Differential effects of laminin and merosin on neurite outgrowth by developing retinal ganglion cells

One of the most striking properties of laminin, unique amongst matrix proteins, is its ability to induce profuse neurite outgrowth by cultured embryonic peripheral and central neurons (Manthorpe et al. 1983; Rogers et al. 1983). In the case of CNS neurons, it was unclear what the physiological relevance of this in vitro behaviour might be. Thus, laminin was thought to be confined to basal laminae and excluded from the CNS and, therefore, beyond the reach of growing axons. More recent work, however, describing spatial and temporal patterns of laminin immunoreactivity at non-basal lamina sites along axon projection pathways in the developing avian (Cohen et al. 1986, 1987; Halfter and Fua, 1987) and mammalian CNS (Letourneau et al. 1988; Liesi and Silver, 1988; McLoon et al. 1988), are consistent with it playing a similar axon growth-promoting role in vivo to that seen in vitro. We have shown however, that for embryonic chick retinal ganglion cells (RGC), the ability to grow neurites on EHS laminin in vitro is lost during development around the time when, in vivo, RGC axons grow into their target in the brain, the optic tectum (Cohen et al. 1986, 1987). This is a consequence of the loss or modification of RGC laminin receptors (Cohen et al. 1989; Neugebauer and Reichardt, 1991), in a process possibly governed by the target (Cohen et al. 1989). The identity of the cell surface receptors for laminin expressed by neurons has recently come under close scrutiny (Edgar, 1989; Reichardt and Tomaselli, 1991). In particular, antibody blocking studies have suggested that <q31 integrins mediate neurite outgrowth on laminin by both central and peripheral neurons in culture (Bozyczko and Horwitz, 1986; Cohen et al. 1986, 1987; Tomaselli et al. 1986; Hall et al. 1987). Of the eight different α sub-units (αl-α7 and α_v identified that can associate with β1, four-α1, α2, α3 and α6,-can function as laminin receptors on a variety of cell types (Sonnenberg et al. 1990), although only the α6β1 dimer is specific for laminin. So far, however, there is only evidence from work with PC 12 cells for the involvement of al and o3 in mediating neurite outgrowth on laminin (Turner et al. 1989; Tomaselli et al. 1990), and no information is available on α sub-unit function in neurons. The recent demonstration that embryonic chick RGC express a6 mRNA and protein, and that these show a substantial decline in amount with maturation (de Curtis et al. 1991), however, provides strong circumstantial evidence that the aS pl integrin dimer mediates the interaction of these cells with laminin.

The structure of laminin was originally established for the molecule secreted by the Engelbreth-Holm-Swarm (EHS) sarcoma. EHS-laminin is a trimer made up of A, Bl and B2 subunits, all separate gene products that display substantial homology (Sasaki et al. 1988; Timpl, 1989). The purification and molecular cloning of homologues of the Bl chain, s-laminin (Hunter et al. 1989) and of the A-chain, called merosin or M-chain (Ehrig et al. 1990), from normal tissues, has focussed attention on the structural diversity of laminin and its functional implications. Thus, a recent immunohistochemical study with chain-specific antibodies has revealed an unexpected extent of heterogeneity in the laminin isoform composition of basal laminae at different sites in adult tissues (Sanes et al. 1990). These findings raise the possibility that different combinations of laminin subunits (of which there are at least four possible trimers), may confer on laminin, at different sites, distinctive properties determined by subunit composition. Thus, peripheral nerve endoneurium, a preferred substratum for regenerating peripheral axons after nerve injury, is one of the few select sites expressing M-chain in place of A-chain. Also, injured adult rat RGC axons, which normally fail to regenerate, will re-grow into sciatic nerve implants grafted into the optic nerve (Vilegas-Perez et al. 1988; Berry et al. 1988). In this context, it is not known whether rodent RGC exhibit an age-dependent response to laminin similar to avian RGC, however, and there are reports (Kleitman et al. 1988a,6) that even early embryonic (day 15), rat RGCs fail to extend neurites on laminin.

In this paper we show that early embryonic mammalian RGC share with avian RGC the ability to extend neurites on laminin in vitro. Using a blocking monoclonal antibody GoH3, to the human and murine α2 integrin sub-unit (Sonnenberg et al. 1988), we also show that laminin-induced neurite outgrowth by mammalian RGC is mediated primarily by α6μ21 integrins. The laminin-respon-siveness of rodent RGC, like that of chick RGC, declines during embryonic development and is lost before birth. Merosin, unlike EHS-laminin, promotes neurite outgrowth by RGC isolated from chick embryos at advanced stages of development (up to hatching age), and from postnatal rats. Growth of chick RGC on both laminin and merosin substrata is blocked by the anti-β1 integrin antibody CSAT at all ages.

For the preparation of either chick, rat or mouse dissociated embryonic retinal cultures, the dissected tissue was incubated at 37°C with papain (10 Uml⁻¹; Worthington), in a Hepes-buffered salt solution containing cysteine (Leifer et al. 1984), for between 15-25 min depending on embryonic age. The tissue was then transferred to a calcium- and magnesium-free Hanks buffered salt solution (CMF; Gibco) containing 5 mg ml⁻¹ BSA (fraction V, fatty-acid free; Sigma) and 50μgml⁻¹ DNAase I (Sigma) and triturated 10 times through a narrow-bore plastic pipette tip to give a single cell suspension. This was diluted in modified Bottenstein and Sato’s culture medium containing Ham’s F12 (Gibco) and 2% fetal calf serum (BSF2; Cohen et al. 1986). 20000-40000 cells were plated onto 13 mm diameter glass coverslips or plastic LabTek slides (Nunc), precoated with an aqueous solution of poly-D-lysine, (PL; 2mgml⁻¹; Sigma), alone, or with PL followed by EHS-laminin (Gibco BRL) or human placental merosin (Telios Pharmaceuticals), both at 5/igmU¹ in F12 (Cohen et al. 1987). When blocking antibodies were to be used, these were added to the wells of LabTek slides diluted in BSF2 medium at double the required concentration, prior to the addition of an equal volume of cell suspension.

For explant cultures, embryonic and postnatal rat and mouse retinae were dissected and placed in Eagles minimal essential medium (Gibco), buffered with 25 DIM Hepes (MEM-H), cut into approximately 0.5 mm squares and plated on pre-coated coverslips (see above) in BSF2 medium supplemented with methyl cellulose (Johnson et al. 1988).

RGC were purified from papain-dissociated chick or rat retinal cell suspensions using species specific anti-Thy-1 antibody-mediated plate adhesion (‘panning’), essentially as described by Barres et al. (1988), except that in the final step adherent Thy-1 positive cells were simply removed by vigorous washing in BSF2. Highest purities (>90 % RGC) were achieved for chick retina aged between embryonic day (E)10 and E14, and for rat retina between birth and postnatal day (P) 10.

Antibody blocking. Monoclonal antibodies GoH3 (culture supernatant; a gift of A. Sonnenberg), and CSAT (ascites fluid; a gift of A. Horwitz), to be used for blocking neurite outgrowth, were initially dialysed overnight against F12, to remove sodium azide, and diluted in BSF2 at double the final concentration, i.e. 1:100 for both antibodies.

Retinal cultures were labelled with a number of antibodies to detect the expression by RGC of the following antigens. Thy-1: mouse monoclonal antibody MRC-OX-7 against the rodent molecule (Beale and Osborne, 1982) and rat monoclonal antibody 3OH-12 (Ledbetter and Herzenberg, 1979) against murine Thy 1.2, and a mouse monoclonal antibody to chick Thy-1 (Cohen et al. 1986). Neurofilaments: mouse monoclonal antibodies RT97 (Wood and Anderton, 1981) and 2H3/3A10 (Developmental Studies Hybridoma Bank, USA), for the 200 and 155 kDa subunits and 68K, a rabbit antibody to the 68 kDa subunit (gift of P. Hollenbeck). GAP-43: rabbit antibody to mammalian GAP-43 (Curtis et al. 1991). All antibodies for immunofluorescence staining were diluted prior to use in MEM-H and 10 % fetal calf serum with 0.05 % azide. For the demonstration of Thy-1 surface antigen, cultures were lightly pre-fixed in 2 % paraformaldehyde for 10 min at room temperature, washed in PBS and incubated for 30 min in diluted primary antibodies, then with biotinylated antimouse, anti-rat or anti-rabbit antibodies (Amersham; 1:100) for 30 min, and finally with fluorescein-conjugated Streptavidin (Amersham; 1:100), for 30 min. The cultures were washed and then permeabilised by fixing for 5 min in 100% methanol at-20 °C, washed and antibodies to intracellular antigens then applied for 30 min. After washing in PBS, labelling was completed with a 30 min incubation in rhodamine-conjugated anti-mouse or anti-rabbit antibodies (Cappel). Finally, the cultures were washed in PBS, mounted in glycerol containing Dabco (Sigma) to prevent fading of the fluorescence, and viewed on a Zeiss Axiophot microscope.

Explant cultures of E15 rat retina consistently failed to extend neurites on a substratum of PL (Fig. 1A and Fig. 3) and neurofilament-positive processes were confined within the body of the expiants. By contrast, 100 % of expiants of this age were able to grow neurites on a PL+laminin substratum (Fig. IB and Fig. 3). The majority of neurites originated from RGC as shown with the monoclonal anti-Thy-1 antibody OX-7, an RGC-specific marker (Fig. 1C). Similar results were obtained when E14 mouse retinal tissue was cultured (data not shown).

In cultures of dissociated retinal cells from either E15 rat (Fig. 2A,B), or E14 mice (Fig. 2C,D), laminin also induced neurite outgrowth from Thy-1⁺ RGC.

We have previously shown that embryonic chick RGC lose their ability to extend neurites on laminin as they mature (Cohen et al. 1986,1987). We examined the possibility that rodent RGC display a similar phenomenon, by culturing embryonic rat retinal tissue between E15 and E19. Fig. 3 shows that a small decline in the proportion of expiants extending neurites on laminin occurred between E15 and E17, (approximately 20%). By E19 however, the ability of RGCs to extend neurites on laminin was almost completely lost and at this stage the small extent of outgrowth was indistinguishable from that seen on PL (Fig. 3). Retinal expiants from P0 through to P7 behaved in an identical manner to those from E19 retina (data not shown).

Previous work has shown that neurite outgrowth on laminin by chick embryonic RGC is mediated by β1 integrins. To investigate further the specificity of the integrin receptor subclass involved, we used a rat monoclonal antibody (GoH3) directed against the murine a6 integrin (Sonnenberg et al. 1988). In these experiments, cultures of dissociated mouse E14-E16 retina were grown on laminin substrata in the presence and absence of antibody GoH3 or a rabbit anti-β1 integrin antibody (gift of Dr K. Tomaselli). The extent of neurite outgrowth in these cultures was estimated after 24 h by measuring the individual neurite lengths of neurofllament-positive neurons using image analysis. Fig. 4 shows that both of the anti-integrin antibodies greatly reduced the extent of neurite outgrowth: Rab anti-βl, by >80 % (Fig. 4C); GoH3 by approx. 50% (Fig. 4B). Also, there was a substantial fall in the proportion of RGC with processes: Rab anti-/31: 84% inhibition; GoH3: 75% inhibition (data not shown).

Both EHS-laminin and merosin were effective in eliciting neurite outgrowth from E6 chick RGC in mixed retinal cultures (Fig. 6A). In contrast, double-labelled Thy-1⁺/68K⁺ E9 or older chick embryo RGC, either in mixed cultures, or purified by panning on anti-Thy-1 antibody coated dishes, failed to extend neurites on laminin, as previously reported (Cohen et al. 1987, 1989), but grew profusely on merosin (Figs 5, 6A). Thus, between 50-60% of RGC extended neurites on merosin from E9 to hatching age, compared with only 5-15 % on laminin. Also, rat Thy-l⁺/GAP-43⁺ RGC, from animals as old as P15, grew lengthy neurites on merosin (data not shown). Thus, the loss of response to laminin by older RGC was not simply a consequence of the possible onset of trophic factordependent survival (Rodriguez-Tebar et al. 1989). In fact, many RGC, including cells purified by panning, grown on merosin survived for more than 4 days in culture in the absence of added factors.

To establish whether the interaction between RGC and merosin was mediated by fil integrins, chick RGC neurons were cultured on laminin or merosin with or without monoclonal antibody CSAT (1:200 dilution ascites fluid). Fig. 6B demonstrates that CSAT antibody caused almost complete (>85 %) inhibition of neurite outgrowth on merosin at all ages tested (E9-newly hatched).

In the present study we have shown that early embryonic mammalian RGC share the ability previously demonstrated in avian RGC, to extend neurites in vitro on a laminin substratum. Similarly, this growth response on laminin in mammals is restricted to a comparable phase of early RGC development, prior to target encounter by RGC axons. Our findings appear to contradict those reported earlier by Kleitman et al. (1988a), in which cultured expiants of embryonic rodent retina failed to grow neurites on laminin. The explanation for this discrepancy does not reside in a difference between the age of animals employed by these workers and those in our study, since they used day 15 embryos, a stage at which we have shown RGC are responsive to laminin. However, other differences, including those of both techniques of explant culture and substratum preparation, may account for the conflicting findings.

Monoclonal antibody GoH3 was originally identified as recognising the human ri<5 integrin subunit on platelets mediating binding to laminin (Sonnenberg et al. 1988). Here we show for the first time by function blocking, that this antibody defines a major receptor involved in murine RGC neurite outgrowth on laminin. Since neurite outgrowth on laminin is also inhibited by anti-β1 integrin antibodies, it is likely that on RGC α6 subunits are associated with β2 molecules as functional cell surface heterodimers. Moreover, whilst the α6 subunit is able to associate with two β subunits, β1 or β4, (Sonnenberg et al. 1990), only the α6β1 dimer functions as a laminin receptor. Taken together with recent work describing the synthesis by embryonic chick RGC of a6 integrin mRNA and protein, and their abrupt decline in expression between E6 and E12 (de Curtis et al. 1991), our results are consistent with α6 integrins being the target for the regulation of RGC laminin-responsive functions during development.

Using saturating concentrations of GoH3 antibody, we were unable to achieve greater than 50% inhibition of RGC neurite outgrowth. Since antibodies to β1 integrins were able to inhibit almost completely, however, it is possible that other laminin-binding β1 integrin heterodimers function in the interaction of these neurons with laminin. The present general unavailability of antibodies specific to avian and rodent integrin α chains has not allowed us to make a comprehensive analysis of their functional role. However, we were unable to demonstrate an inhibitory effect of antibody 3A3, directed against the rodent al integrin subunit (Turner et al. 1989), on rat RGC neurite outgrowth on laminin (J.C. and A.R.J., unpublished observations). Together, these findings highlight an interesting difference between CNS and PNS neurons in their respective interactions with laminin. Thus, rodent and human DRG neurons express aipi integrin heterodimers which function in neurite outgrowth on laminin (K. Tomaselli, personal communication).

Our previous correlative studies, in vivo and in vitro, suggested the possible biological significance of the transient responsiveness of embryonic RGC to laminin (Cohen et al. 1986, 1987). In vivo however, the growth cones of chick RGC axons continue exploratory migration within the tectal neuropil for some days after they grow into the tectum, until contact is made with their target neurons (Rager and von Oeynhausen, 1979; Thanos and Bonhoeffer, 1987). The demonstration here that merosin elicits neurite outgrowth from both early and late stage embryonic RGC raises the possibility that it may be involved in the later phases of optic axon pathfinding, including terminal branching in the neighbourhood of their target cells. Our finding that RGC extend neurites on merosin regardless of embryonic age however, suggests that expression of the appropriate receptors is not, unlike RGC laminin receptors, under the influence of tectal factors (Cohen et al. 1989). Morever, the effective inhibition of RGC merosin receptor function by CSAT antibody suggests that different αβ1 integrin dimers mediate the binding of laminin and merosin on RGC.

A related issue concerns the identity of molecules that may promote the regeneration of mature CNS axons. Whilst injured adult optic axons normally fail to re-grow, they will do so if the optic nerve environment at the lesion site is modified by a peripheral nerve graft containing viable Schwann cells (Villegas-Perez et al. 1988; Berry et al. 1988). Also Schwann cells will support RGC axonregeneration, when co-cultured with either embryonic (Kleitman et al. 1988a,6) or adult (Hopkins and Bunge, 1991) rat retinal expiants. While the cell adhesion molecule LI/NgCAM appears to play the major role in Schwann cell-mediated embryonic neurite-outgrowth (Bixby et al. 1988; Seilheimer and Schachner, 1988; Kleitman et al. 19886), antibodies to Ll/NgCAM have no effect on regeneration of adult RGC in these co-cultures (Hopkins and Bunge, 1991). In view of our findings however, that mature RGC regenerate successfully on binding to merosin in vitro, and the evidence that merosin is the major laminin isoform expressed by Schwann cells in vivo (Sanes et al. 1990), it is likely that a closer examination of the molecular details of neuron-merosin interactions will also help to elucidate the mechanisms of axon regeneration.

We thank A. Sonnenberg, A. Horwitz, G. Wilkin and R. Curtis for gifts of antibodies used in this work. We are grateful to Pat Doherty for carrying out the image analysis illustrated in Fig. 4, and to Vanessa Harrison and Zoey Jackson for technical assistance. This work was supported by grants from Action Research and the International Spinal Research Trust.