde, 1980, Levi-Montalcini, 1987) and central cholinergic neurons (Korshing et al., 1985). NGF, synthesized by and secreted from target tissues of sympathetic, sensory, and cholinergic basal forebrain neurons (Thoenen and Barde, 1980;Levi-Montalcini, 1987), binds to both low and high affinity receptors (formed by ~7 5~~~' and p140NGFr/ t r M protooncogene product) (Johnson et al., 1986;Kaplan et al., 1991;Hempstead et al., 1991), transduces appropriate signals in nerve terminals (Berg et al., 1991;Cordon-Cardo et al., 1991;Hempstead et a[., 1991;Kaplan et al., 1991), is internalized in membrane-bound vesicles of peripheral nerves and transported along microtubules up the axon to the cell body, where it exerts its biological activity (Thoenen and Barde, 1980). In agreement with a target derived neurotrophic role for NGF, a good correlation exists between levels of NGF in target tissues and degree of their innervation by NGF-sensitive fibers (Levi-Montalcini (1987), but see Davies et al. (1987)). However, in male mouse submandibular glands, guinea pig and rabbit prostate, the expression of NGF does not correlate with the level of innervation, suggesting that NGF exerts non-neurotrophic activities in these tissues, and that it evokes biological responses also in non-neuronal cells (for review, see Levi-Montalcini (1987)). Indeed, NGF affects growth and histamine release from mast cells (Levi-Montalcini, 1987), modulates human B lymphocyte differentiated functions (Otten et al., 1989), and regulates the onset of meiosis in rat seminiferous epithelium through Sertoli cells (Parvinen et al., 1992).
Human keratinocytes and lining epithelial cells can be serially cultivated (Rheinwald and Green, 1975;De Luca et al., 1990a;Romagnoli et al., 1990). Under the appropriate culture conditions (Green et al., 1979), keratinocytes reconstitute in vitro transplantable sheets of epithelium and maintain virtually the same differentiation features and gene expression pattern of their in vivo counterparts (for reviews, see Green (1980) and De Luca and ), such as to represent an ideal system to investigate epidermal physiology. We and others have shown that epidermal keratinocytes regulate melanocyte growth and differentiation through the paracrine secretion of several polypeptides (De Luca, 1988a, 1988bGordon et al., 1989;Halaban et al., 1988;De Luca et al., 1993). In particular, basal keratinocytes synthesize and secrete biologically active NGF (Di Marco et al., 1991;Yaar et al., 1991), which plays a pivotal role in regulating melanocyte migration and differentiation in epidermal morphogenesis (Yaar et al., 1991;, and induces the widespread sprouting of sensory nerve fibers during wound healing. ' In this paper we show that (i) normal human keratinocytes themselves bind NGF through both low and high affinity receptors, (ii) keratinocytes express the ~7 5~~~' and a new member of the trk family (that we termed trkE), but not the trkA protooncogene, (iii) NGF, in addition to its well known effects as a survival and differentiation factor, is a potent autocrine stimulator of keratinocyte proliferation, (iv) K252a, a selective inhibitor of the tyrosine protein kinase activity of the trk family of protooncogenes Berg et al., 1992;Knusel and Hefti, 1992), is sufficient to nearly abolish the autocrine keratinocyte growth stimulation.
Hybridoma cells (aD11) were subcloned and cultured in DMEM containing 10% fetal calf serum, glutamine (4 mM), and penicillinstreptomycin (50 IU/ml). Clones used in these experiments were selected by their capacity of inhibiting NGF induced neurites in chick embryo dorsal root ganglia (Di Marco et al., 1991). Supernatants were prepared by conditioning hybridoma cell medium for 72 h in DMEM, glutamine (4 mM), and penicillin-streptomycin (50 IU/ml) and tested on chick embryo dorsal root ganglia before use.
For equilibrium binding studies, two ranges of lZ5I-NGF concentrations (4-100 p~ and 1-15 nM) were used to explore high and low affinity sites. Subconfluent primary or secondary cultures of normal human keratinocytes were trypsinized as described above. Isolated keratinocytes were plated (1 X lo5 cells/well) in 24-well plates in keratinocyte growth medium, in the absence of fetal calf serum and feeder-layer. After 16 h, cells were washed twice in binding buffer (ice-cold DMEM containing 0.1% bovine serum albumin). 9 -N G F solutions were prepared in binding buffer and added (150 pl, final volume) to keratinocyte cultures. After 2 h at 4 "C under gentle agitation, cells were rinsed five times in binding buffer and solubilized in 0.5 ml of 1 N NaOH for 4 h at 37 "C. The solutions were then ycounted (Gamma 5500, Beckman Instruments, Inc.). Nonspecific binding was measured in the presence of 100 pg/ml nonradioactive NGF. Scatchard analysis was performed following the EBDA/LI-GAND program (McPherson, 1985). For nonequilibrium binding studies (aimed to specifically detect high affinity binding sites), cells were incubated in the presence of lZ5I-NGF (4-100 PM) for 5 min at 4 "C under gentle agitation. Nonspecific binding was measured in the M. De Luca and R. Cancedda, unpublished data. Dr. A. Cattaneo, personal communication.
presence of 100 pg/ml nonradioactive NGF. Samples were then processed as above.
RNA Blotting-The EcoRI 0.8-kilobase fragment of human ~7 5~'~' cDNA was a gift from Dr. Moses Chao, Cornel1 University Medical College, New York. The BamHI-EcoRI fragment of the plasmidpDM-17, corresponding to the entire cytoplasmic portion of the human p14ONGF'/trlzA protooncogene (p14OtrW), was a gift from Dr. Mariano Barbacid, Bristol-Myers Squibb Pharmaceutical Research Institute, Princeton, NJ.
Total cellular RNA was isolated by lysing cultured keratinocytes in primary or secondary culture with 4.2 M guanidine thiocyanate followed by cesium chloride gradient centrifugation as described (Di Marco et al., 1991). Poly(A+) RNA was prepared by an oligo(dT)cellulose column (Pharmacia) chromatography. Two mg of total cellular RNA, dissolved in 1 ml of 20 mM Tris-HC1, pH 7.4, 2 mM EDTA, 0.1% SDS, were denatured in boiling water for 2 min followed by addition of an equal volume of 1 M sodium chloride. The sample was then applied to the column and the poly(A-) fraction was eluted with 50 ml of 10 mM Tris-HC1, pH 7.4, 1 mM EDTA, 0.5 M sodium chloride, 0.1% SDS. The poly(A+) fraction was eluted using 2 ml of H20 kept at 37 "C and precipited with 300 mM sodium acetate, pH 5, and 2.5 volumes of absolute ethanol. Twenty pg of total RNA or 5-10 pg of poly(A+) RNA were size fractionated through a 1% agarose gel and transferred to a nylon membrane (Genescreen Plus, Du Pont-New England Nuclear) in 1.5 M sodium chloride, 0.15 M sodium citrate. After immobilization by shortwave UV exposure, blots were prehybridized at 42 "C for 3 h in 50% deionized formamide, 0.75 M sodium chloride, 25 mM sodium phosphate, 5 mM EDTA, 0.2 mg/ml salmon sperm DNA, 0.5% SDS. Hybridization buffer was identical to the above buffer with the addition of the indicated (see "Results") 32P-labeled probes (2 X lo6 cpm/ml) and 10% dextran sulfate. A final wash was done at 65 "C for 30 min in 15 mM sodium chloride, 1 mM sodium phosphate, 1 mM EDTA, 0.1% SDS. All filters were autoradiographed on x-ray films (hyperfilm-MP Amersham) with intensifying screens at -70 "C.
Polymerase Chain Reaction (PCRI-Reagents were purchased from Perkin-Elmer. PCR was performed using two distinct oligonucleotide primers, i.e. 5'-GTCTTCCTTGCTGAGTGC-3' (3a) and 5'-CTTG-GCATCGGGTCCATG-3' (4a). One pg of total RNA obtained from K562 cells and normal human keratinocytes was reverse transcribed using 2.5 p M random hexamers or 0.75 p M 4a downstream from oligonucleotide in 5 mM MgCl, 5 mM KCl, 10 mM Tris-HC1, pH 8.3, dNTPs, 1 mM each, 1 unit of RNase inhibitor, and 2.5 units of reverse transcriptase in a final volume of 20 pl. The reverse transcription reaction was performed in a Perkin-Elmer Cetus DNA thermal Cycler at 42 'C for 15 min, 99 "C for 5 min, and then soaked at 4 "C for 5 min. For the PCR reaction, downstream and upstream primers were used at a concentration of 0.15 mM in the presence of 2 mM MgCl, 50 mM KC1, 10 mM Tris-HC1, pH 8.3, and 2.5 units of AmpliTag DNA polymerase. The downstream primer (4a) was omitted in the PCR reaction when used in the reverse transcription reaction. The mixture was subjected to PCR amplification in the Perkin-Elmer thermal Cycler apparatus for 35 cycles. The temperatures were: 2 min at 95 "C and 1 min at 60 "C for 35 cycles, 7 min at 60 "C for 1 cycle. Ethidium bromide-stained 1.5% agarose gel was used to visualize PCR fragments.
Indirect immunofluorescence on cell cultures was performed as previously reported (De Luca et al., 1990b;Marchisio et al., 1991). Briefly, the primary antibody (10-30 pglml) was layered on fixed and permeabilized cells and incubated in a humid chamber for 30 min.
After rinsing in PBS, 0.2% bovine serum albumin, coverslips were incubated in the appropriate rhodamine-tagged secondary antibody (Dakopatts, Copenhagen, Denmark) for 30 min at 37 "C in the presence of 2 pg/ml fluorescein-labeled phalloidin (Sigma). Coverslips were mounted in Mowiol (Hoechst AG, Frankfurt/Main, Germany) and observed in a Zeiss Axiophot photomicroscope equipped with epifluorescence lamp and usually with planapochromatic oil immersion lenses. Fluorescence images were recorded on Kodak T-Max 400 films exposed at 1000 IS0 and developed in T-Max Developer for 10 rnin at 20 "C.
For immunoperoxidase staining, punch biopsies from normal skin of healthy volunteers or from the leading edge of healing wounds of otherwise healthy volunteers, were snap frozen in liquid nitrogen and sectioned (4-6 pm) in a cryostat; sections were air dried and fixed in a mixture of chloroform/acetone (1:l) at 4 "C for 10 min. Immunoperoxidase staining was performed following a procedure previously described (Pellegrini et al., 1992).
Growth Assay-The natural alkaloid K252a was purchased from Calbiochem. EGF was from Austral Biologicals. a D l l mAb was used as nonconcentrated conditioned medium of subcloned hybridoma cells. ME20.4 mAbs were purified immunoglobulins.
Assays were performed in defined medium. Keratinocytes from subconfluent primary cultures were plated in basal medium (keratinocyte growth medium depleted of fetal calf serum, EGF, cholera toxin, and containing 10 ng/ml insulin) onto 96-well plates (1 X 10' cells/ well). Indicated factors (see "Results") were added 12 h after plating. After an additional 12 h, 5 pCi/ml ['Hlthymidine (Amersham, 90 Ci/ mmol) were added. Cells were trypsinized 12 h later, collected with the aid of a cell harvester, and incorporated radioactivity was p counted. The NGF-dependent autocrine keratinocyte proliferation was also evaluated by plating keratinocytes (4 X 10' cells/well) onto 6 multiwell plates (precoated for 24 h with hybridoma supernatants) in basal medium, in the presence of indicated factors (see "Results").
After 4 days, cells were stained with rhodamine B (De Luca et al., 1988a, 1988b and photographed, or trypsinized and counted.

RESULTS
NGF Binding Assays-Saturation binding assays were performed in normal human keratinocytes as described under "Experimental Procedures." As shown in Fig. 1, two ranges of lZ5I-NGF concentration (4-100 PM and 1-15 nM) were used to explore high (upper panels) and low (lower panels) affinity binding sites. Scatchard analysis (rightpanels) pointed to two classes of binding sites with a dissociation constant (Kd) of 1.086 (k0.108) X IO-" (upper panel) and 8.698 (k0.693) x IO-' (lower panel), respectively. Calculation of the B,,, indicated the presence of 934 -+ 38 high affinity and 223,200 +-12,173 low affinity receptors per cell. Nonequilibrium binding conditions are particularly suitable for demonstrating high affinity binding sites (Eveleth and Bradshaw, 1992). As shown in Fig. 2, binding assays performed in nonequilibrium conditions (reaction stopped 5 min after 1251-NGF addition), emphasized the presence of high affinity receptors on normal human keratinocytes. Values obtained from the Scatchard analysis of nonequilibrium binding assays cannot be considered accurate estimation of the binding constants of the NGF sites (Green et al., 1986;Eveleth and Bradshaw, 1992). Comparable data, both in terms of dissociation constants and binding sites, were obtained from three different keratinocyte strains (not shown).
NGF Receptors in Normal Human Keratinocytes-The expression of p7fiNGF' and p140irkA was investigated in skin sections and in cultured normal human keratinocytes. Equal amounts of poly(A+) RNA were isolated from keratinocytes either in the exponential phase of growth or at different time after confluence (reconstitution of multilayered sheets of COhesive epithelium), and hybridized with the human specific ~7 5~'~' probe. As shown in Fig. 3, a 3.8-kilobase transcript, consistent in size with ~7 5~~~' mRNA detected in the human melanoma cell line Hs-294T (lane 1 ), was present in growing keratinocyte colonies ( l a n e 2 ) and in reconstituted epithelium Nonequilibrium binding studies specifically detect high affinity binding sites. Each point was averaged from triplicates. Variation among triplicates was <3%. Comparable data were obtained from three different keratinocyte strains. 1 and 6 days after confluence (lunes 3 and 4, respectively). The level of expression of ~7 5~~~' mRNA strongly decreased when a confluent differentiated and fully stratified epithelium was obtained (lunes 2-4). The expression of p7fiNGFr was also determined by immunocytochemistry using the mAb ME20.4 (Ross et al., 1984). Fig. 4 (panel 6) shows expression of p7!jNGF* in a colony of growing keratinocytes plated on lethally irradiated 3T3-52 cells. Positivity was limited to keratinocytes (compare the F-actin staining in panel a) and was evenly diffused over their surface, with no obvious receptor cluster- ing. In skin sections of normal donors (Fig. 5,panel a), p7!jNGn was evenly distributed over the surface of basal keratinocytes FIG. 4. Immunofluorescence. Immunofluorescence detection of and absent from suprabasal terminally differentiated cells ~7 5~'~ in a small colony of human keratinocytes growing on a feeder-(see also Bothwell (1991a), Wheeler and Bothwell (1992), and layer of lethally irradiated 3T3-52 fibroblasts. The colony was stained Fantini and Johansson (1992)). Interestingly, the expression for F-actin with fluorescein-tagged phalloidin (panel a ) and simulta-of biolo&c~ly active NGF is limited to basal keratinocytes neously for ~7 5~' " using the mAb ME20 .4 (panel b). The positivity for p75NGR was limited to keratinocytes and evenly diffused Over their and decreases after full epithelial maturation as well (Di surface. No evidence for the formation of receptor aggregates was Marc0 et ale, 19911, Suggesting a. role for NGF during the obtained. Bar denotes 5 pm. kb, kilobase. exponential phase of keratinocyte growth. In sections of small

Autocrine Stimulation of Human Keratinocyte Growth by NGF
punch biopsies taken at the regenerating boundary of burnt human skin (Fig. 5, panel b), ~7 5~~~' was also detected at the surface of rather swollen basal cells; its positivity, however, was less homogeneous than in normal epidermis, with strongly positive cells alternating with cells showing weaker surface staining. The high affinity binding properties of NGF have been associated with the expression of the protooncogene trkA Kaplan et aL, 1991). To determine the level of trkA expression in keratinocytes, 10 pg of poly(A+) RNA were isolated from keratinocytes in primary or secondary culture in their exponential phase of growth, and hybridized with a human p140trkA cDNA probe. The 3.2-kilobase trkA transcript, present in human K562 leukemia cells (Fig.  6, lane 1, arrows), was undetectable in keratinocytes (Fig. 6,  lanes 3-6). Instead, the probe gave multiple but weak bands ranging from 1.8 (lower arrowhead) to 3.9 kilobases (upper arrowhead). The expression of p140trkA was also investigated by PCR (Fig. 7). K562 (lams 2 and 3) and keratinocyte (lanes 4 and 5 ) mRNA were reverse transcribed to cDNA using either random hexamers (lanes 3 and 5 ) or 4a oligonucleotide primer (lanes 2 and 4 ) and subsequently used as template in PCR with oligonucleotide 3a and 4a as primers. The predicted band of 258 base pairs was present in K562 mRNA (lanes 2 and 3, arrow), but was absent in normal human keratinocytes (lanes 4 and 5 ) . One ng of PDM-17 plasmid DNA gave the expected 258-base pair fragment after PCR amplification using the 3a and 4a oligonucleotide primers (not shown). Futhermore, antisera against the 14 carboxyl-terminal residues of p140"" did not immunoprecipitate the receptor from metabolically labeled keratinocytes, and did not stain normal human skin section or cultured keratinocytes (not shown), further demonstrating the absence of trkA in human epidermal cells.
Given the apparent absence of trkA in keratinocytes, the presence of specific high affinity NGF binding sites, and the sequence homology between the different members of the trk family, we therefore decided to further analyze the multiple bands recognized (in low stringency conditions) by the specific trlvi probe. By screening normal human keratinocyte cDNA libraries with the p140"" probe, we have identified, by sequence comparison, a new member of the trk protooncogene  strains (lanes 3-6) were hybridized (in low stringency conditions) with the human ~1 4 0 " '~ cDNA probe. Arrow indicates the 3.2-kilobase trkA transcript. Arrowheads indicates the 1.8-and 3.9-kilobase transcripts detected in keratinocytes. Equal amounts of poly(A') RNA were loaded lanes 2-6 as assessed by ethidium bromide staining and &actin hybridization (not shown). The film has been exposed for 15 days.  mRNA (lanes 2 and 3 ) and normal human keratinocytes mRNA (lanes 4 and 5 ) . The  of Huma 9 and IO), demonstrate the presence of a novel autocrine loop in normal human keratinocytes. The strong inhibition of aDl1 in the presence of other growth stimuli suggests that this autocrine loop is crucial for keratinocyte growth. However, this complicates the demonstration of a direct effect of exogenous NGF on keratinocyte growth. Cultured keratinocytes form colonies, each colony being the progeny of a single basal keratinocyte (Rheinwald and Green, 1975). If exogenous NGF is added to cells (in the absence of feeder-layer, and in defined medium) only a few hours after plating, when the concentration of endogenous NGF is presumably still very low (Fig. 11, close bars), a significative increase in [3H]thymidine incorporation is observed compared to the control (Fig. 11, open bars). When keratinocyte colonies become larger, the exogenous NGF effect is no longer visible (not shown).
Roles of ~7 5~'~' and trk-Several reports indicate that the stimulation of the p140trkA in cell lines responsive to NGF is necessary and/or sufficient to elicit a biological response (Cordon-Cardo et al., 1991;Loeb et aL, 1991;Ibanez et al., 1992). However, other reports highlight the crucial role of the association of p140trkA and ~7 5~~~' in regulating biological activities of NGF (Hempstead et al., 1990(Hempstead et al., , 1991Lee et al., 1992; for recent reviews, see Meakin and Shooter (1992) and Barker and Murphy (1992)). In intact cells, the natural alkaloid K252a (at nanomolar concentrations) inhibits the action of NGF and other neurotrophins without affecting other growth factors acting through stimulation of their tyrosine kinase receptors (Berg et al., 1992;Tapley et al., 1992;Knusel and Hefti, 1992). K252a acts by selectively inhibiting the tyrosine kinase domain of the trk family of protooncogenes (Berg et al., 1992;Tapley et al., 1992). ME20.4 is a mAb which recognizes human (but not rat) ~7 5~~~' and inhibits NGF binding to the low affinity receptor (Ross et al., 1984), also in normal human keratinocytes (not shown).
We investigated the effect of K252a and ME20.4 on the NGF-mediated autocrine growth stimulation of normal human keratinocytes. Assays were performed in the same conditions as in Fig. 9 (panel B ) . As shown in Fig. 12 (closed   circles) there was a dose-dependent inhibition of [3H]thymidine incorporation by nanomolar concentrations of K252a. The alkaloid gave a 50% inhibition at a concentration as low as 5 nM. Matching the effect of the anti-NGF mAb (see squares) did not inhibit [3H]thymidine incorporation even at a concentration of 500 ng/ml. These data indicate that the trk tyrosine kinase stimulation in these cells is necessary and sufficient to induce the NGF-dependent autocrine loop. The EGF-dependent enlargement of colonies and thinning of their interior (Barrandon and Green, 1987b) and the insulin-like growth factor-1 dependent stimulation of keratinocyte growth (Barreca et al., 1992), were not affected by the simultaneous addition of EGF or insulin-like growth factor-1 and K252a (50 nM) (not shown).

DISCUSSION
Epidermis, the outermost layer of skin, consists of stratified squamous epithelium accounting for only one-fortieth of skin thickness. Yet, its function is essential in maintaining the stability of the interior milieu of the entire organism and in protecting the body against environmental hazards. Epidermis survives physical and chemical traumas by a continuous self-renewal. To do this, small progenitor cells Green, 1985, 1987a), forming the innermost epidermal basal layer, undergo mitosis to replace terminally differentiated cornified cells (see Watt (1989), for review). Within a growing keratinocyte colony, migrating and proliferating cells are located at the narrow peripheral rim of the colony, whereas internal cells are destined for terminal differentiation (Barrandon and Green, 1987b). In particular situations, such as healing of large wounds, basal keratinocytes have to strongly increase their migration and growth rates.
The regulation of normal human keratinocyte growth is a complex phenomenon involving both nutritive support from dermal blood vessel and a network of paracrine and autocrine loops which act mainly within the local skin environment. Fibroblasts are a source of several polypeptides binding to receptors endowed with tyrosine kinase activity and stimulating keratinocyte functions. For example, hepatocyte growth factor (scatter factor) is a potent modulator of epithelial cell motility (Stoker et al., 1987;Matsumoto et al., 1991), while fibroblast-derived keratinocyte growth factor and insulin-like growth factor-1 stimulate keratinocyte proliferation (Finch et al., 1989;Barreca et al., 1992). Indeed, optimal keratinocyte clonal growth in vitro is obtained when cells are plated on a feeder-layer of lethally irradiated fibroblasts (Rheinwald and Green, 1975) in the presence of factors which cooperate in sustaining their growth (see Green (1980)). However, when cultured human keratinocytes reach a critical density, they become capable of autonomous growth (Cook et al., 1991a(Cook et al., , 1991b.6 These autocrine-paracrine loops have been related to the synthesis and secretion of transforming growth factor a (which modulates keratinocyte migration and consequent growth) (Barrandon and Green, 1987a;Coffey et al., 1987), basic fibroblast growth factor (Halaban et al., 1988;Shipley et al., 1989), andamphiregulin (Cook et al., 1991b), which, however, acts through the EGF/transforming growth factora receptor (Johnson et al., 1993).
In this report we demonstrate that keratinocyte-derived NGF (Di Marco et al., 1991), in addition to its fundamental role in regulating migration and dendritic arborization of surrounding melanocytes (Yaar et al., 1991;De Luca et al., 1993), plays a key role in the autocrine keratinocyte growth stimulation. This has been proven by (i) presence of high affinity receptors on keratinocyte plasma-membrane, (ii) direct effect of NGF on [3H]thymidine incorporation, (iii) p~tent inhibition of autocrine keratinocyte growth by a D l l mAbs (able to inhibit human NGF biological activity), and by a trk-specific inhibitor, the natural alkaloid K252a. It is worth noting that both a D l l and K252a gave an 80% inhibition of keratinocyte growth in defined medium, and in the presence of other growth stimuli such as EGF, cholera toxin, and a high concentration of insulin, suggesting that the NGF-dependent autocrine loop is crucial for the optimal keratinocyte growth, and that the absence of NGF make other growth stimuli unfit for eliciting their activity. Keratinocyte-derived NGF is also responsible for the widespread sprouting of sensory nerve fibers observed during the healing of wounds.' This further contributes to keratinocyte proliferation by the paracrine secretion of nerve fiber-derived vasoactive-intestinal polypeptide, which activates keratinocyte adenylate cyclase and stimulates its growth (Haegerstrand et al., 1989). Thus, NGF cooperates in defining the epidermal cytokine network, which regulates, within the local epidermal environment, growth and differentiation of different epidermal cell types as well as epidermal immunological functions (see Luger and Schwartz (1990), for review).
The most important NGF biological effect has always been related to survival and maintenance of differentiated functions of vertebrate neurons (see Levi-Montalcini and Angeletti (1968), Levi-Montalcini (1987), and Thoenen (1991), for reviews). Indeed, NGF promotes the survival of ganglionic cells (neurotrophism) by preventing their physiological death, and it induces neurite elongation from peripheral ganglia and their proper orientation (neurotrophism). These data clearly demonstrate a growth promoting activity of NGF outside the nervous system (Cattaneo and McKay, 1990). NGF induces normal human epithelial cells, in primary culture, to enter the S-phase of the cell cycle and sustains their proliferation.
NGF is the prototype of a family of related molecules, called neurotrophins (see Thoenen (1991)). NGF, brain-derived neurotrophic factor, and neurotrophins (NT-3, -4, and 5) are structurally related molecules which bind to a common low affinity receptor, the ~7 5~~~' (Meakin and Shooter, 1992;Barker and Murphy, 1992). Based on some reports, high affinity binding and specificity appear to be conferred by a heterodimeric complex of ~7 5~~~' and the products of the trk family of protooncogenes (trk A , B , and C) Squint0 et al., 1991;Soppet et al., 1991;Lamballe et al., 1991;Bothwell, 1991b;Hempstead et al., 1991). Alternatively, trk monomers might bind NGF with low affinity (Kaplan et al., 1991), while trk dimerization  or ~7 . 5~~~'mediated NGF endocytosis Bradshaw, 1988, 1992;Kahle and Hertel, 1992) could confer high affinity binding properties. Normal human keratinocytes have approximately 1000 high affinity receptors and more than 200,000 low affinity binding sites per cell in conditions where NGF internalization is blocked. Whether the low affinity binding capacity is conferred only by ~7 5~~~' or also by trk monomers, remains to be determined. Moreover, contrasting data exist in the literature concerning the role of ~7 5~~~' and trk in mediating signal transduction and biological activity of NGF in target cells. The stimulation of the p140trkA in cell lines responsive to NGF is necessary and sufficient to elicit a biological response (Cordon-Cardo et al., 1991;Loeb et al., 1991;Ibanez et al., 1992). In a recent paper, even the physical association between p140trkA and ~7 5~~~' has been refuted . Other reports, instead, highlight the crucial role of the association of p140trkA and ~7 5~~~' in regulating biological activities of NGF (Hempstead et al. (1990, 19911, Lee et al. (1992, for recent reviews, see Meakin and Shooter (1992) and Barker and Murphy (1992)). Of particular interest are data obtained in transgenic mice, showing that targeted mutation of the gene encoding ~7 5~~~' leads to deficits in the peripheral sensory nervous system (Lee et al., 1992). Eveleth and Bradshaw (1992) linked signaling dysfunction to altered NGF receptor mediated endocytosis and degradation.
Our data, obtained with the trk specific inhibitor K252a and the mAb ME20.4 against the p75NGF', clearly demonstrate that, at least in normal human keratinocytes, (i) NGF-mediated autocrine growth stimulation requires the activation of the trk tyrosine kinase receptor, (ii) such a stimulation is necessary and sufficient to induce keratinocyte proliferation, and (iii) ME20.4 alone does not interfere with NGF biological activity. However, ~7 5~~~' is up-regulated in the exponential phase of keratinocyte growth (Fig. 3) as well as in injured Schwann cells (Taniuchi et al., 1986), it is abundantly expressed during the very early stages of embryonic development, in skin as well as in other organs (Ernfors et al., 1988), and it is associated with several members of the trk family (Meakin and Shooter, 1992). This suggests a regulatory role for ~7 5~~~' which awaits experimental confirmation.
The trkA protooncogene product has been considered as the high affinity NGF receptor . Interestingly, although NGF elicits biological activities in peripheral tissues, in uiuo trkA expression studies in mice has shown that the expression of trkA mRNA is restricted to sensory cranial and spinal ganglia of neural crest origin, becoming a specific marker of neural crest derived sensory neurons . Thus, the absence of trkA in normal human keratinocytes, even after PCR amplification, is not surprising. However, the presence of specific and high affinity NGF binding sites in human keratinocytes, and the sequence homology between the different members of the trk family, led us to the discovery of a new member of the trk family (trkE), which generates a 3.9-kilobase transcript, present in keratinocytes and absent in K562 cells. To our knowledge, no member of the trk family of protooncogenes has yet been described in normal human cells. These observation raise interesting questions on the different roles of trkA and trkE in humans and suggest that NGF activity in different human tissues might be mediated by different members of the trk family of protooncogenes.