In vitro analysis of interactions between sensory neurons and skin: evidence for selective innervation of dermis and epidermis

The importance of interactions with the local environment for the extension and guidance of nerve fibres during neurogenesis has been demonstrated in vivo as well as in vitro (see Goodman, Raper, Ho & Chang, 1982; Johnston & Wessells, 1980; Katz & Lasek, 1980, for reviews). In particular, many in vitro studies have employed neurons from the peripheral nervous system to try to understand the part played by target tissues in nerve growth control (Bonhoeffer & Huf, 1980; Ebendal, 1981; Ebendal & Jacobson, 1977; Pollack, Muhlach & Liebig, 1981 … ).

Few investigations, however, have been dedicated to the detailed analysis of interactions between nerve fibres and skin cells in vitro (Andres & Van der Loos, 1983; Ebendal, 1977; Lumsden & Davies, 1983; Verna & Saxod, 1979A). Although conclusions derived from the artificial environment of tissue culture necessarily remain questionable, in vitro studies may provide suggestive glimpses of events occurring during the patterning of cutaneous nerves and the development of sensory receptors. In birds, for instance, the nature of the morphogenetic interactions between sensory nerve fibres and mesenchymal cells which lead to the formation of sensory end organs (such as Herbst corpuscule; see Saxod, 1978 for a review) and the origin and significance of the scarce intraepidermal innervation are still unknown.

In order to extend previous studies (Saxod & Mauger, 1976 ; Verna & Saxod, 1979b) on the development and regulation of the pattern of bird cutaneous nerves we have analysed the interactions between dorsal root sensory neurons and cutaneous cells from chick embryos in serum-free cocultures. We used serum-free supplemented medium in order to reduce the heavy outgrowth of non-neuronal cells of the dorsal root ganglion which hinders quantitative estimations of nerve fibre growth, and to prevent modifications in the interactions between the cultivated cells due to binding of serum factors on the cell surface. We find that neurons interact differently with dermal than with epidermal cells. While nerve fibres readily extend over dermal cells, forming close membrane associations with some of them, they demonstrate a strong avoidance reaction with epidermal cells by changing their direction of extension. A brief preliminary note on these results was recently published (V erna & Saxod, 1983).

L15 culture medium and soybean trypsin inhibitor were purchased from Boehringer Mannheim France, Nerve Growth Factor from Laref (Switzerland); transferrin and insulin from Collaborative Research, Inc. (U.S.A.); progesterone, putrescine and seleneous acid from Fluka AG (Switzerland); and poly-L-lysine HBr (M_r 150000–300000) from Sigma.

Lumbosacral dorsal root sensory ganglia (DRG) and dorsal skin were dissected from 7-day chick embryos.

Skin was suspended for 15 mins at 4 °C in Ca²⁺–Mg²⁺-free PBS containing 0·5 % trypsin, then separated into dermis and epidermis. These tissues were rinsed in Ca²⁺–Mg²⁺-free PBS and allowed to stand for 5 mins in L15 medium containing 0·33 % trypsin inhibitor.

Using fine forceps, each tissue was dissociated into small pieces (1mm diameter). Seven to eight tissue fragments of either dermis or epidermis were cultured with the same number of DRG fragments in such a way that each explant of cutaneous tissue was adjacent to two DRG at a distance of about 1 mm. In order to avoid disturbance in the positioning of the explants, a piece of dialysis cellophane was laid over them for the first 24 h (Verna & Saxod, 1979a).

The serum-free supplemented culture medium was prepared according to Bottenstein, Skaper, Varón & Sato (1980) and consisted of: L15 medium (× 1); glucose (6mg/ml); penicillin (125i.u./ml); N.G.F. (10 μg/ml); transferrin (5 μg/ml); insulin (5 μg/ml); progesterone (6·3 ng/ml); putrescine (8·8 μg/ml) and seleneous acid (4 ng/ml).

35 mm tissue culture dishes (Falcon) were either coated with a thin layer of soluble bovine skin collagen (gift from the C. E. R. A. D., Lyon, France) or exposed to a poly-L-lysine solution(10 μg/ml in distilled water). Collagen coating was achieved by spreading one drop of a 0·5 mg/ml collagen solution in distilled water on to each dish and allowing it to dry for 2–3 h. Culture dishes were exposed for 3 h to the poly-L-lysine solution and rinsed with L15 medium just before use.

The cultures were incubated at 37 °C in a humidified atmosphere and maintained for 1 to 10 days without change of medium.

Time-lapse cinemicrography was used to analyse neurite behaviour and growth. For this purpose, selected areas of the cultures were observed with a Leitz Diavert inverted microscope and filmed at one frame every 5–15 minutes with a 16 mm Bolex camera using Codex pan rapid film. Recordings were made from the 1st until the 4th day of culture. Cinematographic records were projected, frame by frame, on an x–y coordinate digitizing tablet (Hipad) connected to an Apple III microcomputer. The coordinates of successive positions of the growth cone were recorded with the aid of the tablet electronic pen, and stored in the computer. These data were then processed (using a computer program written by Y. Usson) to calculate various parameters of the neuritic growth.

Cj, velocity between two consecutive positions (Xj, Xj + 1) of the growth cone; d, distance between these two positions; t, time interval between 2 successive measurements.

d, distance between the first (XI) and the last position (Xn) of the growth cone; T, total duration of the observation.

α_i is the angle between the directions of two successive segments, two consecutive positions of the growth cone defining a segment; N is the number of angles.

Student’s t-test and Fisher’s f-test were used to compare means and variances of data respectively. Growth parameters were calculated for a total of 122 nerve fibres in cocultures (47 with dermis and 75 with epidermis).

Cultures were quickly rinsed in a 0·1 M-phosphate buffer (pH 7·4), fixed for 30 mins at room temperature in a 0·1 M-phosphate buffer containing 6·25 % glutaraldehyde and postfixed for 1 h in 1 % osmium tetroxide in phosphate buffer. Some cultures were then stained 10 mins in 0·3 % tannic acid (according to Rees, 1978).

Areas of particular interest were selected, cut and stained with uranyl acetate and lead citrate.

This medium formulation, as reported for the non-neuronal cells of peripheral ganglia (Bottenstein et al. 1980) strongly reduces the proliferation and migration of dermal cells. Consequently, despite a good survival of these cells after 10 days of culture, confluency is never reached and the singly migrating cells remain close to the explant periphery (Fig. 1). The cells display a fibroblastic morphology, with a flat triangular or polygonal shape on both culture substrates. A well-developed endoplasmic reticulum, microtubules and bundles of submembraneous microfilaments are common features of their cytoplasm. Cell migration is characterized by short-range translocation periods separated by stationary phases during which the dermal cells nevertheless exhibit an intense protusive activity.

During the first days of culture, the lack of serum factors in the medium does not hinder the survival and development of the epidermis. By day 4, however, some cells begin to die and a rapid increase in rate of cell death leads to almost complete necrosis of the explant after 8 days of culture.

After plating, epidermal explants attach quickly and spread on the culture substrate. The cells are tightly associated and migrate in sheets (Fig. 2). The spreading and the motile activity of epidermis is far greater on collagen than on poly-L-lysine. Consequently, on collagen the epidermal sheet is large and rarely consists of more than two cell layers. On the other hand, epidermis grown on poly-L-lysine is thicker (three or more cell layers), in particular on the edge of the sheet, and often contracts as the cultures age.

The uppermost layer of the epidermal sheet is made of typical peridermal cells with numerous microvilli on the surface facing the culture medium (Fig. 3). Kératinocytes are joined one to another by well-differentiated desmosomes. From place to place, melanoblasts lie between kératinocytes.

The radially directed neurite outgrowth from spinal ganglion explants occurs within 12 hours of culture. Extending nerve fibres may be isolated, or grouped in bundles. Because of the serum-free culture medium, the early neuritic outgrowth takes place without the accompanying non-neuronal cells (Bottenstein et al., 1980; Yavin & Yavin, 1980) which, as cultures age, will migrate out from the explant. However the population of migrating non-neuronal cells remains low throughout the period of culture and their displacement is restricted to the periphery of the ganglion explant.

The initial orientation of neurite extension is not changed by encountering cells migrating from the periphery of the dermal explant (Fig. 4). Transient contacts are established between growth cones and dermal cells, and later on neurites end up in the mass of the dermal explant. Neurites can elongate both over and under dermal cells as well as over the culture substrate between them without significantly modifying their directions of extension.

The rate of neuritic growth was significantly greater on poly-L-lysine than on collagen (Vi, 61·4 μm/h versus 35·9 μm/h; Vm, 53·4 μm/h versus 27·8 μm/h) and the extension more directionally consistent (Am, 34·7° on poly lysine versus 51·3° on collagen) (Table 1).

We saw no significant modifications in these growth parameters before and after nerve fibres encountered dermal cells, but as our sample of neurites was small this observation deserves a more detailed study.

When the filopodia of a growth cone come into contact with a dermal cell, either they withdraw within a few minutes and the neurite keeps elongating and ignores the cell, or a closer apposition is realized between the membranes of the growth cone and the cell. In this case, membranes run parallel for a length of 5 to 15 μm and are kept apart by an intercellular space of 20 nm (Fig. 5). No particular junctional specialization is observed. Slight changes of neurite trajectory can occur when, after a contact, the neurite is dragged along by a dermal cell during its phase of short-range translocation (Verna & Saxod, 1979a). Thus neurite–dermal cell contact does not prevent the dermal cell migration and is strong enough to suggest a possible role in nerve fibre growth. Moreover, it appears that these contacts are not formed with every cell encountered and it will be of special interest to determine whether growth cones are distinguishing a special class of dermal cells.

When spinal ganglia and epidermal explants are cultured side by side on the substrate, no neurites extend from the side of the ganglion touching the epidermis (Fig. 6). Moreover, in epidermal cultures plated with a ganglion cell suspension, no neurons attach and extend neurites on the epidermal sheet. These events occur only on the culture substrate (Fig. 7).

In order to determine the fine growth behaviour of neurites during their encounters with epidermis, spinal ganglia explants were put 1 mm away from the epidermal tissue. The initial radially directed outgrowth of neurites was similar to that observed in cocultures with dermis. However, when growth cones reach the close vicinity of the epidermal sheet, nerve fibres modify their direction of growth and deviate around the epidermis (Fig. 8). Neurites never extended over the epidermis, and the great majority did not penetrate deeply into it. The expression of this avoidance behaviour is particularly obvious when tissues are grown on poly-L-lysine. Comparison of data gathered from quantitative time-lapse analysis reveals a significant decrease in the rate of neurite extension in the presence of epidermis. This is associated with a more erratic ordering of the direction of neurite extension than in the presence of dermis (Vi and Vm decrease from 61·4 and 53·4 μm/h respectively to 45·9 and 29·4 μm/h; Am rises from 34·7° with dermis to 58·1° with epidermis; Table 1 and 2A). In order to determine whether these modifications were actually due to the epidermis, growth parameters were calculated for neurites whose growth cones were located 100 μm or less from the epidermis and compared with those calculated for neurites with growth cones further away (Table 2B). In the vicinity of epidermis the growth of nerve fibres was slowed (Vi decreased from 56·4 to 41·6 μm/h and Vm from 48·6 to 21·9 μm/h), and the directionality of extension was markedly altered (Am increased from 38-9° to 70·4°). These changes in neuritic growth behaviour were seen at distances varying from 5 to 120 μm (average distance: 42 ± 4 μm) from the epidermal border. A majority of neurites (65 %) stop growing before reaching the epidermis, and either retract or exhibit successive short-range movements of extension and retraction. In most instances, growth is later reinitiated, but generally not towards the epidermis, so that neurites deviate around it. Nerve fibres reaching the epidermis (35 %) immediately suspend their further extension, and quickly withdraw to progress along the epidermal border. Later-arriving nerve fibres fasciculate with first (‘pioneering’) nerve fibres giving rise to a network of nerve fibre bundles around the sides of the epidermal explant (Fig. 8).

Similar quantitative analysis of cocultures grown on collagen substrates has not provided such clear evidence of significant differences in neuritic growth parameters between cocultures with dermis and epidermis. Nor were these values modified when neurites got closer to the epidermis. Such analyses were made difficult by the intensive migratory activity of the epidermal sheet on collagen which could, in part, account for the discrepancy with results obtained on poly-L-lysine. Indeed, moving epidermal cells generally come rapidly into contact with extending growth cones (the distance between the epidermal sheet and the neurite tip is limited by the field of view of the microscope) and this could happen before the neuritic growth rate could change. Nevertheless, nerve fibres coming into contact with the epidermis grow no further. Growth cones then withdraw and no close or long-lasting associations (such as those observed with dermal cells) are formed. Later on, after several unfruitful attempts to maintain their forward extension and to pass over the epidermal barrier, the growth cones are diverted along the edge of the epidermis (Fig. 9). If the movement of the epidermal sheet results in a growth cone being on its surface, filopodia are withdrawn, protrusive activity is abolished and the growth cone loses its fan-shape morphology and becomes bulbous.

In order to determine whether neurite ‘avoidance’ behaviour requires the presence of living epidermal cells, epidermis was precultivated 1 to-3 days, then killed with a lethal dose of X-rays (10 000 rads using a Secasi Tubic X-ray apparatus), and spinal ganglia were added to the culture. Epidermal cell death begins within a few hours of X-irradiation with a progressive increase in rate. Development and survival of neurons do not appear to be affected by this necrosis as judged by microscopic observation. However, neurite behaviour is significantly different, and no change in the direction of extension occurs either in the vicinity of epidermis, or after contact with the remaining necrotic epidermal cells (Fig. 10).

In a previous study (Verna & Saxod, 1979), we described contacts occurring between neurites and dermal cells cocultured in serum supplemented media and observed that neurites can be dragged away by the mesenchymal cells. Similar results were obtained in the present study with cells grown in serum-free medium.

Furthermore, the encounter with dermal cells led neither to a change in the direction of axon growth nor to a withdrawal of the nerve fibres. Neurites commonly use the dermal cell surface as a substrate on which to grow. The motility of growth cones from dorsal root ganglia neurons is therefore not inhibited in vitro by contact with dermal cells. Similar absence of contact inhibition of movement was reported by Wessells et al. (1980) during in vitro contacts between neurites and non-neuronal cells derived from the same ganglion, whereas Dunn (1971) and Ebendal (1976) observed contact inhibition in their cultures. It thus appears, as recently demonstrated by Nuttall & Zinsmeister (1983) that the neuronal response varies significantly depending on the type of non-neuronal cell encountered. Although growth cones appear to be able to adhere to any dermal cell surfaces, close association seems restricted to only a subpopulation of cells. Moreover, the resulting cellular membrane adhesion is strong enough to resist to the tension created by the displacement of the dermal cell. Taken together, these observations suggest a certain degree of mutual membrane affinity between growth cone and particular dermal cell. However, the absence of morphological criteria allowing us to distinguish these ‘target’ cells from their neighbours makes it difficult to comment on the specificity of these contacts. In birds, cutaneous sensory corpuscules are exclusively located in dermis. It was shown (Saxod, 1978) that their development, and especially that of Herbst corpuscules, requires interactions between the somatosensory nerve endings and the mesenchymal tissue. However, the exact nature of the morphogenetic events involved in the corpuscule histogenesis still remains unknown. That at least some of the contacts observed represent some of these morphogenetic interactions has to be considered.

Ebendal (1977), in an in vivo study of spinal cord ventral root axon growth in the chick embryo, gave evidence of morphologically similar contacts between growth cones and surrounding mesenchymal cells. Based on these and other observations, this author suggested that the axons might be directed by contacts with the surrounding cell surfaces. More recently, different in vivo studies (Al-Gaith & Lewis, 1982) have demonstrated, especially in the insect embryo, that some peculiar cells, located along the neuronal pathways, might serve as ‘guiding cues’ to growing neurites (Bate, 1976; Ho & Goodman, 1982; Goodman et al., 1982; Taghert, Bastiani, Ho & Goodman, 1982; Bentley & Keshishian, 1982). It thus would be of great interest to determine in our culture system whether such a ‘guiding’ role is devoted to some dermal cells and then to study these cellular interactions more precisely in vitro.

Many studies (Chamley, Goller & Burnstock, 1973; Chamley & Dowel, 1975; Ebendal & Jacobson, 1977; Ebendal, Jordell-Kylberg & Soderstrom, 1978; Erânkô & Lathinen, 1978; Pollack et al., 1981; Muhlach & Pollack, 1982) on the influence of target tissues on axonal growth, have demonstrated a preferential extension of neurites towards target cells, so that the ‘avoidance’ behaviour of nerve fibres with respect to epidermis was unexpected. Nevertheless, such behaviour has been described in cocultures associating tissues of different parts of the nervous system (Bray, Wood & Bunge, 1980; Ebendal, 1980; Crain & Peterson, 1982; Peterson & Crain, 1982; Smalheiser, Peterson & Crain, 1981, 1982). In these cocultures, the turning response of neurites generally occurs after the contact with the associated tissues, leading to the assumption that the paucity or the absence of specific recognition cues within the inappropriate tissue explants might be responsible for the change in the direction of axon extension. However, this hypothesis does not explain our results, especially those from cocultures on poly-L-lysine substrate. On this substrate, the deflection in neurite extension is, for the majority of nerve fibres, triggered at a distance from epidermis (some of the nerve fibres contacting the epidermal layer may extend from spinal ganglia neurons which normally provide the intraepidermal innervation). This behaviour still occurs in cocultures of spinal ganglia and whole skin in which nerve fibres elongate over migrating dermal cells located at the periphery of the epidermal sheet but yet deviate around it at a distance. The disappearance of this deviation reaction if epidermal cells were necrotic indicates that normal metabolic activity of these cells is an important prerequisite for the appearance of this phenomenon. A possible explanation of these results is the production by epidermal cells of one or several substances which, once released in the culture medium, affect neurite growth behaviour. Following the discovery of the nerve growth factor (Levi-Montalcini & Hamburger, 1953) and its action on nerve fibre extension, numerous studies have been dedicated to the characterization of other substances displaying such trophic effects (Lumsden & Davies, 1983; see Varon & Adler, 1980; Berg, 1984 for review). Although there is still no clear demonstration, there is a growing body of evidence for the existence of such chemotactic gradients within the developing embryo. However, these putative substances always display a positive neurotrophic influence leading to an oriented growth of neurites towards the source. It is thus important to investigate further the peculiar behaviour reported in this study, and to define the mechanisms involved in the ‘avoidance’ reaction.

On a collagen substrate, the absence of neurite deflection at a distance from the epidermis might be due to the strong motility of epidermal cells, which could bring them into contact with the growth cones well before a change in neurite behaviour took place. However, considering the hypothesis that a substance is released from epidermal cells, its concentration in the close vicinity of the edge of epidermis will be lower (also, the epidermal sheet edge is much thinner on collagen than on poly-L-lysine) and consequently have less effect on neurite behaviour than in cocultures on poly lysine. On the other hand, as observed for some neurite-promoting growth factors found in conditioned media (Collins, 1978; Collins & Garrett, 1980; Adler & Varon, 1980, 1981), such a substance may only act when bound to a peculiar substrate and there may be a preferential binding to polylysine rather than to collagen. In the embryo, where no polylysine is found, this substance may bind to some non-collagenous component of the extracellular matrix.

Nerve fibres contacting the epidermis formed no close membrane adhesions with epidermal cells, and most of the time neurites neither extended over it nor penetrated deeply within it, but progressed on the culture substrate along the edge of the epidermal cell sheet. Similar results in the mouse system, on a collagen substrate, have been recently obtained by E. Peterson (N. Smalheiser, personal communication). To explain this, one can assume that:

• i) in agreement with Letourneau’s observations (1975, 1979), nerve fibres preferentially remain and extend on the culture substrate because of its higher adhesiveness (growth cone-epidermal cell adhesiveness might be decreased by changes in the growth cone plasma membrane triggered by some epidermal factor; see Schubert et al. 1978 who demonstrated the influence of N.G.F. on P.C. 12 cell–substratum adhesiveness);

• ii) that the lack of some adhesion molecules (such as CAM, see Edelman, 1983, 1984 for reviews) or the absence of recognition cues on the epidermal cell surface impede the anchorage and thus the progression of growth cones.

The observations reported here give rise to many questions, in particular whether such an ‘avoidance’ mechanism acts during the development of sensory cutaneous innervation in vivo, or is due in vitro to a peculiar state of epidermal cells in response to serum-free culture conditions. In this latter case, epidermal cells may then release some chemical agents responsible for the ‘avoidance’ reaction observed. Nevertheless, the ultrastructural morphology of epidermal cells closely resembled that described in the epidermis of 7-day chick embryos by Mottet & Jensen, 1968. Moreover, the behaviour and the morphology of these cells do not profoundly differ, at least during the first four days of culture, from those occurring in serum medium (in which the nerve fibres still deviate around the epidermis; Verna, unpublished data) and in various other culture conditions (for a review, see Holbrook & Hennings, 1983).

In considering the possible role of this phenomenon in the normal development of bird cutaneous innervation, it should be noted that birds are peculiar among vertebrates with respect to cutaneous innervation in that: i) ‘Merkel corpuscules’ are found exclusively in the dermis and ii) intraepidermal nerve fibres giving rise to free nerve endings (most of them, if not all, being cold receptors) are not very numerous (see Saxod, 1978 for a review). In this latter case, what impedes the penetration of nerve fibres into the epidermis? This could be due to the presence of a mechanical barrier (such as the basement membrane); to the absence of putative target cells in the epidermis (such as Merkel cells, supposed to act as targets for arriving nerves, Scott, Cooper & Diamond, 1981); or, finally, to some ‘repulsion’ effect triggered by contact with epidermal cells or by some diffusible factors released by them. Recently, Feinberg, Repo & Saunders showed that the ectoderm of early chick embryos plays a role in the establishment and maintenance of the avascular zone (100 ± 20 μm thick) in the underlying mesodermis by controlling the position of the blood vessels. It is therefore conceivable that epidermis also controls the cutaneous nerve pattern formation. Thus, the avoidance of epidermis by nerve fibres in vitro may well reflect events which occur normally in the embryo, and, therefore give insights into the way the dermal and epidermal innervation is realized. In order to correlate these in vitro observations with normal development, an in vivo study of the location of nerve fibres in the dorsal skin of the chick embryo is currently in progress. Preliminary histological results show that, in 6- to 7-day embryos, main nerve fibre bundles are located beneath the epidermis at an average distance of 24 ± 4 μm (n = 20) and it is important now to determine the fine positions of nerve fibre terminals in the skin.

This work was supported by M.I.R. 83-V099 and 82E0680 grants.