Studies on Extracellular Matrix Components That Promote Neurite Outgrowth

An important determinant in the development of mul ticellular organisms is the extracellular matrix (ECM) upon which cells attach, migrate, and differentiate. It is likely a class of substances that affect neuronal de velopment will be found to be associated with the ECM. During early stages of neuronal development, strong spatial and temporal correlations are be the appearance of fibronectin and migration of granule cells to form the external granule cell layer in the cerebellum and migration of neural crest cells to form the variety of tissues derived from the crest When cerebellar granule cells neural crest cells are cultured in vitro, their abilities to adhere to and migrate on fibroncctin substrata correlate temporally with their migratory be haviors in

likely that a class of substances that affect neuronal development will be found to be associated with the ECM . During early stages of neuronal development, strong spatial and temporal correlations are seen between the appearance of fibronectin and migration of granule cells to form the external granule cell layer in the cerebellum (Hatten e t al. 1982) and migration of neural c rest cells to form the variety of tissues derived from the crest (Thiery et al. 1982). When cerebellar granule cells or neural c rest cells are cultured in vitro, their abilities to adhere to and migrate on fibroncc tin substrata correlate temporally with their migratory behaviors in vivo (Hatten et al. 1982;Rovasio e t al. 1983).
The ECM is also important in later stages of neuronal development. The interaction between a g rowth cone and its substratum can determine the rate at which it grows a nd the paths it follows in vitro (Letourneau 1975). In vivo, axons may also follow routes determined by the substratum (Katz and Lasek 1979). In some cases , this reflects the association of axo ns with already oriented cells, e.g., radial glia in the cerebellum (Rakic 1974) and pioneer fibers in Daphnia (Levinthal et al. 1976).
There are several facto rs that, when bound to culture substrata, stimulate outgrowth from particular classes of neurons. Of these, fibronectin and laminin (Akers et al. 1981 ; Baron-Van Evercooren et al. 1982) are known components of the ECM in vivo. One group of factors is derived from cultured cells (Collins 1978;Adler et al. 1981 ;Coughlin et al. 1981 ;Lander et al. 1982), and, when attac hed to a substratum, these factors promote profuse and rapid neurite outgrowth . Whe n tested o n sympathetic neurons , neurite outgrowth is seen even in the absence of nerve growth factor (NGF). The properties of these factors are discussed in this paper.

Extracellular Matrix Promotes Neurite Outgrowth
Our attention was originally drawn to the ECM by observations that PC12 (pheochromocytoma) cells, rat sympathetic neurons, and other neuronal cell types ex-secreted by corneal endothelial cells in vitro (Fujii et al. 1982;Lander ct al. 1982) . PC12 cells and rat sympathetic neurons exte nd neurites on the ECM even in the absence of NGF, and process outgrowth is no t prevented by preincubation of ECM with antiserum to NGF or inclusion of NGF antiserum in the culture medium.
The behavior of neurons o n ECM was unusual in other ways (Table l). Neurites appeared earlie r and grew more rapidly on ECM than o n polylysine-coated plastic. Within 6 hours after plating onto ECM , more than 80% of the neurons had neurites, many of them several cell diameters in length. In contrast, neurons plated o nto polylysine and cultured with NGF had very few processes by 6 hours or even 12 hours , although most cells extended neurites by 24 hours .
Although neurites appeared early and grew rapidly o n ECM in the absence of NGF, their rate of growth slo wed dramatically after 24 hours and cell viability (estimated by morphological criteria) fell steadily thereafter. By 72 hours, less than 20% of the neurons appeared alive; none survived over 5 da ys. Massive cell death could be avoided only if NGF was present in the culture medium. T hen, good viability was maintained for over a week, the longest time assayed. Thus, ECM can substitute for NGF in inducing short-term process outgrowth, but not in maintaining long-term neuronal viability . Rat sympathetic neurons cultured with NGF on polylysine-coated tissue culture plastic and without NGF on ECM were fixed at various times after plating. Neurite outgrowth was measured by counting random fields and detennining the percentage of presumptive neurons wi1h neurites. Survival was estimated crudely as the percentage of presumptive neurons lacking morphological signs of cell death or in· jury . includ ing cell swelling, loss of adhesiveness. retraction of neurites. and accumulation of intracytoplasmic granules. Over 100 hibit profuse and rapid process outgrowth on the matrix neurons were counted for each point shown above. Rat sympathetic neurons were plated on polylysine-coared tissue culture plastic that had been treated with CMsF· Neurons were cultured without NGF and with or without serum. Neurite outgrowth and survival were detcnnined 85 in Table I.

Identification of Active Factors
Associated with the ECM Two approaches have been used to characterize factors in the ECM that promote neurite outgrowth. Major purified components, such as fibronectin and laminin, have been directly tested for effects on neurite outgrowth, and factors secreted by cells into serum-free conditioned media have been identified, purified, and characterized.

Polycationic Surfaces Treated with Conditioned
Medium Can Substitute for ECM One class of neurite outgrowth-promoting factors can be identified in medium conditioned by growth with corneal endothelial cells. Polylysine-coated tissue culture dishes were exposed to this conditioned medium (CM) and washed thoroughly. Sympathetic neurons plated onto this substratum responded as they did on ECM, by rapidly extending neurites. Serum-free conditioned medium (CMsF) applied in this way also produced an active substratum (Table 2). Therefore, the active substance(s) is synthesized and secreted by corneal endothelial cells and is not a modified or concentrated component of serum. Table 2 also shows the results obtained when neurons plated on CMsF-treated polylysine surfaces were cultured without serum. Since rapid neurite outgrowth was also observed under these conditions, the CMsi:-eoated surface was not acting merely by adsorbing and concentrating some serum component onto the substratum. Instead, it appears to act directly on sympathetic neurons. Figure l shows the appearance of cells on CMsr-treated polylysine-coated surfaces when grown in the presence and absence of serum. Surfaces not treated with polylysine did not promote measurable neurite outgrowth after incubation with CMsF· CMsrtreated Substrata Promote Neurite Outgrowth by Many Different Types of Neurons In addition to rat sympathetic neurons, other neuronal cultures were plated onto CMsr-treated, polyly-sine-coated substrata. Rat and chick sensory neurons and sympathetic neurons all responded in the characteristic manner already described, i.e., with early, rapid, extensive neurite outgrowth (Fig. 2). NGF, which these cells normally require for outgrowth and survival, was not required for the response. Thus, two classes of peripheral NGF-dependent neurons from two different species respond to the factor contained in CMsF· Othe~ have reported that parasympathetic neurons also respond to similar factors (Adler et al. 198 l). At least one class of central neurons responds to CMsf'9treated substrata. Embryonic chick motor neurons were purified in a fluorescence-activated cell sorter from embryos whose limbs had been injected previously with a purified conjugate of Lucifer Yellowwheat germ agglutinin (modified from the procedure of McPheeters and Okun 1980). This conjugate is transported retrogradely and labels the motor neurons but no other cell types in the spinal cord. When these fluorescent neurons were purified on the cell sorter and plated, neurite outgrowth in culture was observed in the first 24 hours on· CMsf'9treated, but not untreated, polylysine-coated substrata (not shown). Short-term survival did not require other factors, but long-term survival was promoted by addition of chick muscle CM. In contrast, when dissociated cells from cerebellum and olfactory bulb of neonatal rats were plated on CMsf'9 treated substrata, no positive or negative effects on growth were observed (not shown). When examining cultures not requiring NGF, such as these, negative results may be equivocal, since neurite outgrowth may normally be quite.rapid and growth-promoting factors may effect only a minimal increase. With this in mind, however, it is interesting to note that of the cell types tested, all of those with peripheral axons in vivo responded to CMsF; those with processes only in the CNS were not noticeably affected.

Many Cell Types Produce Factors with Similar Properties
Serum-free media conditioned by confluent cultures of various cell types were prepared as described (Lander et al . 1982) and assayed for neurite outgrowth-  Figure' 3. Sepharose 6B chromatography of BCE CMsF· BCE CMsF was fractionated by gel filtration, as described.
Fractions were assayed for neurite outgrowth-promoting activity.
promoting act1v1ty in the standard manner, using rat sympathetic neurons. Media conditioned by bovine vascular endothelial cells, bovine vascular smooth muscle cells, bovine adrenal · cortical cells, human skin fibroblasts, embryonic chick myotubes, rat primary cells that appear to be derived from pericytes (N. Dekker et al ., unpubl.), and the cell lines C2 (mouse skeletal muscle), PTK-1 (kangaroo rat epithelium), A-431 (human vulva carcinoma), RN-22 (rat Schwannoma~, N-18 (mouse neuroblastoma), and PC12 (rat pheochromocytoma) all possessed activity indistinguishable from that of CMsF. and these factors appear to have similar biochemical properties. Eirst, the factors that have been 60 so 40 ..

•.
.. ·· . .. 30 20 10 0 examined appear to be large molecules. As exemplified in Figure 3, the neurite outgrowth-promoting factors that have been examined eluted near the void volume in penneation chromatography on Sepharose 6B (exclusion limit 4 x 106 daltons for globular proteins; l x 106 daltons for polysaccharides) . "' CMsF from several sources has -also been fractionated by isopycnic sedimentation in CsCl density gradients under nondissociating conditions. The results in Figure 4 show that the factors from a variety of sources have densities in CsCl between l.30 and l.40. The activities in CM from bovine corneal endothelial (BCE), chick mesenchymal , and putative pericyte cultures ap-

1.20
GRADIENT POSITION Figure 4 . lsopycnic sedimentation in associative CsCI gradients. BCE CMSF from different cell types was centrifuged to equi· librium in CsCI containing 0.4 M GuHCI, as described in Lander et al. (1982). Fractions were collected and the density of each was measured ( · · · · -). After dialysis, fractions were assayed for neurite outgrowth-promoting activity. Neu rite outgrowth was pear to be single peaks of activity with average densities of 1.35. The activities in CM grown with RN-22 and PC 12 cells appear to be bimodal with density peaks at 1.37 and 1.32. The presence of two peaks may reflect molecular heterogeneity. Alternatively, this appearance may be an artifact: If a molecule roughly cosedimenting with the factor also bound polylysine, it might outcompete the factor for binding to the substratum, thus artificially depressing the level of activity observed in certain fractions. It is unclear at this point which alternative is correct. Since these activities have densities between those of pure proteins and carbohydrates, it appeared that these factors might contain carbohydrate, as well as protein.
To obtain more detailed information, corneal endothelial .CMsF. metabolically labeled with [3H]amino acids and [3'SJsulfate, was prepared and centrifuged in CsCI. With snme qualifications (Branford-White 1980), the sulfate label is specific for glycosaminoglycans and, therefore, marks the position of proteoglycans in these gradients . The leucine label marks proteins and, therefore, also glycoproteins and proteoglycans. When concentrated CMsF, to which aliquots of leucine-labeled and sulfate-labeled CMsF were added, was centrifuged in CsCl, the neurite outgrowth-promoting activity banded with a peak of [3'SJsulfate and [3H]amino acidlabeled material , suggesting that the neurite outgrowthpromoting factor is or is associated with a sulfated gly-  To help distinguish between these possibilities, CM was prepared in the presence of drugs that interfered selectively with the synthesis of ind ividual classes of carbohydrate. The results in Figure 5 show that neurite outgrowth-promoting activity could be detected in CMsF prepared by growth in the presence of 2.5 mM pnitrophenyl-tJ-o-xyloside, but the density of this activity was shifted to 1.30 g/mJ. Because tJ-o-xylosides compete with xylose-primed core proteins as templates for glycosaminoglycan synthesis, cells exposed to these drugs secrete proteoglycans that contain fewer and/or shorter glycosaminoglycan chains (Stevens and Austen 1982) and are therefore less dense. This result indicates that the biologically active factor includes a proteoglycan, since synthesis of mucins and N-linked carbohydrates is not sensitive to the presence of tJ-o-xylosides. xylosides.
To identify the properties of the molecules essential for activity of the proteoglycan-associated factors, CMsF from BCE cells and PC 12 cells and active fractions from CsCl gradients were subjected to various treatments-prior to assay . The corneal endothelial factor is inactivated by low or high pH or by heating to 80% (Lander et al. 1982). The results in Table 3 show that the activities of the corneal endothelial and PC 12-derived factors are destroyed by trypsin. No decrease in the activity of BCE CMsF was seen after  In ~. CMSF was exposed to various 1rea1ments and assayed. In b, 50-µrn aliquots of heparioase (6.0 mg/ml in 0. 1 sodium ;cetate [pH 7 .O}) were applied to columns containing 100 µI of glycosaminoglycan-«>njugatcd Sepharose and eluted, at 4 °C, in 200 µI of buffer. Controls were no< chromatographed, but were simply diluted with buffer to l .5 mg/ml and 0.15 mg/ml. Samples were added to 4 volumes of panially purified factor and the mixture was incubaled for 4 hr a1 30°C and assayed for neurite ou1growth·promoling aclivi1y . A dilution of the factor affording maximal sensitiviiy was used. Data are averages of duplicate assays± deviation from the mean. exposure to collagenase or neuraminidase. To identify the class of proteoglycan associated with neurite outgrowth-promoting activity, active fractions were pooled and digested with chondroitinase ABC or heparinase. The results in Table 3 show that the neurite outgrowth-promoting activity is not sensitive to chondroitinase ABC, which degrades the chond roitin and dermatan sulfates, but is sensitive to heparinase, which degrades the remaining class of xylose-initiated glycosaminoglycans, the heparan sulfates. If aliquots of the heparinase were first passed over glycosaminoglycan-Sepharosc columns, the eluate that was not retained on chondroitin sulfate-Sepharose retained ability to inactivate the factor, whereas the eluate that was not retained on heparan sulfate-Sepharose no longer possessed the ability to inactivate the neurite outgrowth-inducing factor. Therefore, the degradative enzyme binds heparan sulfate, but not chondroitin sulfate, and the factors ap-pear to be heparan sulfate-proteoglycans or heparan sulfate-proteoglycans complexed with other molecules.

Purification and Characterization of One Factor
The neurite outgrowth-promoting activity from corneal endothelial CMsF labeled with [3H]leucine and [3SS]sulfate has been purified by fractionation with ammonium sulfate, polyethylene glycol, DEAE-cellulose chromatography, and sucrose gradient velocity sedimentation. Results are summarized in Table 4 . When the entire peak of active material from the velocity sedimentation was pooled, the factor was found to be purified approximately 40-fold over 3 H-labeled material and 30-fold over 3 ss-labeled material. These numbers must be considered low estimates, since some of the flanking fractions pooled in with the peak clearly contained contaminating material. The sedimentation Pooled labeled and unlabeled C MsF was trealed with ammonium sulfa1e (48% satura1ion). and the precipitate was re· covered by centrifuga1ion . The pellets were redissolved in 50 mM Tris-HCI (pH 7.4), and polyethylene gly~I 6000 was added to a final ooncen1ra1ion of 14% (w/ v). The prccipi1a1e was recovered by centrifugation. and the pellet was resuspended in Tris-saline buffer (0. 1 M NaCl/50 mM Tris-HCI [pH 7 .4]). Ma1erial that failed to redissolve was removed by centrifugation. The solution was then mixed with 0 .5 ml of a slurry of DEAE-cellulose equilibrated in Tris-saline buffer and shaken overnight. After uni:>ound material was eluted, the DEAE-«llulose was washed with 4 volumes of Tris-saline buffer and eluted with I M NaCl/SO mM Tris-Cl (pH 7.4). The material thus eluted was mixed with 2 volumes of 50 mM Tris-Cl (pH 7 .4) and sedimented on a linear 5-20% sucrose gradien1 in Tris-saline buffer.
1 Neuri1e outgrowth-promoting ac1iviiy: One unit per milliliter gives a hair-maximal response. behavior of the factor is noteworthy: At very low concentration , most of the neurite outgrowth-promoting activity is found at about 14S-15S, with a smaller peak at about l 7S. At higher concentration, the factor sediments farther, with two peaks often seen at 17S and 19S. At the highest concentrations tested, a very broad peak at 21S was seen. In every case examined, regardless of where the peak of outgrowth-promoting activity appeared, the major peak(s) of fast-sedimenting 35 S and 3 H cosedimented with it. This unusual sedimentation behavior, probably reflecting aggregation, has therefore proved useful in firmly identifying the labeled peaks with the active factor.
Characterization of the Factor ~ Significant amounts of reasonably purified factor could be obtained quickly, taJcing advantage of its unusual sedimentation behavior, by centrifuging it to concentrate it, sedimenting it at very high concentration on a sucrose gradient (where it runs near 21S), pooling the active material, and resedimenting it on a sucrose grad-ient at low concentration. As shown in F igure 6a, material from the first sucrose gradient, rerun at low concentration, behaves as a peak at 14S with a small shoulder at 17S . Because peaks are not sharply separated on sucrose gradients, the pooled 21S material from the first gradient was contaminated with a small fraction of the peaks found in other regions; for this reason, a shoulder at 6S-7S is seen, representing residual contamination with the major high-molecular-weight protein secreted by the corneal endothelial cells. Little other contamination is readily apparent.
The results in Figure 6 (b-d) indicate the effects of digestion with various enzymes on the material shown in Figure 6a. Incubation with chondroitinase .ABC did not reduce measurably the size of the labeled material (Fig.  6c). In contrast, incubation with highly purified heparitinase (selective for low-sulfate-substituted heparan sulfate-proteoglycans) reduced the rate at which 3 H and 35 S sediment in sucrose (Fig. 6d). These results support previous evidence that a heparan sulfate-proteoglycan is an integral part of the neurite outgrowth factor. The size of the factor also appeared to be reduced by incubation with collagenase (Fig. 6b), but the possibility that the collagenase contained small amounts of other proteases has not been completely ruled out.
The samples analyzed by sedimentation (Fig. 6a-d) were also subjected to SDS-acrylamide gel electrophoresis and analyzed by fluorography (Fig. 7). The results show more than one labeled component. A very broad band of high-molecular-weight labeled material is seen on these gels that is insensitive to collagenase or chona b c d e Figure 7. Acrylamide gel analysis of purified CMsF factor after enzymatic digestion. Acrylamides were run using a 2.9-15% acrylamide exponential gradient gel with a 2.8% acrylamide stacking gel, cast on gel-bond PAG (Marine Colloids, Inc.). The buffer systems were those of Laemmli (1970). Samples were boiled for 15 min in sample buffer containing SOS, but not mercaptoethanol. The gel was stained, to visualize protein standards, and then fluorographed by the PPO/DMSO procedure (Bonner and Laskey 1974). (a) Control; (b) heparitinase; (c) chondroitinase ABC; (d) collagenase; (e) control.
droitinase ABC, but is eliminated by heparitinase. This material is clearly the heparan sulfate-proteoglycan.
Other bands are visible that are not seen in gels of material from sucrose gradient fractions that do not contain neurite outgrowth-promoting activity, and these appear to be proteins or glycoproteins that purify with the heparan sulfate-proteoglycan. A sharp band is visible in samples digested with heparitinase. This could be either the core protein of the heparan sulfate-proteoglycan or another protein (glycoprotein) whose presence was masked in undigested fractions by the broad heparan sulfate-proteoglycan band. Possible models for this class of factors are shown in Figure S. Figure SA is a model of a heparan sulfateproteoglycan that exists free in solution and is not complexed with other molecules. The data at this time do not support this model. Figure SB is a model of the same proteoglycan associated with other molecules that are not required for activity, which resides solely in the heparan sulfate-proteoglycan. Alternatively (Fig. SC), the factor could consist of a complex formed by a heparan sulfate-proteoglycan and other molecules, with the integrity of the complex being required for activity. Such other molecules probably would not include hyaluronic acid or other proteoglycans, since the factor is resistant to chondroitinase ABC. A protein(s) or glycoprotein(s) would be more likely. To explain why only an intact complex would be active, it is not necessary to assume that neurons must recognize more than one element of the complex. One possibility is that neurons recognize only the protein or glycoprotein component, but that the presence of a proteoglycan is necessary to anchor the complex to an appropriate substratum (e.g., polylysine, ECM). Certain molecules known to be secreted by corneal endothelial cells, such as fibronectin and laminin (Gospodarowitz et al. 1981), are plausible candidates for the nonproteoglycan portion of such a complex, but proteins with appropriate M,s have not been identified in the SDS-acrylamide gels of the partially purified factor.
If, as in Figure S, A and B, a proteoglycan is the only molecule essential for promoting neurite outgrowth, the protein core of the proteoglycan must be necessary for the molecule's function, since activity is eliminated by trypsin. Compatible with this is the observation that In each drawing, a heparan sulfate proteoglycan is represented by a protein core <•> to which heparan sulfate side chains (---) are attached. Neurite outgrowth-promoting activity may be associated with a free proteoglycan (.A), or with one that is bound to a protein or glycoprotein (EEi) that is not eseential for activity (8). The third drawing (C) depicts a similar complex, in which the proteoglycan and the glycoprotein are essential for activity. Laminin applied to untreated tissue cuhu re plastic 6 ± 0.9 9 ± 2. 1 8 ± 2.0 2 Fibronecti n applied to PL YS-<:oated tissue culture plastic Solutions of purified laminin or fibronectin (0.05 ml) in Dulbecco's PBS were applied to wells for 12 hr at 4°C . After washing of the substrata. rat sympathetic neurons were cultured upon them for 13 hr. Data represent the mean o f duplicate assays and are the percentage of neurons with neurites. hcparan sulfate alone is not active. The observation that 4 M GuHCl also destroys activity (Lander ct al . 1982) suggests that the proteoglycan may be subject to denaturation that is not readily reversible; this would almost certainly involve the core protein.
If, as in Figure 8C, the factor is a complex. active only in its intact form , then inactivation by trypsin may reflect the loss of any protein-containing component. Purified heparan sulfate would be ex.pected to lack activity. since it lacks components of the complex.. Ex.posure to 4 M GuHCI could destroy activity either because a component of the complex. is denatured by these conditions or because the complex. is unable to reassemble efficiently after being dissociated.

Effect of Other ECM Components on Neurite Outgrowth
Fibronectin, a glycoprotein important in cellular adhesion (Culp et al. 1979) , has been reported to promote the outgrowth of neurites from fetal human sensory neurons ( Baron-Van Evercooren et al. 1982) and embryonic chick retinal neurons (Akers et al. 1981 ), but fibronectin was not an effective promoter of neuritc outgrowth by rat sympathetic neurons in our assay system (Table 5). To examine the possibility that fibroncctin might be a functionally important part of the heparan sulfate-protcoglycan-assoc;iated neurite outgrowth-promoting factors, BCE CMsF was incubated with gelatin-Sepharose and the unadsorbed fraction was assayed for activity in the neurite outgrowth assay. The results in Table 6 show that the activity was not adsorbed to gelatin and is therefore unlikely to contain significant amounts of fibronectin. An antiserum to fibronectin does not block the activity of a similar factor (Collins 1978).
Laminin, another glycoprotein important in cellular adhesion, is a major component of the ordered ECM secreted by endothelial cells and promotes the outgrowth of neurites from fetal human sensory neurons ( Baron-Van Evercooren et al. 1982) and chick central and peripheral neurons (M. Manthorpe et al., in prep.). The results in TableJ 5 show that purified laminin also induces neurite outgrowth by rat sympathetic neurons. Laminin is most effective on a polylysine-coated substratum, but also has a significant effect when applied to untreated tissue culture plastic. In neither case is the responsiveness of neurons comparable with that seen on BCE CMsF-coated substrata, either in percentage of responsive neurons or in length of processes ex.tended by neurons that respond. To examine the possibility that laminin is a functionally critical part of the heparan sulfate-proteoglycan-associated neurite outgrowth-promoting factors , an anti-laminin serum was prepared and tested for its effectiveness in blocking neurite outgrowth on laminin-coated, PC 12 CMsF·Coated, and BCE CMsF·coated substrata. The results in Table 7 show that concentrations of anti-laminin serum that completely block neurite outgrowth on laminin-coated substrata do not reduce outgrowth on either PC12-or BCE-derived CMsF-coated substrata. The antiserum does not prevent outgrowth on CMsF-coated substrata even when included in the neuronal growth medium, so laminin is not likely to be a functionally important part either of the heparan sulfate-proteoglycan-associated factors or of the surfaces of the growth cones of sympathetic neurons.
Although laminin and fibronectin do not appear to be functionally essential components of the heparan sulfate-proteoglycan-associated neurite-promoting factors, both are synthesized by many of the cell types that secrete neurite outgrowth-promoting factors, including BCE cells, RN22 cells (Palm and Furcht 1983), and PCl2 cells (K. Tomaselli, unpubl.). Furthermore, both glycoproteins bind glycosaminoglycans. especially heparan sulfate (e.g., Hay 1981 ). Preliminary experiments using radioimmunoassays indicate that some laminin  may be associated with the purified factor from BCE CMsF and may be deposited on PC 12 CMsp-coated substrata, but not enough to be visible either as an inducer of neurite outgrowth or as identifiable protein bands in SDS-acrylamide gels.

DISCUSSION
Evidence exists that components of the ECM have strong stimulatory effects on neurite outgrowth by rat sympathetic neurons in vitro. The properties of the neuritc outgrowth-promoting factors detected in CM suggest that they are very closely related to each other and to factors described by others (Collins 1978;Dribin and Barrett 1980;Adler ct al. 198 1 ;Coughlin ct al. 1981 ;Henderson et al. 1981). In particular, the factors investigated in this report have densities in CsCI gradients intermediate between those of carbohydrate and protein and are sensitive to d igestion by heparinase and heparitinase, but not chondroitinase ABC. This suggests that each of these factors contains a heparan sulfateproteoglycan as an element essential for activity.
Heparan sulfate-proteoglycans are ubiquitous molecules that may have a great variety of biological functions. They are widely distributed in tissues, including the brain (Toledo and Dietrich 1977), are found on the surfaces of many cell types (Keller et al. 1978), and are present in basement membranes (Hassel et al. 1980;Kanwar and Farquhar 1979). The pol ysaccharide portions of heparan sulfate-proteoglycans are diverse in length, degree of sulfation, and uronic acid composition (Lindahl et al . 1977;Oldberg et al . 1979;Radhakrishnamurthy et al. 1980). Some appear to be integrally associated with cell membranes; others may bind cell-surface receptors (Kjellen et al. 1980;Hurst et al. 198 l). Hepa ran sulfates may be a negative regulator of cell proliferation (Kraemer and Tobey 1972;Chiarugi et al. 1976;Cohn ct al. 1976). They may play a role in neuronal development, since levels are much higher in developing animals than in adults (Margolis et al. 1975).
Heparan sulfate-proteoglycans may also function in the control of cell motility and adhesion, since heparan sulfate is present in the newly formed adhesion sites of fibroblasts , glioma cells, and neuroblastoma cells (e.g., Culp et al. 1979Culp et al. , 1980. Heparinase digestion of the cell surface prevents the !'onnal spreading of fibroblasts on fibronectin-coated surfaces and indicates that heparan sulfate-proteoglycans are critical mediators of cell adhesion (Laterra et al. 1983). The factors described in this report may, in fact, function by increasing adhesion between the neuronal plasmalemma and the substratum.
The one neurite outgrowth-promoting factor that has been purified and studied in detail in this report appears to be a complex containing a hcparan sulfate-proteoglycan and other proteins (or glycoproteins), and sed iments with a high sedimentation coefficient. The properties of this factor, therefore, resemble those of the adhesion-mediating particles described by Schubert andLaCorbicre (1980a,b, 1982), which contain a mixture of glycosaminoglycans, collagen , and glycoproteins. The factor from BCE CMsF. though , appears to have a simpler composition because it contains no sulfated glycosaminoglycan that is sensitive to chondroitinase ABC . The BCE~erived factor also does not aggregate in the presence of Ca+ • , unlike the particles described by Schubert and LaCorbiere (1 982) . The other similarities between these particles, though, suggest that both may promote cell aggregation and adhesion to substrata.
The results also make it clear that not all factors in CM that promote neurite outgrowth attach efficiently enough to a polycationic substrata to be detected in the neurite outgrowth assay. Both fibronectin and laminin promote neurite outgrowth by some neuronal cell types (Akers et al. 1981;Baron-Van Evercooren et al. 1982;M . Manthorpe et al., in prep.) and can be detected immunologically in media conditioned by many of the cells used in these investigations (e.g., Palm and Furcht 1983). Yet antibodies to fibronectin (Collins 1980) or laminin (Table 7) do not diminish noticeably the outgrowth of neurites induced by CMsP"rlerived CM, and neurite-promoting activity has not been seen at lower densities in CsCI gradients. Thus, the standard assay for neurite outgrowth detects only a subset of those molecules that may play a role in vivo. Only polyanionic molecules can be expected to attach efficiently enough in the presence of competing molecules to be detected as inducers of neurite outgrowth in unfractionated CM. Fibronectin and laminin can be studied in isolation only because they are major, well-characterized glycoproteins of the ECM. Clearly , other molecules that are important may exist and require more sophisticated assays for detection.
There is some evidence for selectivity in the response of different classes of neurons to the factors discussed in this report. Laminin markedly induces neurite outgrowth by both central and peripheral neurons (M. Manthorpe et al. , in prep.), but the heparan sulfate-proteoglycan-associated factors have measurable effects in our assay only on neurons that extend processes into the periphery (Lander et al. 1982). Fibronectin appears to have measurable effects on an even smaller subset of neurons. These results are tentative because the heparan sulfate-proteoglycan-associated factors, laminin, and fibronectin differ in potency in this assay.

Possible Role of Factors In Vivo
Antibodies to neurite outgrowth-promoting factors derived from mouse heart, rat pheochromocytoma, and rat Schwannoma have been made that block the response of neurons to these factors and should be useful in future studies on development in vivo (Coughlin and Kessler 1982;M. Manthorpe, pers. comm.;Matthew and Patterson, this volume). Laminin is produced by both Schwannoma cells and astrocytes in vitro (Liesi et al. 1983;Palm and Furcht 1983). Laminin is a prominent constituent . of the endoneurium of peripheral nerves in vivo (Palm and Furcht 1983) and hence is appropriately located to be important in the regeneration and myelination of peripheral nerves. Laminin is not visible at early times duri ng regeneration of newt limbs and hence seems unlikely to guide the pioneer fibers into the blastema of these limbs (Gulati et al. 1983). Fibronectin is present at early times in regenerating blastema and hence could be an important component of the substratum on which pioneer fibers enter peripheral tissues (Gulati et al. 1983). Although fibronectin is not a prominent constituent of the endoneurium of peripheral nerves, enough is present for it also to be important in promoting axon growth during regeneration. The role of laminin in CNS development is not clear. Even though laminin is produced by embryonic astrocytes in culture (Liesi et al. 1983), it has been detected in embryonic rat brain only in association with . capillaries (K. Val!!ntino, unpubl.).
The synthesis of neurite outgrowth-promoting factors may be regulated i3 targets by innervating neurons. Denervation of the chick limb results in a striking increase in neurite outgrowth-promoting activity detected in extracts of the Ii~!> (Henderson et al. 1983). The growth state of neurons also may correlate with the level of these factors . Denervation of the sympathetic ganglion results in increased levels of antigenic substances that are cross-reactive with the neurite outgrowth factor in the sympathetic ganglion (Matthew and Patterson , this volume). The responsiveness of neurons to neurite outgrowth-promoting factors also appears to be regulated by their growth state. Chick ciliary neurons have been shown to lose responsiveness to these factors at approximately the same time as they establish peripheral connections (Collins and Lee 1982). Responsiveness is not regained at later times in vivo in noDWI) ani~. ~ can be regained by culture in vitro. Similar experiments with retinal ganglion cells have shown that prior denervation increases the ability of neurons to extend processes rapidly in vitro (Landretti and Agranoff 1979). Adhesive factors that promote ncurite outgrowth may have additional roles in vivo. · Polycationic substrateadherent factors are required for the survival of chick motor neurons and a class of chick sensory neurons in vitro (Bennett et al. 1980;Edgar and Thoenen 1982). Sensory neurons are able to survive in the presence of much lower concentrations of NGF on a CM-treated substratum than on more conventional substrata (Edgar and Thoenen 1982). Thus, adhesive factors may be important in both modulating neuronal survival and responsiveness to trophic factors during development .
Finally, it is clear that the major components of the ECM described in this report may be important permissive constituents of the substratum that neurites follow in vivo, but their distributions do not explain adequately the precision with which early pathways are established by pioneer fibers or the specificity for different pathways exhibited by nerve fibers during later development. There may be molecules, precisely placed in the extracellular environment, which guide growth cones along the many stereotyped pathways they follow, but if such molecules are there, we must keep looking for them.

Note Added in Proof
We have recently completed the purification of the neurite outgrowth-promoting factor from BCE cells. Our results confirm that the factor consists of several proteins found associated with a heparan sulfate proteoglycan. Preliminary evidence suggests that the protein portion of the factor interacts with neurons, and the proteoglycan portion mediates the binding of the factor to the substratum.