identification of integrin (13131 as a neuronal...

Neuron, Vol. 15,333.-343, August, 1995, Copyright © 1995 by Cell Press

Identification of Integrin (13131 as a Neuronal Thrombospondin Receptor Mediating Neurite Outgrowth Michael F. DeFreitas,*§ Cathleen K. Yoshida," William A. Frezler,t Donna L. Mendrick,$, Robed M. Kypta,* and Louis F. Reichardt* *Department of Physiology Howard Hughes Medical Institute University of California, San Francisco San Francisco, California 94143-0724 tDepartment of Biochemistry and Molecular Biophysics Washington University School of Medicine St. Louis, Missouri 63110 tDepartment of Pathology Brigham and Women's Hospital Harvard Medical School Boston, Massachusetts 02115

Summary

Thrombospondins are a family of extracellular matrix proteins expressed throughout the developing nervous system that promote neurite outgrowth in vitro and help mediate the migration of granule cells across the molecular layer in explants of neonatal cerebellum. The receptors mediating these Interactions have not previously been Identified. In this study, monoclonal antibodies raised to the integHn (][31~1 heterodlmer are shown to inhibit neurite outgrowth by rat sympathetic neurons on thrombospondln-1.0~1~1 is found to be expressed on the cell body, neurltes, and growth cones of sympathetic neurons in vitro and on sympathetic axons passing through the thrombospondin-rich outer sheath of the superior cervical ganglion in vivo, consistent with its role in mediating axon outgrowth. A receptor-Iigand binding assay is used to demonstrate the direct binding of Immunopurified 0~1 to thrombospondin-1. These results demonstrate a direct Interec- tion between the integrin (131~1 and thrombospondln-1, which mediates neurite outgrowth in vitro and is likely to mediate the same Interactions in vivo.

Introduction

During neuronal development, axons extend through a complex extracellular environment composed of a variety of adhesive substrates that promote axon outgrowth. Among these are extracellular matrix (ECM) components, such as laminin, collagen, fibronectin, vitronectin, tenas- cin, and thrombospondin (TSP) (reviewed in Reichardt and Tomaselli, 1991). These ECM components promote attachment and process outgrowth of several different neu-

§Present address: Department of Molecular and Cell Biology, Univer- sity of California, Berkeley, LSA Room 221, Berkeley, California 94720-3200. lIPresent address: Human Genome Sciences, Rockville, Maryland 20850.

ronal cell types in vitro and in many cases are expressed in appropriate spatiotemporal patterns during development to promote and guide axonal outgrowth in vivo.

Although TSP in the adult is mainly found in platelets, polyclonal antibodies to platelet TSP stain the developing nervous system of the mouse at times of axon outgrowth and cell migration, disappearing after such processes are complete (O'Shea and Dixit, 1988). It is now apparent that there are several TSP isoforms (TSP1-TSP4; Bornstein et al., 1991; Vos et al., 1992; Lawler et al., 1993) and TSP-like proteins, such as F-spondin and cartilage oligomeric protein (Klar et al., 1992; Oldberg et al., 1992) en- coded by distinct genes, which demonstrate a high degree of sequence homology with platelet TSP (TSP1). Because of antibody cross-reactivity between the TSPs, the expression patterns of these TSP isoforms have been best stud- ied by in situ hybridization. Iruela-Arispe et al. (1993) have shown that TSP1, TSP2, and TSP3 are all expressed in the developing murine embryonic nervous system, with TSP1 being expressed earliest, followed by TSP2 and later by TSP3. Expression of all three isoforms decreases significantly after birth. The expression pattern of TSP4 is not well characterized (Lawler et al., 1993). Their expression patterns suggest that TSP isoforms may play a role in cell migration and axon outgrowth during embryogenesis. Consistent with its widespread distribution, many distinct types of neurons, including sensory, sympathetic, spinal cord, and retinal neurons, extend processes on TSP1 in vitro (O'Shea et al., 1991; Osterhout et al., 1992; Neuge- bauer et al., 1991). In addition, anti-TSP antibodies inhibit migration of cerebellar granule cells in explants of neonatal cerebellum (O'Shea et al., 1990).

TSP was originally identified as a major component of platelet ~ granules that, upon release, functions in platelet aggregation (for review, see Frazier, 1991; Bornstein, 1992). TSP1, the TSP isoform expressed by platelets, is a homotrimeric multidomain glycoprotein with each mono- mer containing an amino-terminal heparin-binding domain, followed by a procollagen-like region, three properdin-like repeats (TSP type 1 repeats), three epidermal growth factor-like repeats (TSP type 2 repeats), seven Ca2+-binding repeats (TSP type 3 repeats), and a carboxy- terminal cell attachment domain. TSP2 shares this same structural layout, whereas TSP3 and TSP4 lack the procollagen domain and the TSP type 1 repeats and have four TSP type 2 repeats. Cell attachment sites for a variety of nonneuronal cells have been localized to many of these domains in TSP1 (Frazier, 1991; Prater et al., 1991). In particular, neurite outgrowth from rat sympathetic neurons on TSP1 has been shown to be inhibited by an antibody recognizing the type 1 repeats of TSP1 (Osterhout et al., 1992). Attachment of retinal neurons to TSP1, but not neurite outgrowth, is inhibited by heparin, implicating the heparin-binding domain as a neuronal cell attachment site (Neugebauer et al., 1991). Thus, neurons, like other cell types, appear to interact with more than one domain of TSP1.

Neuron 334

For many neurons, recognition of ECM components is due in large part to integrins (reviewed in Reichardt and Tomaselli, 1991). Integrins are heterodimers composed of noncovalently associated a and I~ subunits. Integrins containing the I~1 subunit, which can associate with at least ten distinct a subunits, are particularly important for neuronal interactions with many ECM substrates (Reichardt and Tomaselli, 1991). Previous work has implicated a ~ integrin as a TSP1 receptor on retinal neurons that mediates neurite outgrowth (Neugebauer et al., 1991). Integrins proposed to be TSP1 receptors include czvl~3 in endothelial cells (Lawler et al., 1988), m,bl~a on melanoma cells (Tus- zynski et al., 1989), and a41~ and as~ in activated T cells (Yabkowitz et al., 1993). In this paper, we identify the integrin aal~ heterodimer as a neuronal TSP1 receptor, recognizing the type I repeats, that mediates neurite outgrowth of rat sympathetic neurons on purified TSP1 in vitro. This receptor is shown to be appropriately expressed to mediate such an interaction in vivo.

Results

Sympathetic Neurite Outgrowth on TSP1 Rat neonatal sympathetic neurons from the superior cervical ganglia (SCGs) were grown in serum-free medium for 18-24 hr on plastic coated with TSP1. In agreement with earlier work (Osterhout et al., 1992), sympathetic neurons extended neurites on TSP1 in a dose-dependent manner, requiring a coating concentration of 6-12 pg/ml to support outgrowth (Figure 1A). QuantitaUon of outgrowth either in terms of percentage of neurons with neurites or number of neurites per unit area showed a similarly steep dose dependence on TSP1 concentration, with about 90% of the cells extending processes at a coating concentration of 50 pg/ml. The effects of function-blocking reagents on sympathetic neuron process outgrowth on TSP1 were de- termined using coating concentrations of 10-20 pg/ml, and outgrowth was quantitated in terms of number of neurites per unit area. Outgrowth on purified TSP1 was inhibited by an antibody to TSP1 (Figure 1B), demonstrating the substrate specificity.

Sympathetic Neurite Outgrowth on TSP1 Is Inhibited by Anti-I~ Integrln Antibodies Experiments by Neugebauer et al. (1991) indicate that a l~-class integrin is involved in neurite outgrowth of chick retinal neurons on TSP1. To investigate the possible role of 13~-class integrins in mediating neurite outgrowth of rat sympathetic neurons on TSP1, the effect on outgrowth of a function-blocking monoclonal antibody (MAt)) was exam- ined. This antibody has previously been shown to inhibit glomerular epithelial cell attachment to laminin and collagen (Mendrick and Kelly, 1993). As shown in Figure 1B, this anti-I~ MAb inhibited neurite outgrowth on TSP1 to 27% ± 7.6% of normal. The antibody almost completely inhibited neurite outgrowth on collagen IV (Col IV) and significantly reduced outgrowth on laminin-1 (LN-1), as has been described previously with a polyclonal anti-I~l antibody (Lein et al., 1991). Outgrowth on the recombinant extrecellular domain of the homophilic, Ca2+-dependent cell-adhesion molecule N-cadhedn was, if anything, slightly increased by this antibody, demonstrating that the antibody is not toxic and has no nonspecific effects on neurite outgrowth.

Characterization of pl-Class Integrlns Expressed by Sympathetic Neurons To determine which integrins are potential TSP1 receptors, we used subunit-specific antibodies to identify the integrins present on these neurons. As shown in Figure 2 (lane 1), anti-131 integrin antibody immunoprecipitates a 13~ integrin band with a size of 112 kDa as analyzed by SDS-PAGE, along with four associated ~z subunit bands of 185, 135, 128, and 122 kDa. The pattern seen here with an anti-I~ cytoplasmic domain IgG is also seen with the function-blocking anti-~ MAb, Ha2/11 (data not shown). The 185 kDa band corresponds to the al subunit (data not shown). The 135 kDa band contains ae, as, and possibly other integrin subunits (data not shown). The 122 kDa subunit has not been identified. The 128 kDa band is immunoprecipitated with a polyclonal antibody to the integrin ,z~ cytoplasmic domain (lane 2). These data show that ml~, ~za131, and additional ~-class integrins are expressed by sympathetic neurons.

.~ 100

¢: 75- ..¢

"~ 5o-

i 25"

0"

B ..c 150-1" 1

I 4 0 0 ~ ¢, 75 • "~ 50 * .

Z 25 * * | 0 z ~ 0

102030405060 N 131 3-1 3-2 3-3 ~xl TSP TSP-1 coating c~3

concentrat ion (p.g/ml)

Figure 1. Sympathetic Neurite Outgrowth on TSP1 (A) Neurite outgrowth of sympathetic neurons in response to various coating concentrations of TSP1. Neurite outgrowth was quantitated after 18 hr in culture, both in terms of percentage of neurons with neurites (squares) and number of neurites in a unit area (circles). Error bars represent the range of duplicate wells in this particular experiment. (B) Inhibition of sympathetic neurite outgrowth on purified substrates. Sympathetic neurons were grown on TSP1 (closed bars), laminin-1 (LN-1; open bars), collagen IV (Col IV; gray bars), or N-cadherin (hatched bars) for 15-18

hr in the presence of different MAbs. Antibodies tested were an anti-I~l antibody (Ha2/11), two function-blocking anti-az MAbs (Ralph 3-1 and Ralph 3-2), a non-function-blocking anti-aa MAb (Ralph 3-3), an anti-a, MAb (3A3), and an anti-TSP MAb, A 4.1 (TSP), as well as a no-antibody control (N). Outgrowth was quantitated in terms of neurites per unit area and normalized to the amount of neurite outgrowth seen with no antibody. Results are the mean values from several experiments ± SEM; asterisk, statistically significant to p < .001.

Integrin a~13~ Is a Thrombospondin-1 Receptor 335

a l - ) ~ m 200

a--)~

1~1--> ~ 1 1 6

~ 9 7

1 2 3 4 5 6 7 8 9 10 11

Figure 2. ImmunoprecipitationoflntegrinsfromSympatheticNeurons and Characterization of Ralph 3-2 Antibody

SCGs were dissociated, grown in culture on a LN-1 substrate, biotinylated, and extracted for immunoprecipitation. Immunoprecipitations were done with an anti-~l cytoplasmic domain antibody (lanes 1, 7, and 11), an anti-~x~ cytoplasmic domain antibody (lanes 2, 4, 5, and 10), and an anti-Q3 MAb, Ralph 3-2 (lanes 3, 6, 8, and 9). Lanes 1--3 show immunoprecipitations directly from a SCG extract. Arrows indicate four a bands, including (xl, as, and two other a bands of unknown composition. Also indicated by arrow is the 131 subunit. The unbound extract from the ~z3A cytoplasmic domain immunoprecipitate in lane 2 was depleted twice more with the anti-a~ cytoplasmic domain antibody (lanes 4 and 5), then immunoprecipitated with the anti-a3 MAb Ralph 3-2 (lane 6) or the anti-~l cytoplasmic domain antibody (lane 7). The unbound extract from the Ralph 3-2 immunoprecipitate in lane 3 was depleted twice more with Ralph 3-2 (lanes 8 and 9), then immunoprecipitated with anti-a~ cytoplasmic domain antibody (lane 10) and anti- ~1 cytoplasmic domain antibody (lane 1 1).

Identification of o~13~ as a TSP1 Receptor Preliminary evidence with an anti-a3 polyclonal antibody suggested that a31~1 might be a I~-class TSP1 receptor on sympathetic neurons, but the specificity of the antibody could not be adequately documented. Therefore, MAbs to the native integrin heterodimer, a31~, purified from RuGli cells, a rat glioblastoma cell line, were generated. The MAbs were screened for the ability to stain live RuGli cells and immunoprecipitate the a31~1 integrin heterodimer. Of those MAbs that fit these criteria, 2 were found to inhibit neurite outgrowth on TSP1. These 2 antibodies, Ralph 3-1 (for rat alpha 3) and Ralph 3-2, immunoprecipitate 128 and 112 kDa proteins. These proteins comigrate with the a3 and ~ integrin subunits immunoprecipitated by an antibody to the a~ cytoplasmic domain (Figure 2, lanes 2 and 3). Positive identification of these two bands as as and ~1 was provided by immunodepleting a biotinylated SCG extract via three rounds of immunodepletion with the anti- a~ cytoplasmic domain antibody (Figure 2, lanes 2, 4, and 5) and then immunoprecipitating with Ralph 3-2 (lane 6) or with an antibody to the ~ subunit (lane 7). The anti-a~ cytoplasmic domain antibody depleted the extract of almost all of the Ralph 3-2 antigen and the 128 kDa band

coimmunoprecipitated by the anti-~l antibody, but did not deplete the extract of any of the other integrin bands immunoprecipitated with the anti-~ antibody. The converse of this experiment, immunodepletion with Ralph 3-2 followed by immunoprecipitation with the anti-a~ or anti-I~ antibody (lanes 3 and 8-11), gave similar results. Some anti-a~ cytoplasmic domain immunoreactivity was not removed with Ralph 3-2 immunodepletion, which may reflect either partial inactivation of the Ralph 3-2 epitope by biotinylation or the lack of the Ralph 3-2 epitope on some of the a3 expressed on sympathetic neurons. Likewise, a very small amount of the a3 band is immunoprecipitated by the Ralph 3-2 antibody after anti-a~ cytoplasmic domain immunodepletion, but is seen only with long exposure times. This may represent expression of a small amount of the alternatively spliced B form of the a3 cytoplasmic domain (Tamura et al., 1991). Similar results were obtained for Ralph 3-1 and a non-function-blocking anti-a3 MAb, Ralph 3-3 (data not shown). Thus, Ralph 3-1, Ralph 3-2, and Ralph 3-3 specifically recognize the integrin a3~.

As shown in Figure 1 and Figure 3, Ralph 3-1 and Ralph 3-2 inhibit neurite outgrowth by sympathetic neurons on TSP1 to 22% + 3.2% and 33°/0 + 5.4% of normal, respectively. Neither antibody detectably perturbed outgrowth on Col IV or LN-1, demonstrating specificity. As an antibody control, a function-blocking MAb to the rat integrin (~ subunit, 3A3, was shown to have no inhibitory effect on neurite outgrowth on TSP1. This same antibody strongly inhibits neurite outgrowth on Col IV to 8.4% + 2.3°/0 of control, as has been seen by others (Lein et al., 1991). As another control, a third anti-a3 MAb identified in our screen, Ralph 3-3, did not inhibit neurite outgrowth. Since a3 is only known to associate with the integrin ~ subunit, and the anti-13~ cytoplasmic domain antibody completely depletes a3 from a sympathetic neuron extract (data not shown), the results with the anti-I~ and anti-a3 MAbs implicate integrin a3131 as a novel TSP1 receptor.

Sympathetic Neurlte Outgrowth on a TSP1 Fragment Our data, in agreement with those of Osterhout et al. (1992), show that outgrowth of sympathetic neurons on TSP1 can be blocked by a MAb that recognizes the properdin-like repeats of TSP1 (see Figure 1B). This suggests that the integrin a~61 heterodimer recognizes the properdin-like type 1 repeats or a nearby region. To test this possibility in a functional assay, a TSP1 chymotrypsin fragment (50/70 kDa) containing the procollagen-like domain, the TSP type 1 repeats, and a portion of the TSP type 2 repeats (Prater et al., 1991) was tested for its ability to support SCG neurite outgrowth. This fragment supported neurite outgrowth in a dose-dependent manner (Figure 4A). Outgrowth on the fragment was completely inhibited by the antibody to the 131 integrin subunit (Figure 4B). This result differs from whole TSP1, where neurite outgrowth is inhibited to only 27%. It suggests that either a non-I~l-Class integrin receptor or a I~l-class integrin receptor resistant to this antibody is present on SCG neurons and recognizes a different domain of TSP1. As with TSP1, anti-a3 MAbs Ralph 3-1 and Ralph 3-2 were found to inhibit neurite out-

Neuron 336

TSP

c~ 6 ,,~,

Col %

N O Ab

Figure 3. Antibody Inhibition of Sympathetic Neurite Outgrowth Neurons were grown on TSP1 (A, C, and E) or Col IV (B, D, and F) for 16 hr in the presence of either no antibody (A and B), 50 p.g/ml anti-a3 MAb Ralph 3-2 (C and D), or 50 p.g/ml anti-~zl MAb 3A3 (E and F). Ceils were fixed and photo- graphed under phase optics. Bar, 100 p.m.

ol3

o~1

A 6 0 - L ~

= 40- == e = 20- E ==

O0 -- 10 20 30 40 50 60 Z

Coating concentration (~g/ml)

I I ,,I 25 *

0 ~ N I~l 3-1 3-2 c~l ~3

Figure 4. Neurite Outgrowth of Sympathetic Neurons on TSP1 Frag- ment (A) Neurite outgrowth on increasing concentrations of TSP1 50•70 kDa fragment. (B) Inhibition of neurite outgrowth on 25 p.g/ml TSP1 50•70 kDa frag-

growth on this f ragment to 20%, whereas the control MAb against a l had no effect (Figure 4B). This suggests that a site for integrin a31~1 interact ion is located within this central region of TSP1.

Immunocytochemical Distribution of (~3~1 on Sympathetic Neurons To examine the distr ibut ion of ~3 on neurons, cells from dissociated neonata l SCGs were cultured for 16-20 hr to al low extension of neurites. The cells were double labeled with the anti-or3 MAb Ralph 3-2, and an ant ibody to the neuron-speci f ic marker, neurof i lament 150 kDa subunit. As can be seen in Figure 5, both the neurof i lament-posi t ive neurons and the neurof i lament-negat ive nonneuronal cells expressed a3 as v isual ized with the Ralph 3-2 antibody. Anti-a3 imm unoreact iv i ty was intense throughout the neuri tes and was part icular ly strong in growth cones. Staining was seen in the growth cone f i lopodia, where

ment with MAbs against ~ (Ha2/11), ct3 (Ralph 3-1 or Ralph 3-2), or al (3A3). Outgrowth was normalized to that seen with no antibody present (N). Results are mean _+ SEM; asterisk, statistically significant to p < .001.

Integrin ~3131 Is a Thrombospondin-1 Receptor 337

Figure 5. Cellular Localization of ~3 (A) and (B) are pictures of the same field as in (C) and (D). Sympathetic neurons were double immunostained with anti-neurofilament and either Ralph 3-2 (A and B) or mouse IgG (C and D). Neurofilament (A and C) was visualized with Texas red. Ralph 3-2 (B) and mouse IgG (D) were visualized with fluorescein isothiocyanate (FITC). Bar, 50 p.m.

neurofilament staining was absent, as well as in the numer- ous small filopodial spikes seen along the length of the neurite. To rule out the possibility that neuronal ~z3 expression was induced by conditions of cell culture, mechani- cally dissociated sympathetic neurons were stained after allowing them to attach to a polyornithine-coated substrate for I hr. Under these conditions, 81% _ 12% of the neurons expressed detectable levels of Q3, with the level of expression varying greatly (data not shown).

To confirm the expression of e~ on sympathetic neurons in vivo, whole SCGs were sectioned and stained with anti- a3 MAb Ralph 3-2, anti-neurofilament antibody, and an anti-TSP1 polyclonal antibody. As seen in Figure 6B, the anti-or3 antibody stained the borders of cells within the ganglion, as well as axons extending within the outer sheath. Because of the high level of expression of ~z3 within the ganglion, individual axons within the ganglion that expressed c¢3 could not be resolved. However, in some in- stances strong expression of as could be seen to colocalize with neurofilament staining (Figures 6A and 6B, arrows). Strong TSP staining was found in the outer sheath as well and colocalized with the axons expresssing a3 (Figure 6C). Thus, the in vitro and in vivo localization patterns of a3~1 are consistent with its playing a role in axon outgrowth.

~131 Binding Assay To demonstrate a direct ligand-receptor interaction between the integrin heterodimer c¢3~1 and TSP1, a binding assay was developed using immunopurified a~6~ and purified TSP1. Because of the difficulty of getting a sufficient amount of integrin a~6~ from sympathetic neurons, RuGli cells were used as a source of rat ~zsl~l integrin. RuGli cells were surface biotinylated and then extracted in octylgluco- side-containing buffer. The as~ integrin was immunopuri- fled from this extract with an antibody to the a~ cytoplasmic domain coupled to protein A-Sepharose. The integrin was recovered under nondenaturing conditions by elution with the synthetic ct~ cytoplasmic domain peptide. As shown in Figure 7, the purified receptor contained two major biotinylated proteins of 130 and 112 kDa (lane 1) that, when reduced, migrate at 116 and 121 kDa, respectively (data not shown). These same subunits were immunoprecipitated from the purified receptor with an antibody to the ~ cytoplasmic domain (lane 2). An antibody to the (~s subunit, another possible TSP1 receptor subunit (Yabkowitz et al., 1993), did not immunoprecipitate any bands from the immunopurified a3131 preparation (lane 4), but did immunoprecipitate a5131 from a RuGli cell extract (lane 5).

Neuron 338

Figure 6. Localization of (z3 and TSP1 in the SCG Cryostat sections of a neonatal SCG were immunostained with anti-neurofilament antibody (A), anti-a3 MAb Ralph 3-2 (B), polyclonal anti-TSP1 antibody (C), or control mouse IgG (D). (A) and (B) are pictures of the same field that has been double labeled. (C) and (D) are from nearby sections. Neurofilament and TSP1 were visualized by Texas red immunofluorescence. Anti-(z3 MAb Ralph 3-2 and mouse IgG were visualized by FITC immunofluorescence. Bar, 50 ~m.

Purified (~3J~1 was tested for binding to substrates coated with ECM proteins in conditions similar to those used in neurite outgrowth assays. In the presence of 1 mM Ca 2. and 1 mM Mg 2., the immunopurified a31~ was found to bind to its reported ligand, LN-1 (Gehlsen et al., 1988, 1989; Elices et al., 1991), but not to vitronectin or ovalbumin (Figure 8A). Figure 8A also shows that a3~ bound TSP1 with approximately the same efficiency as LN-I. Similar binding can be demonstrated with nonbiotinylated (~31~ detected with an antibody to the ~ cytoplasmic domain, although with a much reduced signal to noise ratio (data not shown). The comparative efficacy of binding to TSP1 and LN-1 was found to vary with the batch of TSP1 (166%-72% range). The binding efficiencies of different batches of TSP1 correlated with the ability to support neurite outgrowth and probably reflect the lability of an essen- tial binding site in this protein.

Integrin interactions with ligands typically depend on divalent cations (Elices et al., 1991). As shown in Figure 8B, binding of integrin (x31~ to TSP1 or LN-1 occurred in 1 mM Ca 2+, 1 mM Mg 2., or 1 mM Mn 2+, but not in the presence of EDTA. The dependence on Ca 2+ showed substrate dependence, with binding to LN-1 being slightly greater at 0.1 mM than at 1 mM, while binding to TSP1 was much reduced at 0.1 mM Ca 2+.

As confirmation of the specificity of this interaction be-

tween O,3~ 1 and TSP1, both function-inhibiting anti-(~3 MAbs, Ralph 3-1 and Ralph 3-2, but not control antibodies, were found to decrease significantly the binding of the purified integrin to both LN-1 and TSP1 (Figure 8C). The anti-TSP1 MAb against the type 1 repeats, A 4.1 (Prater et al., 1991), also inhibited binding of purified integrin to TSP1 but not LN-1. This supports the cellular assays that localized the binding site to the region of the type 1 repeats. As further confirmation of specificity, the bound integrin was eluted off the substrate with SDS and EDTA after binding and analyzed by SDS-PAGE. The 130 kDa a3 band and the faint 112 kDa I~1 band were eluted from LN-1- and TSPl-coated substrates, but not from a bovine serum albumin (BSA)-coated substrate (see Figure 7B). a313~ binding to fibronectin, but not laminin and collagen, is reported to be inhibited by a peptide containing the RGD sequence of fibronectin (Elices et al., 1991). However, no RGD peptide sequence is present in the type 1 repeats of TSP1, and binding of ~sl~ to TSP1 was insensitive to the GRGDSP peptide at concentrations up to 1 mg/ml (1.7 raM; Figure 8C).

Discussion

Results in this paper show that the integrin a3131 is expressed on the axons and growth cones of sympathetic

Integrin ~zl3~ Is a Thrombospondin-1 Receptor 339

A B

o~3---~ ~

131->- . . . . . . . . . . . .

1 2 3 4

2 0 0

O ~

~ 1 1 6 O .......................

~ 9 7

5 z = . < . _ i t / ) ( / )

i - m

Figure 7. Immunopurified ~3P~ and Recovery from Coated Substrates (A) Gel analysis of immunopurified ~p,. Biotinylated ~z~pl was immunopurified from RuGli cells and analyzed by SDS-PAGE on a 6% gel under nonreducing conditions (lane 1). Arrows indicate the ~z~ and p, bands. This purified ~z3 extract was subjected to immunoprecipitation by the anti-(~ cytoplasmic domain antibody, either without (lane 2) or with (lane 3) competing I~, cytoplasmic domain peptide, or by an antibody to the ~5 cytoplasmic domain (lane 4). Lane 5 shows an immunoprecipitation from RuGli cells with the anti-as cytoplasmic domain antibody. (B) Recovery of ~p~ bound to substrates. ~31~1 bound to substrate in a binding assay was eluted with 10 mM EDTA and 1% SDS. Material eluted from LN-1 (LN), TSP1 (TSP), or bovine serum albumin (BSA) was analyzed by SDS-PAGE on a nonreducing 6% gel. Positions of Q3 and faint ~ bands are indicated by arrows.

neurons. On these cells it funct ions as a major receptor for TSP1 and recognizes a f ragment of TSP1 containing the properdin- l ike or type 1 repeats. The interact ions of Q3131 have been documented in cells using function- b locking MAbs to cz3 and 13~ and in receptor - l igand binding assays using the purif ied integrin.

Four integrin-subunit specif ic MAbs have been used to analyze the role of integrins in sympathet ic neuri te outgrowth on purif ied substrates of LN-1, Col IV, or TSPI . The impor tance of I~-class integrins was demonst ra ted with an anti-I~l MAb that inhibi ted outgrowth on all three of these substrates, having the greatest effect on Col IV ( - 9 2 % inhibition), a large effect on TSP1 ( - 7 3 % inhibition), and on ly 50% inhibit ion on LN-I . It has, however, been reported that an anti-I~l polyclonal serum inhibits 92% of neuri te outgrowth on LN-1 (Lein et al., 1991), which we conf i rmed with several anti-I~ polyc lonal ant ibodies (data not shown). Since the anti-I~ MAb comple te ly inhibits outgrowth on the TSP1 50/70 kDa f ragment and immunoprecipi tates the same four ~z bands as the anti-I~l cyto- p lasmic domain ant ibody (data not shown), this discrep- ancy between the MAb and polyclonal ant ibody effects on

A

120 100

• ~ 80 ~. 60

40 8 20

0 LN TSP VN OA

B150-I T ]

50

C E LCa Ca Mg Mn

= 12°/ I

,° . . . .

N 3-1 3-2 3-3 I T R c<3

Figure 8. Characterization of e3131 Binding (A) Qs131 was tested for binding to LN-1 (LN), TSP1 (TSP), vitronectin (VN), or ovalbumin (OA). Results from different experiments were normalized to binding on LN-1. (B) a3pl binding to LN-t (open bars) and TSP1 (closed bars) was done in the presence of either 10 mM EDTA (E), 0.1 mM Ca =* (LCa), tmM Ca 2., 1 mM Mg 2+, or I mM Mn 2.. Data from different experiments were combined by normalizing the data on both substrates to binding in the presence of 1 mM Ca 2.. (C) Inhibition of ~z~ binding by MAbs and RGD peptide. Immunopuri- fled a31~l was preincubated for 1 hr with either no antibody (N), Ralph 3-1 (3-1), Ralph 3-2 (3-2), Ralph 3-3 (3-3), mouse IgG (I), or 1 mg/ml GRGDSP peptide (R) before addition to substrates coated with LN-1 (open bars) or TSP1 (closed bars). Some substrates were pretreated with anti-TSP MAb (T) for 1 hr before addition of ~z31~,. Results are normalized to binding in the absence of antibody (N). Results are mean ± SEM; asterisk, statistically significant to p < .001 ; n.d., not done.

outgrowth on LN-1 may reflect an inabil i ty of the MAb to inhibit all integrin heterodimers completely. Because Q3131 is a receptor for the 50/70 kDa f ragment and the anti-I~l MAb complete ly inhibits outgrowth on this f ragment, it is l ikely that the anti-I~l MAb complete ly inhibits integrin cz3131 function in our assays. Compar ison of the inhibit ion with the anti-I~ MAb on whole TSP1 ( - 7 3 % ) versus the 501 70 kDa f ragment (100%) also suggests that SCG neurons can interact with a site outs ide of the 50/70 kDa fragment, and that this interact ion is insensit ive to this MAb. Results with the m- and a3-specific MAbs show that a3131 is a TSP1 receptor and that, as reported by others (Lei n e t al., 1991), m13~ is a Col IV receptor on these neurons. However, the same exper iments also show that nei ther m13~ nor ~31~1 funct ions as a major funct ional receptor for LN-1 on these cells.

In the purif ied receptor b inding assay, we found that purif ied a313~ bound to both its previously descr ibed l igand,

Neuron 340

LN-1, and to TSPI. Binding to both substrates was dependent on the presence of divalent cations, with similar binding in the presence of Ca 2+, Mg 2÷, or Mn 2+. This binding was specific since binding was not seen to vitronectin and ovalbumin, two proteins that are not ligands for a31~1. More- over, binding was inhibited by the same anti-co3 MAbs that inhibited neurite outgrowth on TSP1 but not by a non- function-blocking anti-cz3 MAb. Binding was also inhibited by an anti-TSP1 MAb that recognizes the region containing the type 1 repeats, which have been shown to inhibit outgrowth of sympathetic neurons on TSP1 (Oster- hout et al., 1992). Thus, we demonstrate that cz31~ can bind directly to TSP1 in a manner similar to its binding in a cellular context. Because the anti-a3 antibodies inhibit binding to LN-1 and TSP1, binding to both substrates appears to be mediated by a single binding site or nearby binding sites on ~z31~. Binding was also found to be insensitive to a peptide containing the RGD sequence that has been shown by others to inhibit the binding of cz3~ to fibre- nectin (Elices et al., 1991). This is consistent with evidence presented here demonstrating that ~zs~ interacts with the type 1 repeats of TSP1, which lack an RGD sequence.

These results add TSP1 to the increasing list of known ligands for the integrin ~3~. To date, ~z~ has been described as a receptor for several laminin isoforms (EHS- laminin [LN-1], merosin [LN-2], and kalinin/epiligrin [LN-5/ 6]), collagen, fibronectin, entactin, bacteria-derived invasin, and possibly the integrin c¢2~ (Wayner and Carter, 1987; Elices et al., 1991; Tomaselli et al., 1993; Carter et al., 1991; Weitzman et al., 1993; Delwel et al., 1994; Ded- har et al., 1992; Isberg and Leong, 1990; Symington et al., 1993). Some biochemical evidence has also suggested that ~1~1 heterodimers may bind to each other in a homophilic manner (Sriramarao et al., 1993). In neurons that express ~ , however, ~1~1 does not appear to play a major role in neurite outgrowth on most of these substrates. Antibodies to ~1~ can inhibit the interaction of a neuronal cell line with LN-1, but only after ¢x~l~, another LN-1 receptor, has been blocked (Tomaselli et al., 1990). With sensory neurons, outgrowth on LN-1 is marginally affected by antibodies to ~ , although outgrowth on LN-2 (merosin) is inhibited by - 5 0 % (Tomaselli et al., 1993). Our data show that on sympathetic neurons ~1~ does not play an important role in mediating neurite outgrowth on LN-1 or Col IV. In addition, we have found no effect of our antibodies on outgrowth on LN-5 (kalinin) or fibronectin (data not shown). The strong in hibitory effects of the anti-~3 antibodies on neurite outgrowth on TSP1 suggest that on neuronal cells the most important interaction of ~31~ may be with TSP isoforms rather than other ligands.

Since c¢~ is expressed in most cultured cell lines and in some primary cells, it is somewhat surprising that ~z~l~ has not been reported to be a TSP1 receptor. In fact, integrins ~z41~ and ~zsl~, but not ~ 1 , have been reported to be TSP1 receptors in activated T cells (Yabkowitz et al., 1993), despite the expression of ~1~ in these cells and its function in T cell attachment to epiligrin (Wayner et al., 1993). This suggests that the functional specificity of ct~l~ may be cell type dependent. Precedent for such cell type dependence has been found with the integrin ~2~1, which

functions as a collagen receptor in some cells and as a collagen and laminin receptor in others (Elices and Hemler, 1989; Kirchhofer et al., 1990; Chan and Hemler, 1993). This may also explain why neuronal c¢31~1 does not appear to play a major role in interactions with previously identified a3~1 ligands found with nonneuronal cells. Cell type differences might include alternative splice forms, alternative posttranslational modifications, or differences in membrane or cytoskeletal composition. Alternatively, the anti-~3 MAbs used in other studies may not inhibit binding to TSPI. Although the two function-blocking antibodies found here inhibit binding of a3131 to both TSP1 and LN-1, we have found a MAb that inhibits binding of a31~1 to LN-1 but not TSP1 (unpublished data). A third possibility is that the primary function of ~z~6~ is not in cell attachment, which most studies measure, but in other adhesive interactions such as neurite outgrowth or cell spreading.

Integrin a313~ recognizes multiple isoforms of its ECM ligands. For example, a313~ appears to recognize Col I as well as Col IV and Col VI (Wayner and Carter, 1987; Gehl- sen et al., 1989; Elices et al., 1991) and also appears to recognize LN-1, LN-2 (merosin), and LN-5/6 (kalinin/ epiligrin) (Wayner and Carter, 1987; Tomaselli et al., 1993; Carter et al., 1991; Weitzman et al., 1993; Delwel et al., 1994). This suggests the possibility that a31~ may be a receptor for other isoforms of TSP. The domains of TSP1 found in the 50•70 kDa fragment, which we have identified as containing an ~3131-binding site, are present in TSP2 and show an - 60% sequence identity with those in TSPI. This region is not present in TSP3, TSP4, or cartilage oligomeric protein, but is present in the TSP-like protein F-spondin, which is expressed in the ventral spinal cord (Klar et al., 1992). Thus, we hypothesize that TSP2 and possibly F-spondin are candidates as additional ligands for ~31~.

The expression of ~ deduced from immunofluorescent staining supports the hypothesis that this ligand-receptor interaction may play an important role in neural development. cz3 is expressed on the surface of growth cones and their filopodial extensions and is found on sympathetic axons in vivo. Immunofluorescent staining with a polyclonal antibody raised against TSP1 reveals that TSP is concentrated in the regions through which sympathetic axons expressing a3 travel. TSP1 is also expressed by endothelial cells of blood vessels, a target of sympathetic axons. Thus, the expression of these molecules is appropriate for roles in both axon extension and target interactions. Interestingly, although very little expression of c~3 is seen in neuronal tissue of the neonatal rat brain, intense staining was found in the external granule cell layer (data not shown). Since antibodies to TSP isoforms inhibit granule cell migration (O'Shea et al., 1990), cz31~ is a candidate to function as a TSP receptor in those neurons. Because TSP1 is not expressed in the cerebellum, this would re- quire that ~z31~ recognize another TSP isoform, possibly TSP2, which is appropriately expressed (Iruela-Arispe et al., 1993). Thus, 0f,3~1 could serve as a receptor on other types of neurons for TSP isoforms and play a role in other developmental events for which TSP isoforms are thought to be important. The importance, in vivo, of c¢3~ in mediat-

Intsgrin Q313~ Is a Thrombospondin-1 Receptor 341

ing in te rac t ions o f n e u r o n s wi th TSP1 and o t h e r TSP iso-

f o rms requ i res f u r the r s tudy.

Experimental Procedures

Mutedaio BSA was RIA grade fraction V from Sigma Chemical Company (St. Louis, MO). Synthetic peptides of the integrin subunit cytoplasmic domains of chick ¢¢3 (CYEAKGQKAEMRIQPSETERLIDDY), human p, (CDTGENPIYKSAVTTVVNPKYEGK), and human (z5 (CYGTAMEKAQ- LKPPATSDA) were synthesized by Dr. Chris Turck at the Howard Hughes Medical Institute, University of California, San Francisco. The GRGDSP peptide was purchased from GIBCO-BRL (Grand Island, NY). Hank's balanced saline solution (BSS), phosphate-buffered saline (PBS), Ca 2+- and Mg2+-frse PBS (CMF-PBS), fetal bovine serum (FBS), penicillin/streptomycin, L-15 medium, and insulin were from the UCSF Cell Culture Facility. DME and RPM I media were purchased from Medi- atech (Hemdon, VA). Nutddoma-SP was purchased from Boehringer Mannheim (Indianapolis, IN). Nerve growth factor (NGF) was prepared from rat submaxillary glands purchased from Pal Freez Siologicals (Rogers, AR) as described by Mobiey et al. (1976). A protease inhibitor cocktail, CLAP, was prepared as a 1000x stock with chymostatin, leupeptin, antipein, and pepstatin, each at 7 mg/ml in dimethylsulfox- ide. Except where indicated, chemicals were purchased from Sigma.

Sub . ra tes Human plateiet TSP was purified as described by Dixit et al. (1984). The 50/70 kDa fragment of human platelet TSP was generated by limited digestion of TSP with chymotrypsin as described by Prater et al. (1991), with the following modifications. The chymotryptic digest of purified platalet TSP was diluted 1:4 to lower the salt concentration to <50 mM. The digest was then loaded on a hepadn-Sepharose column equilibrated with 20 mM Tris-CI (pH 7.4). The 50/70 kDa fragment was eluted with a salt gradient at -0 .1 M NaCI. Gel analysis of this fragment by SDS-PAGE showed no contaminating bands. LN-1 (¢¢lP~y~) was prepared from the Engelbreth Holm Swarm tumor as de- sCribed by Timpl et al. (1979). Human Col IV was purchased from Fluka Biochemika (Buchs, Switzerland). Bovine vitronectin was purchased from Calbiochem (San Diego, CA). An extracellular recombinant fragment of chicken N-cadherin expressed in COS cells will be described elsewhere (R. M. K., unpublished data). For neurite outgrowth assays and receptor binding assays, substrates were coated on 96-well Lin- bro Titertek dishes from Flow Laboratories (McLean, VA). TSP and N-cadherin were diluted in Hank's BSS for coating. All other substrates were diluted in CMF-PBS. Substrate (75 pl per well) was incubated on the 96-wall dishes overnight at 4°C, then blocked with 1% BSA in PBS for 1 hr at room temperature.

Cell Culture RuGli cells (from Dr. Randy Kramer, UCSF) were grown in DME with 10% FBS and 100 Ulml penicillin/streptomycin. Cells were passaged with CMF-PBS containing 10 mM EDTA.

SCGs were dissected from Sprague-Dawiey neonatal (P0) rat pups purchased from Bantin and Kingman (Fremont, CA). SCGs were dissociated by manually removing the outer sheath and digested for 1 hr in 1 mg/ml cellagenase/dispase (UCSF Cell Culture Facility) in CMF- PBS. Digested SCGs were washed in CMF-PBS and triturated to dis- socials the cells. After two washes in in CMF-PSS, the cells were resuspended in serum-free 1_-15 complete medium as described by Hawrot and Patterson (1979), except that the N1 serum-free additives of Bottanstein et al. (1980) were substituted for rat serum. NGF was added to a final concentration of 100 ng/ml, glutamine to a concentration of 2 mM, and BSA to a final concentration of 1 mg/ml. Cells were grown in a 5% CO2 atmosphere at 37°C.

Hybridoma cells were grown in DME supplemented with 10O/o FBS and 100 Ulml penicillin/streptomycin. For IgG purification from supernatants, hybridomas were grown in serum-free medium composed of 1:1 DME:RPMI with 1% Nutddoma-SP and 100 Ulml penicillin/streptomycin.

Antibodies The anti-~ cytoplasmic domain serum, anti-Q~ cytoplasmic domain serum, and anti-ct~ cytoplasmic domain serum have been described

previously (Tomasalli et al., 1988; de Curtis et al., 1991). These anti- sara were affinity purified on cytoplasmic peptides coupled to thiopro- pyI-Sepharose (Pharrnacia, Piscataway, NJ) via their amino-terminal cysteine residue. Antibodies were eluted with 0.1 M glycine-CI (pH 2.3) and dialyzed against CMF-PBS. The hamster anti-I~l MAb, Ha2/ 11, has been described previously (Mendrick and Kelly, 1993) and was purified from superoatants of cells cultured in protein-free medium (Ultradoma PF, Whittaker Bioproducts, Inc., Walken/ille, MD) by am- monium sulfate precipitation and dialysis against PBS. The anti-TSP1 MAb, A 4.1, has been described previously (Prater et al., 1991) and was purified from hybridoma supernatant on a column with goat anti- mouse IgM (Zymed, South San Francisco, CA) covalently coupled to protein A-Sepharose (Pharmacia). The anti-TSP1 polyclonal antibody used for immunofluorescent staining, named Carol, was raised against platelet TSP, IgG purified on protein A-sepharose, and adsorbed against flbdnogen. It is specific for TSP1 by immunobiot and enzyme- linked immunosorbent assay (ELISA). IgG of the anti-a1 MAb, 3A3 (Turner st al., 1989), was purified from escites on MAC protein G discs (Amicon, Beverly, MA). Anti-bovine 150 kDa neurofilament subunit antibodies were purchased from Cbemicon International (Temeoula, CA).

Generation of Function-Blocking Anti-a3 MAbs Immunopurified ¢¢~81 (3.5 p.g; prepared as described below, but without biotinylation) was mixed with RIBI adjuvant (RIBI Immunochemical Research, Hamilton, MT) and injected into the hind footpads of BALB/c mice (Simonsen Laboratories, Gilroy, CA) previously anssthe- tizod with avertin. The mice were boosted every 4-6 days, twice with 10 p.g of immunopurified e31~1 in adjuvant, and once with 5 × 10 s live RuGli cells. The mice were boosted again with immunopurified (xs~, in adjuvant, 4 days before fusion.

The monoclonal fusion was done according to standard procedures (Hadow and Lane, 1988). Briefly, lymph nodes were removed, mechan- ically dissociated, and washed with medium. They were then fused with an equal number of SP2/0 myeloma cells with 0.5 ml of polyethyl- ene glycol and 10 pl of 2.25 M CaCI=. After washing with medium, cells were plated at limiting dilution in wells prsplated with 0.5 x 10 s to 3 x 105 peritoneal washout feeder cells per milliter (prepared as described in Hariow and Lane, 1988).

Resulting hybridoma supernatants were first screened by ELISA on live RuGli cells. Positive clones were grown in 1% Nutridoma and screened by immunoprecipitation of biotinylated RuGli cell extracts and SDS-PAGE. A total of 32 MAbs were obtained that immunoprecipitated only the eal~ heterodimer. These were subcloned and grown in 1% Nutddoma. IgG was purified from Nutridoma supernatants on a protein G-Sepharose 4 Fast Flow (Pharmacia) column and tested in function-blocking assays. Antibodies were tested at 50 p.g/ml, and 2 MAbs (Ralph 3-1 and Ralph 3-2) were found to inhibit neurite outgrowth on TSPI.

Neurlte Outgrowth Assays Dissociated sympathetic neurons, suspended in serum-free L-15 complete medium, were diluted to 2000-6000 cells/ml (see Celt Culture). Antibodies were diluted into serum-free medium at twice the desired final concentration. Each of the MAbs was normally used at a 50 p~g/ ml final concentration, except for Ha2/11, which was used at 100 I~g/ ml. An equal volume of cell suspension was added to the antibody aliquots and allowed to absorb to the ceils for 30 rain at room temperature. After preadsorption, the cells were resuspended, and 100 p.I was centrifuged onto each well of a substrata-coated 96-well dish at 500 x g for 10 min and grown in a 5O/o CO2 atmosphere for 15-18 hr. Cells were fixed by adding 200 p.I per well of 10% formalin, 50/o sucrose in PBS.

For quantitation of neudte outgrowth, the number of neurites with lengths greater than or equal to five cell body diameters in a unit area was counted. Only neurites originating within a unit area were counted. Care was taken to count similar areas in different wells, since neurite outgrowth can vary within a well (e.g., center versus edge). Although quantitation of the number of neurites per unit area and the percentage of neurons with neurites gave similar results (see Figure 1), the former method was used for routine quantitation. Assays were done in duplicate at least three times, and results were normalized to a no-antibody control. Statistical analysis was done with StatView software (Abacus Concepts, Inc., Berkeley, CA) by ANOVA using Scheffe's F procedure.

Neuron 342

Blotlnyletlon of Cell Surface Proteins For biotinylation of RuGli cell surface proteins, cells were grown to confluency in 75 cm 2 tissue culture flasks. For biotinylation of sympathetic cultures, approximately ten dissociated SCGs were plated in a well of a 6-well Primaria dish (Becton Dickinson Labware, Franklin Lakes, N J) coated with 5 p.g/ml LN-1 and grown for 4 days in culture to allow the production of a dense axonal plexus. Biotinylation was performed by a modification of the procedure described by Miyake et al. (1991). Briefly, confluent 75 cm z tissue culture flasks of RuGli cells or 6-well dishes of dissociated SCG cultures were washed twice with warm PBS, after which 2 ml of PBS/0.1 M Na-HEPES (pH 8.0) was added to each flask or well with 80 p.I of 10 mg/ml sulfo-NHS-biotin (Pierce Chemical Company, Rockford, IL) in PBS. After rocking for 1 hr at room temperature, cells were washed twice with PBS, removed from the flask or well with CMF-PBS with 10 mM EDTA, washed again with CMF-PBS with 10 mM EDTA, and washed once with PBS to remove unreacted sulfo-NHS-biotin.

Immunopreclpitstlon from Biotinylated Extracts Biotinylated RuGli cells or sympathetic neurons were solubilized in buffer A (lO/o Triton X-100 in PBS with CLAP and 1 mM phsnylmethyl- sulfonyl fluoride [PMSF]) with 50 mM Tris-CI (pH 7.5). The extracts were spun at 10,000 x g for 15 min, and 10 p.g of antibody was coupled to 30 I~1 of protein A-Sepharose for each immunoprecipitation. Biotinylated extracts were added and incubated overnight at 4°C. The Sepharose was then washed five times with buffer A and once with PBS with 1 mM PMSF and CLAP. Sample buffer was added, and the samples were boiled for 5 min, then run on 6% SDS-PAGE gels according to standard protocols. The proteins were transferred to nitro- cellulose (Schleicher & Schuell, Keene, NH), which was subsequently blocked with 5°/0 Carnation nonfat dry milk, 30/0 BSA fraction V (Sigma), 0.1% Tween-20 in PBS for I hr, washed three times with PBS, and incubated with horseradish peroxidase-streptavidin (Zymed Labora- tories, Inc., South San Francisco, CA) in 0.1% Tween-20/PBS for 1 hr. The blot was washed once in Tris-buffered saline (TBS), twice in TBS with 1% Triton X-100, 1% SDS, 0.5% deoxycholate, and twice again in TBS for 5 min each. The blot was visualized with enhanced chemiluminescence reagent (Amersham, Buckinghamshire, England).

Immunocytochemlstry Dissociated neurons were grown on 16-well glass chamber slides (Nunc, Naperville, IL) coated with 1 mg/ml polyornithine in CMF-PBS for 30 rain, followed by 10 p.g/ml Col IV for 1 hr or overnight. Cells were cultured for 18-20 hr, fixed with 10% formalin, 0.1% Triton X-100 in PBS for 15 min, and stained according to standard procedures (Harlow and Lane, 1988). For staining whole ganglia, SCGs were dissected from neonatal rats, fixed in 4O/o paraformaldehyde for 1 hr, and equilibrated with 30% sucrose in PBS overnight at 4°C. SCGs were then frozen in liquid nitrogen and cryostat sectioned. Sections (10 p.m) were fixed with methanol at -200C for 10 min and stained according to standard procedures. Secondary antibodies were FITC-conjugated goat F(ab')2 anti-mouse IgG and Texas red-conjugated goat F(ab')2 anti-rabbit IgG (Jackson ImmunoResearch, West Grove, PA). Slides were mounted under coverslips with FITC-Guard (Tastog, Inc., Chi- cago, IL).

o4131 Immunopurlflcatlon To immunopurify as131, 10 mg of affinity-purified anti-a~ cytoplasmic domain antibody was cross-linked to 3 ml of protein A-Sepharosa (Pharmacia) with 20 mM dimethlypimelimidate (Pierce) in 0.2 M Na- borate (pH 9.0) as described by Harlow and Lane (1988) and blocked with 0.2 M ethanolamine (pH 8.0). Biotinylated or nonbiotinylated RuGli cells from four confluent flasks were extracted in buffer B (50 mM N-octyl-~-D-glucopyranoside, 50 mM Tris--CI [pH 7.5], 15 mM NaCI, 1 mM PMSF, CLAP) with 1 mM CaCI2 and 1 mM MgCI2, centrifuged at 10,000 x g for 15 min, and incubated with CL-4B Sepharose (Phar- macia). The supernatant was then incubated with anti-a~-Sepharosa for at least 4 hr. Following six washes in buffer B, a31~, was eluted for 1 hr with the aa, cytoplasmic domain peptide (1 mg/ml in buffer B).

Receptor Binding A m y Wells were coated and blocked as described above. Receptor purified as described above was diluted in buffer B with 1 mM CaCI2 and 1

mM MgCI2 at 1:10 to 1:50, respectively. The solutions were added to the substrata-blocked dishes and allowed to bind at room temperature for 2 hr. The wells were then washed five times with buffer C (25 mM N-octyl-I~-D-glucopyranoside, 50 mM Tris-CI [pH 7.5], 15 mM NaCI, 1 mM PMSF, CLAP) with 1 mM CaCI2 and 1 mM MgCI2 and incubated with streptavidin-conjugated horseradish peroxidase (Zymed) diluted in the same buffer, for I hr at 40C. After washing five times more with buffer C with 1 mM CaCI2 and 1 mM MgCI2, binding was assayed by developing with 3,3',5,5'-tetramethylbenzidine reagent (Kirkegaard and Perry, Gaithereburg, MD) and stopping the reaction with an equal volume of 1 M H3PO4. Absorbance at 450 nm was read in a microtiter plate reader (Flow Laboratories). Results were zeroed to wells with no receptor added. For experiments testing different divalent cations, 1 mM CaCI2 and 1 mM MgCI2 were replaced by the divalent cation tested. For experiments with antibodies added, receptor was used at a 1:50 dilution with antibody at 150 p.g/ml. The receptor and antibody were incubated at room temperature for 1 hr before the binding assay. Statistical analysis was as described for the neurite outgrowth assay.

Acknowledgments

The authors thank Dr. Randy Kramer for providing the RuGli cells and Dr. Ivan de Curtis and Ms. Kristine Venstrom for generating the antisera against the integrin cytoplasmic domains. We thank Dr. Paul Patterson and Dr. Gisela Weskamp for teaching the SCG dissection. We also thank Dr. Deborah Finlay for useful comments and discus- sions regarding the binding assay and Dr. Ulrich M011er for critical reading of the manuscript. M. F. D. was supported by U. S. Public Health Service grants T32 GM07449 and EY07120. This work was supported by USPHS grants NS19090 and NS-POl-16033 (L. F. R.), HLt4147 ON. A. F.), and DK38995 (D. L. M.). L. F. R. is an investigator of the Howard Hughes Medical Institute.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 USC Section 1734 solely to indicate this fact.

Received January 25, 1995; revised May 3, 1995.

References

Bornstein, P. (1992). The thrombospondins: structu re and regulation of expression. FASEB J. 6, 3290-3299.

Bornstein, P., O'Rourke, K., Wikstrom, K., Wolf, F. W., Katz, R., Li, P., and Dixit, V. M. (1991). A second, expressed thrombospondin gene (Thbs2) exists in the mouse genome. J. Biol. Chem. 266,12821-12824. Bottenstein, J. E., Skaper, S. D., Varon, S. S., and Sato, G. H. (1980). Selective survival of neurons from chick embryo sensory ganglionic dissociates utilizing serum-free supplemented medium. Exp. Cell Res. 125, 183-190.

Carter, W. G., Ryan, M C., and Gahr, P. J. (1991). Epiligrin, a new cell adhesion ligand for integrin a3~1 in epithelial basement mem- branes. Cell 65, 599-610.

Chan, B. M, and Hamlet, M. E. (1993). Multiple functional forms of the intogrin VLA-2 can be derived from a single a = cDNAclone: inter- conversion of forms induced by an anti-~l antibody. J. Cell Biol. 120, 537-543.

de Curtis, I., Quaranta, V., Tamura, R. N., and Reichardt, L. F. (1991). Laminin receptors in the retina: sequence analysis of the chick integrin a8 subunit. Evidence for posttranslational regulation. J. Cell Biol. 113, 405-416.

Dedhar, S., Jewell, K., Rogiani, M., and Gray, V. (1992). The receptor for the basement membrane glycoprotein entactin is the integrin a3~1. J. Biol. Chem. 257, 18908-18914.

Delwel, G. O., de Melker, A. A., Hogervorst, F., Jaspars, L. H., Fles, D. L. A., Kuikman, I., Lindblom, A., Paulsson, M., Timpl, R., and Son- nenberg, A. (1994). Distinct and overlapping ligand specificities of the a3~1 and a61~1 integrins: recognition of laminin isoforms. Mol. Biol. Cetl 5, 203-215.

Dixit, V. M., Grant, G. A., Santoro, S. A., and Frazier, W. A. (1984). Isolation and characterization of a heparin-binding domain from the

Integrin a3~1 Is a Thrombospondin-1 Receptor 343

amino terminus of platelet thrombospondin. J. Biol. Chem. 259, 10100-10105.

Elices, M. J., and Hemler, M. (1989). The human integrin VLA-2 is a collagen receptor on some cells and a collagen/laminin receptor on others. Prec. Natl. Acad. Sci. USA 86, 9906-9910.

Elices, M. J., Urry, L. A., and Hemler, M. E. (1991). Receptor functions for the integrin VLA-3: fibronectin, collagen and laminin binding are differentially influenced by ARG-GLY-ASP peptide and by divalent cations. J. Cell Biol. 112, 169-181.

Frazier, W. (1991). Thrombospondins. Curr. Opin. Cell Biol. 3, 792- 799.

Gehlsen, K. R., Dillner, L., Engvall, E., and Ruoslahti, E. (1988). The human laminin receptor is a member of the integrin family of cell adhesion receptors. Science 241, 1228-1229

Gehlsen, K. R., Dickerson, K., Argraves, W. S., Engvall, E., and Ruo- slahti, E. (1989). Subunit structure of a laminin-binding integrin and localization of its binding site on laminin. J. Biol. Chem. 264, 19034- 19038.

Harlow, E., and Lane, D. (1988). Antibodies: A Laboratory Manual (Cold Spring Harbor, New York: Cold Spring Harbor Laboratory).

Hawrot, E., and Patterson, P. H. (1979). Long-term culture of dissociated sympathetic neurons. Meth. Enzymol. 58, 574-584.

Iruela-Arispe, M. L., Liska, D. J., Sage, E. H., and Bornstein, P. (1993). Differential expression of thrombospondin 1, 2, and 3 during murine development. Dev. Dynam. 197, 40-58.

Isberg, R. R., and Lecng, J. M. (1990). Multiple ~ chain integrins are receptors for invasin, a protein that promotes bacterial penetration into mammalian cells. Cell 60, 861-871.

Kirchhofer, D., Languino, L. R., Ruohslahti, E., and Pierschbecher, M. D. (1990). ~z21~ integrins from different cell types show different binding speciflcities. J. Biol. Chem. 265, 615-618.

Klar, A., Baldassare, M., and Jessell, T. M. (1992). F-spondin: a gene expressed at high levels in the floor plate encodes a secreted protein that promotes neural cell adhesion and neurite extension. Cell 69, 95- 110.

Lawler, J., Weinstein, R., and Hynes, R. O. (1988). Cell attachment to thrombospondin: the role of ARG-GLY-ASP, calcium, and integrin receptors. J. Cell Biol. 107, 2351-2361.

Lawler, J., Duquette, M., Whittaker, C. A., Adams, J. C., McHenry, K., and DeSimone, D. W. (1993). Identification and characterization of thrombospondin-4, a new member of the thrombospondin gene family. J. Cell Biol. 120, 1059-1067.

Lein, P. J., Higgins, D., Turner, D. C., Flier, L. A., and Terranova, V. P. (1991). The NC1 domain of type IV collagen promotes axonal growth in sympathetic neurons through interaction with the a~l~, integrin. J. Cell Biol. 113, 417-428.

Mendrick, D. L., and Kelly, D. M. (1993). Temporal expression of VLA-2 and modulation of its ligand specificity by rat glomerular epithelial cells in vitro. Lab. Invest. 69, 690-702.

Miyake, K., Weissman, I. L., Greenberger, J. S., and Kincade, P. W. (1991). Evidence for a role of the integrin VLA-4 in lympho- hemopoiesis. J. Exp. Med. 173, 599-607.

Mobley, W. C., Schenker, A, and Shooter, E. M. (1976). Characteriza- tion and isolation of proteolytically modified nerve growth factor. Bio- chemistry 15, 5543-5552. Neugebauer, K. M., Emmett, C. J., Venstrom, K. A., and Reichardt, L. F. (1991). Vitronectin and thrombospondin promote retinal neurite outgrowth: developmental regulation and rote of integrins. Neuron 6, 345-358.

Oldberg, A., Antonsson, P., Lindblom, K., and Heingard, D. (1992). COMP (cartilage oligomeric matrix protein) is structurally related to the thrombospondins. J. Biol. Chem. 267, 22346-22350. O'Shea, K. S., and Dixit, V. M. (1988). Unique distribution of the extracellular matrix component thrombospondin in the developing mouse embryo. J. Cell Biol. 107, 2737-2748. O'Shea, K. S., Rheinheimer, J. S., and Dixit, V. M. (1990). Deposition and role of thrombospondin in the histogenesis of the cerebellar cortex. J. Cell Biol. 110, 1275-1283.

O'Shea, K. S., Liu, L.-H. J., and Dixit, V. M. (1991). Thrombospondin and a 140 kd fragment promote adhesion and neurite outgrowth from embryonic central and peripheral neurons and from PC12 cells. Neu- ron 7, 231-237.

Osterhout, D. J., Frazier, W. A., and Higgins, D. (1992). Thrombospon- din promotes process outgrowth in neurons from the peripheral and central nervous systems. Dev. Biol. 150, 256-265.

Prater, C. A., Plotkin, J., Jaye, D., and Frazier, W. A. (1991). The properdin-like type I repeats of human thrombospondin contain a cell attachment site. J. Cell Biol. 112, 1031-1040.

Reichardt, L. F., and Tomaselli, K. T. (1991). Extracellular matrix molecules and their receptors: functions in neural development. Annu. Rev. Neurosci. 14, 531-570.

Sriramarao, R., Steffnert, P., and Gehlsen, K. R. (1993). Biochemical evidence for a homophilic interaction of the a3~1 integrin. J. Biol. Chem. 268, 22036-22041.

Symington, B. E., Takada, Y., and Carter, W. G. (1993). Interaction of integrins a31~1 and a21~,: potential role in keratinocyte intercellular adhesion. J. Cell Biol. 120, 525-535.

Tamura, R. N., Cooper, H. M., Collo, G., and Quaranta, V. (1991). Cell type-specific integrin variants with alternative a chain cytoplasmic domains. Prec. Natl. Acad. Sci. USA 88, 10183-10187.

Timpl, R., Rohde, H., Robey, P. G., Rennard, S. I., Foidart, J.-M., and Martin, G. R. (1979). Laminin--a glycoprotein from basement mem- branes. J. Biol. Chem. 254, 9933-9937.

Tomaselli, K. J., Damsky, C. H., and Reichardt, L. F. (1988). Purifica- tion and characterization of mammalian integrins expressed by a rat neuronal cell line (PC12): evidence that they function as oJ~ heterodim- eric receptors for collagen IV and laminin. J. Cell Biol. 107, 1241- 1252. Tomaselli, K. J., Hall, D. E., Flier, L. A., Gehlsen, K. R., Turner, D. C., Carbonetto, S., and Reichardt, L. F. (1990). A neuronal cell line (PC12) expresses two 13~-class integrins-~l~l and asJ3,--that recognize different neurite outgrowth-promoting domains in laminin. Neu- ron 5, 651-662.

Tomaselli, K. J., Doherty, P., Emmett, C. J., Damsky, C. H., Walsh, F. S., and Reichardt, L. F. (1993). Expression of I~1 integrins in sensory neurons of the dorsal root ganglion and their functions in neurite outgrowth on two laminin isoforms. J. Neurosci. 13, 4880-4888

Turner, D. C., Flier, L. A., and Carbonetto, S. (1989) Identification of a cell surface protein involved in PC12 cell-substratum adhesion and neurite outgrowth on laminin and collagen. J. Neurosci. 9, 3287-3296.

Tuszynski, G. P., Karczewski, J., Smith, L., Murphy, A., Rothman, V. L., and Knudsen, K. A. (1989). The GPIIb-Illa-like complex may function as a human melanoma cell adhesion receptor for thrombospondin. Exp. Cell Res. 182, 473-481.

Vos, H. L., Devarayalu, S., de Vries, Y., and Bornstein, P. (1992). Thrombospondin 3 (Thbs3), a new member of the thrombospondin gene family. J. Biol. Chem. 267, 12192-12196.

Wayner, E. A., and Carter, W. G. (1987). Identification of multiple cell adhesion receptors for collagen and fibronectin in human fibrosarcoma cells possessing unique ~z and common ~ subunits. J. Cell. Biol. 105, 1873-1884.

Wayner, E. A., Gil, S. G., Murphy, G. F., Wilke, M. S., and Carter, W. G. (1993). Epiligrin, a component of epithelial basement mem- branes, is an adhesive ligand for a~-posi t ive T lymphocytes. J. Cell Biol. 121, 1141-1152.

Weitzman, J. B., Pasqualin, R., Takada, Y., and Hemler, M. E. (1993). The function and distinctive regulation of the integrin VLA-3 in cell adhesion, spreading, and homotypic cell aggregation. J. Biol. Chem. 268, 8651-8657.

Yabkowitz, R., Dixit, V. M., Guo, N., Roberts, D. D., and Shimizu, Y. (1993). Activated T-cell adhesion to thrombospondin is mediated by the e4~1 (VLA-4) and ~x5[~1 (VLA-5) integrins. J. Immunol. 151,149- 158.

identification of integrin (13131 as a neuronal...

Documents