β₄ integrin expression in myelinating Schwann cells is polarized, developmentally regulated and axonally dependent

Integrins are a family of cell surface receptors that mediate adhesion of cells to other cells or to extracellular matrix (ECM) (reviewed by Hynes, 1992). These widely expressed heterodimeric receptors are composed of one (of at least fourteen) α and one (of eight) β subunit, non-covalently associated and both membrane spanning. The adhesive interactions mediated by integrins may serve to position cells relative to one another. In addition, upon engagement with ligand, integrins may also transmit signals to the cell interior that affect cytoskeletal organization, protein phosphorylation, calcium oscillations or gene expression (reviewed by Juliano and Haskill, 1993). Moreover, the affinity of integrins for ligand may be modulated by the activity of other cellular receptors, such that integrin-dependent adhesion may be coordinated by other receptor pathways. In summary, the integrins are a diverse group of receptors that can link cell differentiation with cell morphology and position in the organism.

The α₆β₄ integrin has unusual structural characteristics and a restricted tissue distribution, both of which suggest a unique role in differentiation. The β₄ subunit has an unusually large cytoplasmic domain that is not homologous to other β subunits, includes four fibronectin type III-like repeats and may interact with cytoskeleton proteins (Hogervorst et al., 1990; Suzuki and Naitoh, 1990; Tamura et al., 1990). In addition, unlike some β integrins that can interact with multiple α subunits, β₄ has been shown to interact with only α_6. Although α₆ is expressed ubiquitously in association with β₁, the expression of β₄ is more restricted, so that α₆β₄ appears primarily in normal epithelial cell types, including basal keratinocytes, gut epithelium, mammary gland epithelium and Schwann cells (SCs) in peripheral nerve (Sonnenberg et al., 1990; Natali et al., 1992a).

SCs are the only glial cells in peripheral nerve and are derived from the portion of the neural epithelium that gives rise to the neural crest (Le Douarin, 1982). The morphology and molecular phenotype of SCs undergo remarkable changes depending on whether they either simply ensheath or go on to form a myelin sheath around an axon (Mirsky and Jessen, 1990). During development, SCs migrate along bundles of axons and proliferate, probably in response to an axonal mitogen (Webster and Favilla, 1984). As SCs cease migrating, they synthesize a basal lamina (Billings-Gagliardi et al., 1974), which is composed of laminin, merosin, type IV collagen, fibronectin, nidogen/entactin and heparan sulfate proteoglycan (Sanes and Cheney, 1982; Tohyama and Ide, 1984; Bannerman et al., 1986; Leivo and Engvall, 1989; Sanes et al., 1990). In developing nerves, some SCs ensheath axons in a 1:1 relationship and go on to form a myelin sheath. These myelinating SCs express high levels of myelin-specific genes (Lemke and Axel, 1985; Stahl et al., 1990). Other SCs ensheath bundles of unmyelinated axons; these non-myelinating SCs express distinct molecular markers such as nerve growth factor receptor, neural cell adhesion molecule (N-CAM), L1 and growth-associated protein-43 kD (GAP-43); but none of the myelin-specific proteins (Mirsky and Jessen, 1990; Curtis et al., 1992). Transplantation experiments of peripheral nerves that are predominantly composed of myelinating or non-myelinating SCs have shown that axons determine whether a SC has a myelinating or non-myelinating phenotype (Aguayo et al., 1976a; Weinberg and Spencer, 1976). Thus, there are a number of potentially distinct physical interactions between SCs and axons as SCs proliferate, migrate, ensheath axons and form myelin sheaths during the development of peripheral nerve.

In addition to interacting with axons, SCs interact with ECM components of peripheral nerve and these interactions also i n fluence SC behavior. The effects of ECM have been studied largely in vitro, in which laminin and fibronectin have been shown to regulate SC proliferation (Baron Van Evercooren et al., 1982; McGarvey et al., 1984; Muir and Manthorpe, 1992). Furthermore, there is compelling evidence that SCs must deposit a basal lamina in order to ensheath and myelinate axons fully (Eldridge et al., 1989) and, in turn, that the full assembly of the basal lamina by SCs requires SC-axonal contact (Clark and Bunge, 1989). To account for this observation, Clark and Bunge have hypothesized that axonal contact induces SCs to synthesize and/or polarize receptors for basal lamina components such as laminin. These data demonstrate that SCs must interact both with their basal lamina and with the axons that they ensheath in order to form a myelin sheath, but the receptors involved in these interactions have not been elucidated.

The special structural features of α₆β₄, as well as its limited tissue distribution, indicate that it may have a specialized role in SC. Since integrins mediate both cell-cell and cell-ECM interactions, it is plausible that α₆β₄ mediates ensheathment and/or myelination of axons. To explore these possibilities, we investigated the expression of β₄, α₆ and β₁ in developing, degenerating and regenerating rat sciatic nerve. Both β₄ and α₆ mRNAs are highly enriched in sciatic nerve. The β₄ and α₆ proteins are co-localized at the abaxonal border of myelinating SCs, which apposes their basal laminae. In contrast, α₆ and β₄ proteins are expressed in a different pattern in non-myelinating SCs. The expression of β₄, but not α₆ or β₁ mRNA, is developmentally regulated in sciatic nerve. Further, the levels of β₄ mRNA and protein, but not α₆ and β₁ mRNA, fall after axotomy, demonstrating that the expression of β₄ is modulated by axons, possibly at a transcriptional level. Finally, in cultured SCs, the level of β₄ mRNA, more than that of α₆, is increased by forskolin (Fsk), which is known to mimic some effects of axons on SCs. These data suggest that α₆β₄ may play a role in the myelination of axons by SCs.

Using aseptic technique, the sciatic nerves of anesthetized (50 mg/kg pentobarbital i.p.), adult (10-13 week old) Sprague-Dawley rats were exposed at the sciatic notch. Some nerves were doubly ligated and transected with iridectomy scissors, and the two nerve-stumps were sutured at least 1 cm apart; this technique prevents axonal regeneration to the distal nerve stump for at least 2 months. Nerve crush was produced by tightly compressing the right sciatic nerve at the sciatic notch with flattened forceps twice, each time for 10 seconds; this technique causes all of the axons to degenerate, but allows axonal regeneration. At 1, 4, 8, 12, 24 or 58 days after nerve injury, the animals were killed by CO₂ inhalation and the distal nerve stumps were removed, stripped of epineurium and immediately frozen in liquid nitrogen. For RNA extraction, the distal nerve stumps of crushed nerves were divided into two equal portions, each about 2 cm in length, which are subsequently called the proximal and the distal portion of the distal nerve-stumps. For protein extraction, the whole distal stump was homogenized. Unlesioned sciatic nerves were obtained from animals of varying ages, from postnatal day 5 (P5) through 90.

SCs were isolated from the sciatic nerves of 3 day old SpragueDawley rats (Brockes et al., 1979). Their numbers were expanded on poly-L-lysine-coated 100 mm tissue culture plates in Dulbecco’s Modified Eagle Medium (DMEM), supplemented with 10% heatinactivated fetal calf serum (FCS), 2 μM forskolin (Fsk) and glial growth factor (Porter et al., 1986). The cells were refed every 3-4 days and subcultured every 7 days. One week before harvesting, cells were fed with DMEM and 10% FCS without Fsk or glial growth factor, to achieve a non-Fsk-induced phenotype. After 4 days, cells were treated with either 0, 4 or 20 μM Fsk for 3 additional days.

RNA was extracted from cultured cells by the acid guanidinium thiocyanate-phenol-chloroform extraction method (Chomczynsky and Sacchi, 1987). RNA was isolated from rat sciatic nerves and from various tissues by CsCl₂ gradient centrifugation (Chirgwin et al., 1979). Equal amounts (10 μg) of total RNA were electophoresed in 1% agarose/2.2 M formaldehyde gels, transferred to nylon membranes (Duralon, Stratagene) in 6 or 20× SSC, and u.v. cross-linked (0.12 joules). Blots were prehybridized, hybridized, washed using standard techniques (Sambrook et al., 1989) and exposed to film at −80°C with an intensifying screen. The following cDNAs were utilized as probes:(1)a 0.8 kb EcoRI fragment from clone 210, a β₄ cDNA clone previously identified in a cDNA library constructed from rat sciatic nerve, using a differential screening strategy designed to recognize clones whose expression is enriched in peripheral nerve (Scott et al., 1991); it demonstrated 84 and 88% homology, respectively, when compared with the sequences of human (Hogervorst et al., 1990) and murine (Kennel et al., 1993) β₄ cDNAs and spans bp 3234 to 3948 of the human sequence (Hogervorst et al., 1990), which encodes a portion of the cytoplasmic tail that is unique among the β integrin subunits; (2) a 0.3 kb EcoRI-HindIII fragment of α₆ cDNA from the rat cell line 804G (Starr and Quaranta, 1992); (3) a 2.2 kb HindIIIScaI fragment of mouse β₁ cDNA (kindly provided by Dr M. Hollers; Shih et al., 1993), (4) a full-length cDNA of rat Po (Lemke and Axel, 1985), (5) a full-length cDNA of rat glyceraldehyde-3-phosphate deydrogenase (GAPDH; Fort et al., 1985). Plasmid inserts were generated from restriction endonuclease digestions, separated by agarose gel electrophoresis and purified by electroelution. ³²P-labeled cDNA probes with specific activities of 1×10⁹ cts/minute/μg were prepared by primer extension with random hexamers using the Prime-a-gene kit (Promega) according to the manufacturer’s instructions.

Spinal cords, sciatic nerves, cervical sympathetic trunks (CST) and the cervical portion of the vagus nerve were isolated and fixed in 4% paraformaldehyde in 0.1 M phosphate buffer, cryoprotected in 20% sucrose and embedded in OCT (Miles). Cryostat-cut sections (8 μm) were air-dried and rinsed in PBS (spinal cord and nerve roots), or additionally postfixed in cold acetone (sciatic nerve, CST and vagus nerve). N o n s p e c i fic binding was blocked by preincubating the sections with 2% FCS, or 20% serum (FCS or goat)/1% BSA, and 0.1% Triton X-100 in PBS, and the sections were incubated with primary antibodies overnight at 4°C. Primary antibodies included: (1) a rabbit polyclonal antiserum anti-α₆A cytoplasmic domain (6845), raised against the last 15 amino acids (IHAQPSDKERLTSDA) of α₆A (Cooper et al., 1991); (2) a rabbit polyclonal antiserum anti-α₆B (382), raised against a synthetic peptide from the carboxy terminus of human α₆B (KDEKYIDNLEKKQWITKWNRNESYS) (Tamura et al., 1991); (3) a rabbit polyclonal antiserum (6945), raised against the last 16 residues of the human β₄ cytoplasmic domain (TSGTLSTHMDQQFFQT) as described (Tamura et al., 1991);(4) a monoclonal antibody recognizing S100 (East Acres Biologicals);(5) a monoclonal antibody recognizing GAP-43 (BoehringerMannheim; Schreyer and Skene, 1991); (6) a rabbit polyclonal antiserum raised against laminin (Sigma) and (7) a rabbit polyclonal antiserum raised against purified β₁ integrin (kindly provided by Dr C. Buck; Solowska et al., 1991). The sections were washed extensively, then incubated with goat or donkey anti-rabbit (fluorescein-conjugated) or goat anti-mouse (rhodamine-conjugated) secondary antibodies (Jackson ImmunoResearch Laboratories), washed, mounted in 0.2 M DABCO (Johnson et al., 1982), and photographed with a Leitz microscope equipped for epifluorescence. Controls included replacing the primary antisera with normal rabbit serum, and blocking the β₄ staining with the s p e c i fic peptide used to raise the 6945 antibody: β₄ antisera was diluted 1:1000 in staining media and incubated at 4°C for 1 hour with 0, 10, 100, and 1000 μg of β₄ peptide.

Normal sciatic nerves from 20-day-old and adult rats, and transected sciatic nerves from adult rats, were dissected and frozen in liquid nitrogen. The membrane-associated and myelin fraction was purified according to the method of Micko and Schlapfer (1978). Protein degradation was prevented by adding protease inhibitors (Chymostatyn, Leupeptin, Apoprotin, Pepstatin 5 μg/ml, and PMSF 0.3 mM). The protein content of the myelin fractions was measured (Lowry et al., 1951) and equal amounts of protein were solubilized in sample buffer, separated using 6% SDS-polyacrylamide gel electrophoresis and transferred onto nitrocellulose sheets. In order to detect β₄ in transected nerve, a 5-fold increased amount of myelin extract was run in a parallel lane. Non-specific binding was blocked by incubating the nitrocellulose strips in 3% (w/v) non-fat dry milk, 0.05% Tween-20 in 10 mM Tris-HCl pH 8.0 and 0.9% NaCl, and by preabsorbing the anti-β₄ antibody with whole homogenate of rat brain (Hirabayashi et al., 1983, modified). Primary anti-β₄ antibodies were counterstained with affinity-purified sheep anti-rabbit alkaline phosphatase-conjugated antibodies (Boehringer Mannheim). The reaction products were visualized with an alkaline phosphatase color development kit (BioRad, Richmond, CA). Controls included replacing the primary antisera with normal rabbit serum, or adding 10 μg/ml of specific β₄ or α₆A peptide to the primary antisera incubation (see above for peptide sequences).

Total RNA from sciatic nerve and SCs was reverse transcribed and amplified by PCR as previously described (Tamura et al., 1991, m o d i fied). The primers, 5′-GTGAGGTGTGTGAACATCAG-3′(sense) and 5′-CATGGTATCGGGGAATGCT-3′ (antisense), flank the alternatively spliced insertion that distinguishes the α₆A and α₆B mRNA isoforms, and amplify products of 500 and 380 nucleotides, respectively.

To evaluate whether β₄ integrin is relatively enriched in peripheral nerve, we compared the steady state levels of β₄ mRNA in various adult rat tissues by northern blot analysis. Since β₄ forms heterodimers only with α_6, we also measured the expression of α₆ mRNA (Fig. 1). The β₄ cDNA probe identified a large transcript, at approximately 6 kb, the predicted size of human β₄ mRNA (Hogervorst et al., 1990; Suzuki and Naitoh, 1990; Tamura et al., 1990). β₄ mRNA expression was highly enriched in sciatic nerve as compared to other tissues. Similarly, the α₆ cDNA probe hybridized with a 6 kb transcript (de Curtis et al., 1991) that was enriched in sciatic nerve, although lower levels were detected also in the spleen, thymus and heart. Thus, the distribution of β₄ mRNA was more restricted than that of α_6. Although the steady state level of α₆ mRNA was somewhat lower than that of β₄, both β₄ and α₆ mRNAs were relatively abundant and enriched in adult rat sciatic nerve.

To determine the cellular localization of β₄ in peripheral nerve, we performed immunohistochemistry on sections of adult rat dorsal and ventral roots with a polyclonal antibody against β₄. β₄ immunoreactivity was found in the pia mater, which surrounds the spinal cord, and in the nerve root sheath, which surrounds and separates the ventral and dorsal rootlets (Fig. 2A). Within the roots, β₄ staining was found around all of the myelinating SCs (Fig. 2C). When viewed in cross-section, the nerve root consists of axons surrounded by their myelin sheaths. Surrounding the myelin sheath is the SC that produced it, all of which resides within a basal lamina tube. Therefore, the myelinating SC appear as a ring with an inner (adaxonal) surface facing the axon, and an outer (abaxonal) surface facing the basal lamina tube, which encircles the SC. This polarization of SCs is analogous to epithelia; the adaxonal surface of the SC corresponds to the apical surface of epithelia, the abaxonal surface of the SC corresponds to the basal surface of epithelia facing the basement membrane. β₄ staining appeared to be localized predominantly, if not exclusively, to the abaxonal surface, which interfaces the basal lamina. The myelin sheath and axon were unstained and appeared as an empty space inside the abaxonal surface (Fig. 2C). By double immunofluorescence, we compared β₄ staining with that of S100β? a calcium-binding protein found in the cytoplasm of SCs (Brockes et al., 1979; Stefansson et al., 1982). After S100β staining, the compact myelin sheath appeared as a ring-like void, separating the adaxonal from the abaxonal cytoplasm of the SC. In SCs, S100β staining was found throughout the cytoplasm, being present at the adaxonal surface and perinuclear cytoplasm, where β₄ staining was not seen (compare Fig. 2C,D). The polarized localization of β₄ to the abaxonal surface of myelinating SCs, which is analogous to the basal surface of epithelial cells (Bunge and Bunge, 1983), is thus consistent with its localization in other epithelia (Sonnenberg et al., 1990).

To determine which isoform of α₆ was expressed and its corresponding location in the nerve roots, we also stained adjacent sections of the nerve roots with either a polyclonal antibody that recognized the α₆A isoform or one that recognized the α₆B isoform (Cooper et al., 1991; Tamura et al., 1991). The α₆A antibody stained the pia mater, nerve root sheaths and the abaxonal surface of myelinating SCs in a pattern identical to that of the β₄ antibody (Fig. 2E). As compared to S100β staining, α₆A staining was not found at the adaxonal surface or perinuclear region of myelinating SCs (compare Fig. 2E,F). We did not detect any staining with the α₆B antibody, suggesting that α₆A is the predominant isoform expressed by SCs. These data show that β₄ and α₆A are colocalized in myelinating SCs, which is consistent with the finding that they form heterodimers in other tissues (Sonnenberg et al., 1990).

To confirm that α₆B is not present in peripheral nerve, as suggested by the immunostaining, we performed RT-PCR, modified from Tamura et al. (1991). We chose oligonucleotide primers flanking the 3′ end of the α₆ coding region that generate products corresponding to both isoforms. From sciatic nerve, we obtained only a 500 bp product, corresponding to the α₆A isoform, and no detectable 380 bp product, corresponding to the α₆B isoform. In contrast, in cultured SCs, we obtained predominantly a 500 bp product, but also a 380 bp product, demonstrating that this method can detect both isoforms (Fig. 3). These data show that α₆A is the predominant, if not exclusive, isoform in adult sciatic nerve, and confirm the recent report by Hogervorst et al. (1993). Finally, as Tamura et al. (1991) reported for embryonic stem cells, the expression of α₆ isoforms may be related to SC differentiation; the less differentiated SCs in culture express some α₆B, while fully differentiated SCs in adult sciatic nerve express only α₆A.

We determined the developmental pattern of β₄ and α₆ mRNA expression in sciatic nerve. Since β₁, the other integrin subunit known to associate with α_6, is expressed in all components of peripheral nerve including SCs (Lefcort et al., 1992, and data not shown), we also determined the regulation of its mRNA expression. Total RNA was isolated from rat sciatic nerves at various times after birth and analyzed by northern blotting. The level of β₄ mRNA increased progressively during postnatal peripheral nerve development (Fig. 4). It was present at low levels at postnatal day 5 (P5), increased remarkably around P15, when the peak of myelin structural gene expression occurs (Stahl et al., 1990), as shown by reprobing the same blot for Po. However, unlike Po levels, β₄ levels slowly increased further until at least P90. In contrast, the levels of α₆ and β₁ mRNAs were relatively constant between P5 and P90 (Fig. 4). The α₆ signal at P90 contained hybridization artifact. These data show that β₄? but not α₆ or β₁ mRNA expression increases in parallel with the mRNA expression of the myelin genes during the first two postnatal weeks. However, the level of β₄ mRNA continues to increase after the levels of myelin-related mRNAs peak, in parallel with the growth of the myelin sheaths (Wood and Engel, 1976).

Since the expression of most myelin-specific genes in SCs is transcriptionally dependent on the presence of axons (Lemke and Chao, 1988; Trapp et al., 1988), we studied the expression of β₄, α₆ and β₁ mRNAs in adult rat sciatic nerve after crush and transection injury. Crush of sciatic nerve causes Wallerian degeneration followed by prompt axonal regeneration, whereas permanent transection causes Wallerian degeneration but prevents axonal regeneration. Therefore, by comparing the resulting expression of β₄, α₆ and β₁ mRNAs in these two paradigms, the relative axonal dependence of their expression can be inferred.

To correlate the progression of axonal regeneration after crush injury and the levels of β₄, α₆ and β₁ mRNAs, we measured them in either the distal stump (day 4), or the proximal and distal segments of the distal stump at 8 days (e.g. 8p and 8d) and thereafter. After crush, β₄ mRNA expression in the distal stump fell remarkably by day 4, reappeared by 12 days and returned to the level of adult unlesioned sciatic nerve by 58 days (Fig. 5A). We compared the pattern of β₄ mRNA expression to that of a myelin structural protein by reprobing the same blot for Po, the most abundant mRNA in peripheral nerve (Lemke and Axel, 1985). The level of Po mRNA also fell rapidly by day 4, reappeared by 12 days and continued to rise returning to unlesioned nerve levels by 58 days (Fig. 5A, see also Gupta et al., 1988; Leblanc and Poduslo, 1990). Thus, the expression of β₄ mRNA, like that of Po mRNA, is maintained by axons in adult sciatic nerve, since it falls remarkably as axons degenerate. Furthermore, the return of β₄ and Po mRNA expression is similar and parallels the time course of axonal regeneration (Shawe, 1954), although high levels of Po mRNA reappeared earlier. In contrast, the steady state level of α₆ and β₁ mRNA did not change remarkably after crush injury (Fig. 5A). The slight changes observed in the α₆ and β₁ mRNA steady state levels appeared to be a reflection of differences in loading, as confirmed by reprobing the same blot for GAPDH mRNA, which is relatively unaffected by nerve injury. Therefore, after crush injury, the expression of β₄ mRNA, but not α₆ and β₁ mRNA, depends on the presence of axons.

To evaluate further the axonal dependence of β₄ mRNA expression, we analyzed sciatic nerves after permanent transection. The level of β₄ mRNA fell remarkably by 4 days, but then increased progressively from 8 to 58 days, even though axons did not regenerate (Fig. 5B). Reprobing this blot for Po (Fig. 5B) demonstrated that the level of Po mRNA fell dramatically by 4 days, and remained at this low level for at least 58 days, indirectly confirming that axons are not present, since Po expression is up-regulated in SCs in the presence of axons (Mitchell et al., 1990). Although the expression of β₄ mRNA reappeared in transected nerves (in the absence of axons), its level at 58 days was less than that in crushed nerves, in which axons are present (compare Fig. 5B to Fig. 5A). In contrast to its effect on β₄ and Po mRNA expression, transection did not alter the steady state level of α₆ or β₁ mRNAs (Fig. 5B). The minor fluctuations in α₆ and β₁ signals were accounted for by differences in loading, as demonstrated by reprobing the blot for GAPDH mRNA. Both the crush and transection data, therefore, suggest that the expression of α₆ and β₁ mRNA is not influenced by the presence of axons in sciatic nerve. Therefore, the expression of β₄ mRNA alone is axonally dependent, since it falls when axons are removed and its return is augmented when axons regenerate.

The level of β₄ mRNA at 8 days post-transection (no axons present) was unexpectedly greater than at 8 days after crush (regenerating axons returning) (compare Fig. 5A to Fig. 5B). Furthermore, at 12 and 24 days after crush, the level of β₄ mRNA is higher in the distal segment than in the proximal segment of the distal nerve stump; this is the reverse of Po mRNA expression, and, hence, of the gradient of axon regeneration (Fig. 5A). As above, these data again suggest that denervated SC, may re-express β₄ mRNA in the absence of axons, so that axons may suppress this re-expression of β₄ mRNA when they first contact these denervated SCs. To test this possibility, we transected the left sciatic nerve 24 days after crush, when β₄ mRNA levels were intermediate. These animals were killed after 2 days, and RNA was prepared from the distal stumps of these re-lesioned nerves and from the contralateral ones, which had been crushed 26 days prior to killing. The above hypothesis was not supported, since the level of β₄ mRNA was decreased in the re-lesioned nerves (data not shown). Although this finding does not eliminate the possibility that regenerating axons suppressed the expression of β₄ mRNA in a portion of the SCs in these nerves, these data demonstrate that, even at 24 days after crush injury in regenerating nerve, the expression of β₄ mRNA in a significant proportion of the SCs is axonally dependent.

To confirm that the modulation of β₄ mRNA expression resulted in the modulation of the β₄ protein, we performed western blot analysis on homogenates prepared from developing and transected sciatic nerves. In normal adult sciatic nerve, the β₄ antibody detected a protein of approximately 200×10³M_r (Fig. 6A, lane 1), which is the expected size (Sonnenberg et al., 1990). The 200×10³M_r band was β₄, since it was abolished by substituting normal rabbit serum for the primary antibody (lane 2), and by competing the β₄ antibody with β₄ peptide (lane 3), but not the α₆ peptide (data not shown). Fig. 6B demonstrates that β₄ expression is lower in homogenates of P20 than in those of adult sciatic nerves, confirming the developmental modulation of the expression of β₄ mRNA (Fig. 4). To determine whether axons modulate the expression of β₄ protein, we also measured β₄ levels in the distal nerve stumps of adult nerves at 4 and 8 days post-transection. When the same amounts of protein were analyzed, the intensity of the 200×10³M_r band dramatically decreased in transected nerves (Fig. 6B), again replicating the corresponding decreased expression of β₄ mRNA. By increasing the amount of homogenate from the 8 day transected nerve five-fold, the intensity of the 200×10³M_r band was approximately the same as that of normal nerve (data not shown), demonstrating that some β₄ was still present. These data confirm that the expression of both β₄ mRNA and protein is developmentally regulated, and axonally dependent.

Jaakola et al. (1993) have found that both SCs and perineurial cells in human sciatic nerve express β₄. To demonstrate that the decreased levels of β₄ in transected sciatic nerve r e flected decreased expression of β₄ by SCs, we performed immunohistochemistry on sections of normal and transected adult sciatic nerve as before. The β₄ antibody labelled the perineurium and the abaxonal surface of myelinating SCs (Fig. 7A,C). The Schmidt-Lantermann incisures did not appear to be labelled (Fig. 7A). The β₄ immunoreactivity of the SCs was clearly reduced by 8 days after transection (Fig. 7B,D), but returned sign i ficantly by 58 days after transection (Fig. 7E). In contrast, β₄ immunoreactivity in perineurial cells was unchanged at either 8 or 58 days after transection (Fig. 7B,D,E). These data confirm the modulation of β₄ expression shown previously by northern and western blot analysis of sciatic nerve. Furthermore, these data demonstrate that SC, but not perineurial, expression of β₄ is axonally modulated.

To understand better the regulation of β₄ and α₆ mRNA expression, we examined SCs cultured in the presence of Fsk, which is a model system of how SCs respond to axonal signals. These SCs express low levels of myelin-specific mRNAs and proteins, but can be induced to express much higher levels by Fsk and other drugs that increase cAMP (Lemke and Chao, 1988; Shuman et al., 1986; Sobue et al., 1986). SCs were cultured in the absence of Fsk for 4 days, then cultured with either 0, 4 or 20 μM Fsk for 3 days, after which total RNA was extracted and analyzed by northern blotting. The level of β₄ mRNA was clearly higher in SCs treated with either 4 μM or 20 μM Fsk (Fig. 8), while the level of α₆ mRNA was only slightly higher in SCs treated with 4 μM Fsk. The same blot was reprobed for Po mRNA, the level of which was also increased with Fsk (Lemke and Chao, 1988). Although Fsk increased the level of both β₄ and Po mRNAs, their levels were remarkably lower in SCs cultured in Fsk than in adult rat sciatic nerve (Fig. 8, see also Lemke and Chao, 1988), confirming that Fsk only partially reproduces the effect of axons on SCs. Therefore, the expression of β₄ mRNA? more than that of α_6, is induced by Fsk in SCs.

The previous results suggested that α₆β₄ integrin may play a specialized role in myelinating SCs. To study the expression and localization of α₆ and β₄ integrin in non-myelinating SCs, we immunostained sections of the cervical sympathetic trunk (CST) and vagus nerve of adult rats. The CST is chiefly composed of non-myelinated axons and their associated SCs, as well as a few, small myelinated axons (Aguayo et al., 1976b); the cervical vagus contains many myelinated axons and many more bundles of non-myelinating SCs than the sciatic nerve (Friedman et al., 1992; King and Thomas, 1971). As shown in Fig. 9A, β₄ immunoreactivity was located at the outer abaxonal surface of the large, myelinating fibers in the vagus nerve, as described above for sciatic nerve and lumbosacral roots. The β₄ immunoreactivity of non-myelinating SCs in the vagus nerve and CST, in contrast, was less intense and was apparently distributed homogeneously. The limited resolution of light microscopy did not allow us to distinguish β₄-immunoreactivity associated with in-folded SC plasma membranes from diffuse cytoplasmic β₄ staining. However, β₄ was not distinctly localized adjacent to the basal laminae (Fig. 9A,B). We performed double immunofluorescence, using a polyclonal antibody recognizing β₄ and a monoclonal antibody recognizing GAP-43, which stains specifically non-myelinating SCs (Curtis et al., 1992). As shown in Fig. 9A and C, β₄ and GAP43-immunoreactivity were co-localized, confirming that nonmyelinating SCs express β₄. Finally, to demonstrate better that the β₄ immunoreactivity was authentic, we preincubated the primary antibody with 10, 100 or 1000 μg of the peptide against which the antibody was raised. Preincubation with as little as 10 μg of peptide greatly reduced the level of β₄ immunoreactivity associated with the myelinating and non-myelinating SCs, as well as with the perineurium (Fig. 9D). The pattern of α₆ immunoreactivity in the non-myelinating SCs was similar to that of β₄ (data not shown).

These results suggest that non-myelinating SCs express low levels of β₄ and α₆, but that these integrins are not highly polarized to the surface of the non-myelinating SCs that apposes its basal lamina. However, this interpretation is limited by the small size of the non-myelinating SCs and the limited resolution of the light microscopy. To address this problem, we stained adjacent sections or vagus nerve and CST with an antibody recognizing laminin, which is an ubiquitous component of basement membranes, including the basal laminae of SCs (Sanes et al., 1982b). As shown in Fig. 9E and F, using this antibody, we could resolve the basal laminae of non-myelinating SCs. Thus, if β₄ and α₆ integrins were highly localized, apposing the basal laminae of non-myelinating SCs, as in myelinating SCs, our technique should have allowed us to detect this localization.

We found that both β₄ and α₆ mRNAs were highly enriched in peripheral nerve. In adult nerves, β₄ and α₆ were co-localized at the abaxonal surface of Schwann cells, which apposes the basal lamina. In spite of these similarities, the expression of β₄ and α₆ appeared to be differentially regulated. The steady-state levels of β₄ mRNA and protein increased remarkably during postnatal development after postnatal day 10, whereas the level of α₆ mRNA remained relatively constant. Furthermore, axotomy caused a sharp fall in the level of β₄ mRNA and protein, whereas the level of α₆ mRNA did not change. Immunohistochemical analysis demonstrated that the fall in the level of β₄ protein was restricted to the endoneurial SCs. Although the level of β₄ mRNA ultimately increased even inthe absence of regenerating axons, these levels were augmented by axons in regenerating nerves. Finally, the level of β₄ mRNA was induced in Fsk-stimulated SCs in culture, an in vitro model in which the expression of other axonally dependent genes is induced by forskolin (Lemke and Chao, 1988). Schwann cells, therefore, express high levels of β₄ mRNA and protein, and this expression requires continuous axon-SC interactions at times when such interactions are required to induce or maintain a mature myelinating SC phenotype.

We found β₄ immunoreactivity in all SC and perineurial cells of adult peripheral nerve, as has been briefly reported (Sonnenberg et al., 1990; Natali et al. 1992a). While this work was in progess, Jaakkola et al. (1993) reported their more extensive analysis of β₄ expression in developing human peripheral nerve. In contrast to our results, they reported that SCs only transiently expressed β₄ at early times in development; only perineurial cells expressed β₄ in adult nerve. Why SCs in adult nerves did not stain in their study is unclear, but we believe that differences in sensitivity related to the antibody, fixative or species could explain this discrepancy. We have carried out western blot, northern blot and immunohistochemical analyses for β₄ in parallel. Thus, in developing nerves, we find that both β₄ mRNA and protein increase together. Since SCs, the predominant cell type in peripheral nerve, account for the majority of nerve mRNA and since SC proliferation is remarkably reduced by the time of increased β₄ expression, these data suggest that β₄ expression increases per cell in developing SCs. In addition, SCs express significant amounts of β₄ by immunohistochemistry. Similarly, in acutely degenerating nerve, the levels of β₄ mRNA and protein fall together, and the intensity of β₄ staining in SCs also falls relative to that of perineurial cells. Finally, we have shown that cultures of purified SCs express β₄ mRNA. These results demonstrate conclusively that SCs express β₄ in both developing and adult rat peripheral nerve, and further work will be required to determine whether this is also the case for human.

The axonally dependent expression of β₄ in SCs in vivo is consistent with the findings of Einheber et al. (1993), who characterized the expression of integrins in SC co-cultured with neurons. They found by immunohistochemical and immunoprecipitation analysis that SCs cultured alone principally express α₆β₁, while myelinating SCs cultured with neurons principally express α₆β₄. Conversely, if the axons from co-cultures were removed by selective axotomy, β₄ staining in previously myelinating SCs fell remarkably. By immunoelectron microscopy, β₄ was concentrated in the outer plasma membrane of the myelinating SCs, which apposes the basal lamina. However, in contrast to our data, this study found that β₄ was also expressed in the Schmidt-Lanterman clefts. Also, they compared the expression of β₄ to that of myelinassociated glycoprotein (MAG), and found that β₄ was not expressed significantly in SCs that had contacted axons but did not express MAG, while we clearly demonstrated β₄ expression in non-myelinating SCs. These differences may be due to technical differences related to the antibodies and staining methods employed in these studies. Alternatively, important differences have been identified between SCs in coculture with neurons and SCs in peripheral nerve depending on the exact culture conditions (Fernandez-Valle et al., 1993). None-the-less, our findings in vivo agree with the major findings of this work in vitro: the expression of β₄ in myelinating SCs is polarized and axonally dependent.

The fall in β₄ mRNA and protein after axotomy is consistent with the observation that axotomy affects the expression of many genes in SCs (reviewed by Mirsky and Jessen, 1990; Scherer and Asbury, 1993). For instance, the mRNA levels of the myelin genes such as Po fall after axotomy and return if, and only if, axons are allowed to regenerate. In contrast, the mRNA levels of other genes, such as NGFR and GAP-43, increase after axotomy (Heumann et al., 1987; Plantinga et al., 1993). These changes in gene expression are thought to reflect activation of new transcription (Lemke and Chao, 1988), which is also likely to be the case for β₄, although changes in the rate of transcription have not been determined directly for any gene in SCs.

Of note, the axonal dependence of β₄ expression in SCs does not extend to the pathological state of chronically axotomized nerve. Unlike Po mRNA, which remained at low levels in the absence of regenerating axons, the level of β₄ mRNA increased progressively in permanently transected nerves. We believe that most of this β₄ mRNA originates in SCs, as β₄ immunoreactivity of perineurial cells remained constant, whereas the β₄ immunoreactivity of SCs increased between 8 and 58 days post-transection. Although microfasciculation of the endoneurium by perineurial-like cells (which might express β₄) is well described in chronically transected nerves, it is not found until 18 weeks after transection (Weinberg and Spencer, 1978), well after the longest time studied here. Furthermore, there is little precedent for axotomy-induced changes in perineurial gene expression. For instance, the mRNA level of neurothelin, a cell adhesion molecule expressed by perineurial cells and not by SCs (Schlosshauer and Herzog, 1990), is unaffected by nerve injury (S. Scherer, personal observation). The finding that perineurial cells are the only cell type expressing β₄ at 8 days post-transection, at a time when β₄ mRNA and protein are almost undetectable by western and northern blot analysis, also suggest that SCs account for the majority of β₄ expression detected in whole peripheral nerve.

Since SCs appear to be the main source of β₄ mRNA in chronically axotomized nerve, then there must be a stimulus besides axons that increase its expression in this setting. One candidate is the basal lamina of SCs. This is a tubular structure that is equivalent to the basement membrane of epithelia. All of the cell types that express β₄, including perineurial cells, have a basement membrane. Furthermore, decreased expression of basement membrane, or its components (such as laminin) has been associated with the loss of polarized β₄ expression, or decreased overall expression of β₄ in breast tumor cells (Natali et al., 1992b). Since SC basal lamina remains intact after axotomy (Thomas and Ollson, 1984), it could account for continued β₄ accumulation in chronically denervated SCs. Similarly, in developing nerve, basement membrane might function as an additional stimulus for β₄ expression beyond that of axons, and could explain continued β₄ mRNA accumulation in development relative to other axonally dependent mRNAs (e.g. Po), which peak earlier in development. Thus, the axons could induce the expression of β₄, resulting in the insertion of α₆β₄ into the abaxonal SC membrane. In the model of Clark and Bunge (1989), these ECM receptors could permit more complete assembly of basement membrane and, therefore, a more potent stimulus for further β₄ expression.

An alternative explanation for the accumulation of β₄ in permanently transected nerves is that β₄ expression might be induced by a factor that is found in axotomized, but not in intact, nerve. Axotomized nerve is a rich source of trophic factors and cytokines (Scherer and Asbury, 1993). For instance, the level of TGF-β₁ mRNA increases in lesioned peripheral nerves (Rogister et al., 1993; Scherer et al., 1993), and TGF-βs increase the expression of β₄ in keratinocytes (Sollberg et al., 1992). While neither basal lamina nor TGFbetas have been shown to cause increased β₄ expression in chronically denervated SCs, there must be another stimulus besides axons, which are not present.

We demonstrated that both α₆ and β₄ are co-localized at the abaxonal border of myelinating SCs, apposing the endoneurial basal lamina in mature peripheral nerve. The localization of α₆β₄ in myelinating SCs is similar to that in epithelial cells, in which α₆β₄ is localized to the basal surface, which apposes the underlying basement membrane. Each axon-SC unit may be viewed as a circular fragment of epithelium, with the external aspect of the unit, the abaxonal SC membrane, covered by a basal lamina (Bunge and Bunge, 1983). Similarly, when epithelial cells are in contact with basement membrane, they form specialized membrane domains, basolateral and apical, and acquire spatial polarity. In corneal keratinocytes, the topography of α₆β₄ expression is regulated such that, in nonpolarized, migrating keratinocytes, it is expressed diffusely over the cell surface while, in polarized, stationary keratinocytes, α₆β₄ is strictly concentrated on the basal surface, in contact with the basement membrane. The polarized pattern of α₆β₄ expression in apposition with the basal lamina, therefore, establishes an intriguing analogy between myelinating SC and epithelial cells, and may support the potential role of α₆β₄ in myelinating SC differentiation.

In contrast, we found that β₁ integrin, the other β subunit that can interact with α_6, was expressed by myelinating SCs, as well as in all of the principal cell types of peripheral nerves: the epineurial fibroblasts, the perineurial cells, the non-myelinating SCs and the axons themselves, in agreement with previous reports (Lefcort et al., 1992, and data not shown). Our northern blot analysis suggested that the levels of β₁ mRNA did not change significantly during development or after nerve injury. This is confirmed by β₁ immunohistochemistry of injured nerve. Besides the loss of axon-related β₁ immunore-activity, the levels of β₁ immunoreactivity did not fall in nerves that had been transected, including in the perineurium, and may even rise in axotomized SCs (S. Scherer, personal observation). These data provide further evidence that β₁ and β₄ are differentially regulated in SCs.

The differential regulation of β₄ as compared to α₆ and β₁ expression during sciatic nerve development also supports a role for α₆β₄ in the differentiation of all SCs. Since α₆ preferentially forms more stable heterodimers with β₄ than with β₁ (Giancotti et al., 1992), the induction of β₄ expression may act to ‘switch’ the profile of integrin heterodimer expression from α₆β₁ to α₆β₄ in SCs of developing nerve. Although we did not directly measure the expression of α₆β₁ or α₆β₄ in developing nerves, the increase in β₄ levels in development suggests that when β₄ appears it could compete with β₁ for association with α₆ in SCs. This is supported by the data that primary SC cultured alone express mainly α₆β₁, while SCs in co-culture with neurons express α₆β₄ (Einheber et al., 1993). Hence, the induction of β₄ expression could have important functional consequences. For example, if α₆β₁, a laminin receptor, is predominant in cultured SC and early developing nerve, then it might be involved in SC proliferation. Laminin is a mitogen for cultured SCs (McGarvey et al., 1984), and SC proliferate robustly in rodent nerve during the first week after birth (Brown and Asbury, 1981). In contrast, although the ligand for α₆β₄ may be related to laminin in carcinoma-derived cell lines, the actual ligand for α₆β₄ in SCs is unknown, as its receptor function may be cell specific, or differentially regulated by the presence of other integrin receptors (Lee et al., 1992). The increase in β₄ mRNA and, therefore, α₆β₄ expression in developing nerve, is well after the peak of SC proliferation and even after the peak of myelin gene expression. It appears to parallel best the growth of the myelin sheaths (Wood and Engel, 1976), suggesting that α₆β₄ might function in myelination.

The possible involvement of α₆β₄ in myelination is consistent with several features of myelinating SCs. First, axons are required for SCs to form and maintain a myelin sheath (Bunge, 1992). Second, SCs need a basal lamina in order to ensheathe axons fully and as well as to form a myelin sheath, and the induction and full organization of basal lamina construction is also axonally dependent (Bunge, 1992). Finally, as Bunge and Bunge (1983) emphasize, myelinating SCs are polarized. They hypothesize that ECM component receptors are necessary for full basement membrane formation by SCs, and that axons induce the polarized expression of these receptors (Clark and Bunge, 1989). α₆β₄ is a strong candidate for this ECM receptor since, in normal nerve, the expression of β₄, and hence α₆β₄, is axonally dependent. Furthermore, the expression of α₆β₄ is polarized on the surface of SCs that apposes their basal laminae, which contain potential ligands of α₆β₄. Therefore, α₆β₄ is appropriately poised to mediate the SC-basal lamina interactions that are necessary to initiate and maintain the sequence of events leading to myelination.

If α₆β₄ plays a role in myelination, it might do so by permitting SCs to change position or shape as required in the process of myelination. One possibility is that α₆β₄ might anchor SCs in position within their basal lamina tubes. Bunge and co-workers have suggested that the SCs must be anchored relative their basal lamina, in order to account for the observed restricted movement of a SC nucleus relative to the axon during myelination (Bunge et al., 1989). In other epithelial cells, α₆β₄ may provide polarized anchorage by interacting with their basement membranes (Carter et al., 1990; Jones et al., 1991; Quaranta and Jones, 1991; Sonnenberg et al., 1991). Second, SCs must have intracellular structural elements to generate the mechanical force necessary to advance the inner lip of mesaxon that encircles the axon. Bunge (1992) has proposed that the basal lamina not only enables SCs to acquire polarity, but that it communicates with the cytoskeleton via a transmembrane receptor in order to effect this change in cell shape. The unusually long cytoplasmic domain of β₄ (Hogervorst et al., 1990; Suzuki and Naitoh, 1990; Tamura et al., 1990), suggests that α₆β₄ could signal from the basal lamina to the necessary intracellular components. Because the integrin α₆β₄ is found in hemidesmosomes (Stepp et al., 1990), a physical association with intermediate filaments is possible (Jones et al., 1986). Furthermore, β₄ has been shown to be associated with cytokeratins (Gomez et al., 1992). However, myelinating SCs express vimentin, nestin and neurofilament protein subunits, but not cytokeratins (Neuberger and Cornbrooks, 1989; Friedman et al., 1990; Kelly at al., 1992). The possibility of physical interactions between α₆β₄ and these types of intermediate filaments remains to be explored. Finally, α₆β₄ integrin, like other integrins, may also act as a receptor activating signalling pathways necessary for SC myelination, although specific evidence predicting this function for α₆β₄ in SCs is lacking. Perturbation of the function or expression of α₆β₄ in SCs will clarify its potential role in SC differentiation and myelination.

We thank Dr Michael Shy for his support to M. L. F. at the beginning of this project, Daniel Roling for excellent technical assistance, Dr M. Hollers for the generous gift of the β₁ cDNA, Dr C. Buck for the generous gift of the β₁ antibody, and Dr J. Salzer for sharing data prior to publication. M. L. F. was supported by IRCCS San Raffaele. This work was also supported by the following NIH grants: NS-01565 (S. S. S.), GM-46902 (V. Q.), and NS-08075 (L. W.); a grant from the Muscular Dystrophy Association (J. K.) and a grant from Telethon, Italy (Projects 251, 466; R. F., L. W.).

Niessen and colleagues (1994) have demonstrated the presence of α6A and β4 integrins in both myelinating and non-myelinating SCs, as well as in the perineurium of human peripheral nerve. Furthermore, they have shown directly by immunoprecipitation that α6Aβ4 heterodimers exist in peripheral nerve, but not in SCs cultured in the absence of Fsk.