Characterization of a neuronal subtype of insulin-like growth factor I receptor.

Primary neuronal cultures from fetal rat brain were utilized to investigate the possible role of insulin-like growth factor I (IGF-I) in neuronal growth and differentiation. 125I-IGF-I binding to intact cultured neurons was specific and saturable with an apparent Kd of 7.0 +/- 1.2 nM and a Bmax of 1.8 +/- 0.3 pmol/mg protein. Binding of 125I-IGF-I to neurons was inhibited by IGF-I, followed by IGF-II and insulin. 7 S nerve growth factor, but not beta-nerve growth factor, also inhibited 125I-IGF-I binding. A similar binding site was detected on brain membranes. Affinity cross-linking of 125I-IGF-I to intact cultured neurons revealed, under reducing conditions, a major binding moiety with an Mr of 115,000 and a minor component at Mr 260,000. The former represents a neuronal type of the IGF-I receptor alpha subunit, whereas the latter probably represents an alpha dimer. The Mr = 115,000 binding component for 125I-IGF-I was also present in membranes prepared from postnatal whole brain. In contrast, the binding moiety in cultured glial cells was of Mr = 135,000, which was identical to the IGF-I receptor alpha subunit of placenta. Thus mature brain, despite its cellular heterogeneity, expresses a structural subtype of IGF-I receptor which appears to be unique to differentiated neurons. Moreover, glial and neuronal cultures secreted a polypeptide which specifically bound IGF-I; the apparent Mr of this binding protein was determined by affinity cross-linking to be approximately 35,000. The presence of neuronal IGF-I receptors and binding proteins suggested that IGF-I may exert neurotrophic effects on developing neurons. This possibility was supported by the observation that IGF-I markedly stimulated neuronal RNA synthesis.

growth medium containing transferrin is sufficient to sustain the growth of cultured central nervous system neurons. Thus, the study of the interaction of IGF-I with its receptors on cultured neurons may yield important clues about the role of growth factors in regulating neuronal development (14,15).
In fact, the availability of IGF-I antibodies and binding proteins has permitted the detection of IGF-I in brain tissue, brain explants, and dissociated neurons (2)(3)(4). The recent detection of IGF-I mRNA in neonatal rat brain (5) suggests that this peptide may be a growth factor endogenous to the developing brain itself, where it may serve an autocrine or paracrine function. This possibility is strengthened by the finding of high-affinity specific receptors for IGF-I in brain membranes (6) and cultured brain cells (7) and by the detection of specific IGF-I binding proteins in explants from fetal brain (8).
The type of brain cell which expresses IGF-I receptors has not been previously established, but neuronal cells apparently respond to IGF-I. The consequences of stimulation of IGF-I receptors depend on the developmental stage of the brain and may include long term trophic effects on undifferentiated neural cells and short term modulation of mature neurons (9)(10)(11)(12)(13). In neuroblastoma SH-S45Y cells, insulin and IGF-11, possibly acting on the IGF-I receptor (9), elicit neurite formation and induce tubulin mRNA production (10). Furthermore, IGF-I receptors in brain persist into adulthood (11) when neurons are no longer proliferative, suggesting a role beyond their developmental function. Specific binding of IGF-I to the anterior pituitary and hypothalamus supports the possibility that this peptide is involved in regulating central nervous system and pituitary function in mature mammals (11). IGF-I does indeed modulate release of somatostatin from hypothalamic tissue, apparently participating in negative feedback on growth hormone release (12) and may act as an endogenous satiety factor (13).
In this report, we characterize a structural subtype of IGF-I receptor present in mature brain and in enriched primary neuronal cultures and we contrast it with IGF-I receptors of glial cells and peripheral tissues. We also describe a soluble IGF-I binding protein produced by cultured neural cells. In addition, we present evidence that IGF-I is neurotrophic in this system because of stimulation of RNA synthesis after brief as well as prolonged exposure to the growth factor. Primary neuronal cultures may, therefore, represent a useful model for investigating the role of IGF-I in neuronal development.
EXPERIMENTAL PROCEDURES Materials IGF-I was obtained from Amgen Biologicals Inc.; other growth factors including insulin, 7 s and 8-NGF, IGF-I1 (MSA), and epidermal growth factor were purchased from Collaborative Research; DSS was supplied by Pierce Chemical Co.; culture media were obtained from Gibco; and Na'=I was obtained from Cinti-Chem. lZ5I-IGF-I was prepared by the chloramine-T method.

Methods
Cell Cultures-Brains of 16-day-old rat embryos were aseptically removed and freed of meninges, mechanically dissociated, and plated in poly-D-lysine-treated 24-well culture dishes in serum-free medium (basal medium of Eagle with Earle's salts and L-g1utamate:Ham's F-12, 1:1, with glucose, glutamine, penicillin, streptomycin, transferrin, and 1 p~ insulin) as previously described (16); these cultures after 5 days in vitro were >95% neuronal, judged by immunohistochemical staining for neurofilament protein. Astroglial cultures were prepared from 16-day-old rat embryonic cortex or neonatal cortex by replating polygonal adherent cells as previously described (17); cultures enriched in fibroblasts were obtained from meninges of the same brains. Both types of glial cultures were grown to confluence in the neuronal medium with the addition of 10% fetal calf serum. Morphological appearance and cell-type specific staining for glial fibrillary acidic protein or fibronectin indicated that these cultures were predominantly (>80%) astroglia or fibroblasts, respectively.
Binding of IGF-I to Intact Cell.-Binding was carried out with cells still adhering to culture dishes. Cells were washed twice with PBS containing 0.1% BSA, pH 7.4 (PBS/BSA). '251-IGF-I (25,000 cpm/ IO6 cells/well; 112 pCi/pg IGF-I), with or without the appropriate concentrations of competing peptides, were added in this buffer in a final volume of 200 pl. Incubation at 15 "C for 90 min was followed by 4 X 30-5 washes with ice-cold buffer. Cells were harvested in 200 p1 of 1 N NaOH and bound lZ5I-IGF-I was measured in a y counter. Nonspecific binding was defined as that obtained in the presence of Binding of IGF-I to Membranes-Crude membranes were prepared from human placenta (18) or from whole brain of Sprague-Dawley rats after removal of nuclei and myelin. They were suspended in PBS/BSA at 0.5 mg of protein/ml and 100-p1 aliquots were incubated with labeled ligand (25,000 cpm/assay; 112 pCi/pg). After 60-min incubation at room temperature, the reaction mixture was layered over an air cushion over 0.3 M sucrose with 0.1% BSA and 25 mM HEPES in 400-pl Eppendorf tubes and centrifuged at 40,000 X g for 20 min. The pellet was obtained by cutting off the tips of the tubes and bound radioactive ligand was determined.
Affinity Cross-linking of '251-IGF-I to Intact Cell. and Membranes-Intact cultured cells (lo6 cells/well) or membranes from placenta or brain (50 pg of protein/well) were labeled with lZ51-IGF-I (500,000 cpm/assay; 112 pCi/pg) with or without competing peptides, and unbound ligand was removed, using the procedures described in the two preceding paragraphs. Disuccinimidyl suberate (DSS), freshly dissolved in dimethyl sulfoxide, was added as a 100 X stock to yield a final concentration of 1 mM (except when indicated otherwise) in 25 mM HEPES, pH 7.4. The cross-linking reaction was quenched after 15 min at 4 "C by the addition of 50 pl of 50 mM Tris-HC1, pH 7.7, then after two washes cells or membranes were suspended in 100 pl of SDS sample buffer containing 40 mM dithiothreitol. Samples were boiled for 3 min and analyzed by SDS-polyacrylamide gel electrophoresis using 7.5% acrylamide gels (19) followed by autoradiography with Kodak Blue Brand or X-AR film. Characterization of IGF-I Binding Protein-Conditioned medium (50 pl) from neurons grown 5 days in serum-free medium was used directly for cross-linking studies; the procedure was as described above, except that the washing steps were omitted. Partial purification of the binding protein was obtained by concentrating conditioned media 10-fold using Amicon cones (Mr = 10,000 cutoff) followed by chromatography on 10-ml AcA 44 columns using PBS/BSA as elution buffer. One-ml fractions were tested at 1:l dilution in PBS/BSA for interference with IGF-I binding on intact neurons, as compared with 5 and 10% concentrated conditioned medium.
Incorporation of PHlUridine by Cultured Neuron.-Cells were grown for 3 days in defined medium. In some wells, the medium was replaced with fresh insulin-free medium. The cultures were grown for three additional days with or without the addition of IGF-I, then [3H] uridine (5 pCi/well) was added to the medium. Incubation was continued at 37 "C for 2 h. Cells were washed extensively with cold PBS and incorporated radioactivity measured no change in specific incorporation into RNA was obtained if cultures were further washed with acid alcohol. Alternatively, the short term effects of IGF-I were examined by adding it together with [3H]uridine followed by a 2-h incubation.

RESULTS
Binding of IGF-Z to Neurons in Primary C~lture-"~~I-1GF-I bound to intact neuronal cells from 16-day-old embryonic brains cultured for 7 days, and up to 70% of the binding could be displaced by unlabeled ligand. The specific binding reached equilibrium after 40 or 60 min at 21 "C or 4 "C, respectively.
Increasing concentrations of *251-IGF-I yielded saturable binding with a B,, of 1.8 * 0.3 pmol/mg protein and an apparent Kd of 7.0 f 1.2 nM at 15 "C ( Fig. 1) and a BmaX of 1.7 pmol/ mg and a Kd of 3.3 nM at 4 "C (data not shown). The corresponding values, obtained when brain membranes were evaluated, were 0.18 pmol/mg protein and 1.1 nM (data not shown).
Specificity of ZGF-I Binding-"251-IGF-I binding to intact neurons was inhibited not only by IGF-I, but also by relatively low concentrations of 7 S NGF, whereas IGF-I1 and insulin were less effective (Fig. 2 A ) . Inhibition of IGF-I binding to brain membranes by related peptides was also in the order IGF-I > 7 S NGF > MSA > insulin (Fig. 2B). With the exception of 7 S NGF, these cross-reactivities are consistent with previously characterized IGF-I receptors. Since the interaction of NGF was unexpected, attempts were made to rule out possible artifacts, such as binding of IGF-I to NGF. NGF apparently did not act as a binding protein for IGF-I, because preincubation of labeled IGF-I with NGF did not alter the elution pattern of '251-IGF-I from an AcA 44 column (not shown). However, we cannot exclude other potential artifacts.
Although 7 S NGF displaced '251-IGF-I binding with a Ki of less than 5 nM, P-NGF obtained from the same commercial source (Collaborative Research Inc.) had no displacing activity at 100 nM (not shown).
Affinity Cross-linking of '251-IGF-I to Neurons and Other Celts-Whereas Iz5I-IGF-I labeled the (Y subunit of placental IGF-I receptor at M , = 135,000, cross-linking of the ligand to receptors from cultured neurons labeled a polypeptide corresponding to M, = 115,000, after subtraction of the M , = 7,500 contributed by the cross-linked '251-IGF-I. 1251-IGF-I could be effectively cross-linked to the M, = 115,000 polypeptide at concentrations between 0.3 and 3.0 mM DSS (Fig. 3A). Unlabeled IGF-I prevented the cross-linking of the radiolabeled ligand in a concentration-dependent manner (Fig. 3B). The cross-linking could also be inhibited by 7 S NGF and insulin at the appropriate concentrations (Fig. 3A). An M , = 260,000 polypeptide was also labeled by '251-IGF-I. Although this Cultures were grown in serum-free medium for 7 days, washed with PBS/BSA and exposed to varying concentrations of lZ5I-IGF-I (112 pCi/pg) at 15 "C for 90 min, and binding was determined as described under "Methods." Each data point represents the mean of three separate experiments, each done in quadruplicate. corresponds in size to the receptor for IGF-11, the fact that binding was inhibited by insulin, which does not interact with the IGF-I1 receptor, but not by low concentrations of MSA itself (not shown), suggests that it may be a cross-linked a! dimer. In nonreducing conditions, the major labeled species was at MI -400,000 (data not shown).

Neuronal IGF-I Receptors
lZ5I-IGF-I could also be specifically cross-linked to an M, = 115,000 polypeptide from brain membranes of rats older than 1 week (Fig. 3C). Under identical conditions, the IGF-I receptor from placenta yielded an M, = 135,000 polypeptide, in agreement with previous work on the cross-linking of lZ5I-IGF-I to the a! subunit of the receptor (Fig. 3C). Interestingly, primary rat astroglial cultures (Fig. 30) and rat fibroblasts (not shown) had receptors with the same apparent M, as the placental receptor. This was true whether glia were prepared from the same 16-day-old embryos, as were the neurons, or from neonatal cerebral cortex. The pattern of 1251-IGF-I crosslinking to cultured neurons was not altered following addition of 10% serum for 2 days, nor was the glial pattern altered by deletion of serum for 4 days, suggesting that serum factors such as protease inhibitors were not responsible for the difference between the neuronal and glial receptor a! subunits. We have also found the high molecular weight receptor in brain of 15-day-old embryos, while neonatal brain appears to have an intermediate form (not shown).
Secretion of IGF-I Binding Protein by Cultured Cells-Since diminished binding of IGF-I was obtained in the presence of conditioned medium, it seemed that cultures might be producing either high concentrations of endogenous IGF-I or a binding protein which reduced the effective concentration of added IGF-I at the receptor. In fact, when partially purified conditioned medium was added back, '251-IGF-I binding was C.

45-
B. 35,000 which is the major specific IGF-I binding site on intact astroglia. E, cross-linking of '"I-IGF-I to conditioned medium (CM). Arrow denotes M, = 35,000 polypeptide which specifically binds ' "I-IGF-I. Autoradiograms were exposed for 2 days at -120 "C for all but astroglia, which was exposed for 2 weeks, using Kodak X-AR with DuPont Cronex Lighting plus enhancing screens. reduced (Fig. 4) and the inhibitory moiety migrated on an AcA 44 gel permeation column with a Kav consistent with an apparent MI of 30,000-50,000. The inhibitory component was also able to undergo a stable interaction with lZ5I-IGF-I, resulting in an apparent increase in the ligand size (Fig. 4).

FIG. 4. IGF-I binding protein secreted by neuronal cultures.
Top, elution profile of '=I-IGF-I from AcA 44 column with (0) or without (0) preincubation with 100 pl of serum-free conditioned medium from primary cultures of neurons 5 days in oitro. Samples were eluted at 4°C from a 10-ml AcA 44 column with PBS/BSA. One-ml fractions were collected and '"I-IGF-I was measured by y counting. Bottom, inhibition of 9 -I G F -I binding by conditioned medium. Conditioned serum-free medium from neuronal cultures was concentrated and fractionated by elution from a 10-ml AcA 44 column using cold PBS/BSA. Binding of '=I-IGF-I to intact cultured neurons was measured in the presence of PBS/BSA alone, of 5% or 10% concentrated conditioned medium, or of 50% of various AcA 44 column fractions. Data are expressed as percent of inhibition of total binding corrected for nonspecific binding in the presence of 50 nM IGF-I (100% inhibition).
These data suggested that the inhibitor was interacting with IGF-I itself rather than with the receptor, indicating the presence of an IGF-I-binding protein. This possibility was confirmed by the cross-linking of '251-IGF-I to a soluble polypeptide with an apparent M, of 35,000. This protein was able to bind '%I-IGF-I selectively such that the binding was displaced by unlabeled IGF-I, but not by NGF or insulin (Fig.   3E). An M, = 35,000 polypeptide was also specifically crosslinked to Iz5I-IGF-I in the intact astroglial preparations (Fig.  3 0 ) and might represent binding protein adhering to the glial membranes with relatively high affinity. The IGF-I-binding protein was produced in much greater quantities by astroglial cultures than by neuronal cultures. It is possible, in fact, that contaminating glia initially present in neuronal cultures are the source of the binding protein.

IGF-I Stimulation of PHIUridine Uptake-Uptake of [3H]
uridine into intact neurons which had been incubated from the third to the sixth day in vitro with IGF-I and/or fresh medium was enhanced ( Table I). Concentrations of IGF-I as low as 1 nM increased RNA synthesis. The stimulatory effect of IGF-I was strongly enhanced in cultures which had fresh medium. This effect was probably due to the removal of IGF-I-binding protein which was secreted by the neuronal cultures, Cultured neurons were grown for 3 days in serum-free medium containing 0.1 p~ insulin and for 3 additional days with fresh medium or the same medium. IGF-I was added either on day 3, providing 3 days of hormone exposure, or on day 6 at the time of [3H]uridine addition. Cultures were incubated for 2 h after the addition of 5 pCi of [3H]uridine, then washed extensively. Data were calculated as percentages of control uptake in the presence of conditioned medium with insulin (44,600 cpm) and are means f S.E. of three values in two separate experiments.  (Table I).

DISCUSSION
Cultured dissociated neurons expressed relatively high levels of IGF-I receptors. Affinity cross-linking studies demonstrated that the LY subunit of the IGF-I receptor of cultured dissociated neurons and adult brain has an M, of approximately 115,000, whereas the IGF-I receptor of glia and fetal brain resembles more the peripheral-type receptor with an apparent M, of 135,000. After these studies were completed, we learned that another group has also observed an M, = 115,000 IGF-I receptor a subunit in brain (20). The finding of the low molecular weight form of the receptor in mature brain seems surprising since glial cells comprise the majority of brain cells. This may be attributed to the lesser abundance of receptors on glia than on neurons (not shown). Moreover, the concentration of IGF-I receptors in intact purified primary neuronal cultures is about 5-fold greater per mg of protein than that in brain membranes; this comparison actually underestimates the contribution of neurons in brain membranes because the latter should be more enriched in membrane receptors than are intact whole cells. These data further suggest that neurons, rather than glia, account for the majority of receptors in brain membrane preparations.
The structural difference between the neuronal and glial receptors may serve as an important marker in the differentiation pathway from progenitor neuroblast. In this regard, it was previously reported that IGF-I receptor in human brain at mid-gestation (14-18 weeks (21)) is structurally identical to placental receptor and that high affinity IGF-I binding sites begin to appear just after mid-gestation (6). Since neuroblasts begin to differentiate at this time, these results may indicate that the appearance of a neuronal-type receptor may represent an important aspect of neuronal differentiation. Our finding that fetal brain and nondifferentiated neuroblastoma N4TG1 cells express the peripheral receptor type' is consistent with this view.
Insulin receptors in brain are also of distinctly lower molecular weight than those in other tissues (22-25) which may be due to brain-specific differences in glycosylation (22,23,27); associated functional differences unique to brain insulin S. K. Burgess, S. Jacobs, and P. Cuatrecasas, unpublished data. receptor include lack of down-regulation (24, 26) and lack of insulin-stimulated glucose transport (24). The functional relevance of the neuronal subtype of IGF-I receptor is not known, nor has the molecular mechanism which is involved in producing a structurally modified receptor been determined. Developmentally regulated changes in post-translational processing have been proposed to account for the presence of structurally modified forms of N-CAM (28, 29) and c-src (30) in post-mitotic neurons. However, these changes in processing might be secondary to differences in primary sequence of the proteins; for example, due to tissuespecific gene processing which yields alternative mRNAs from a single gene. Recent evidence suggests that the gene for csrc is subject to developmentally regulated alternative processing (31). While IGF-I receptor and c-src might be structurally related (32, 33), the possibility that a similar scheme could account for the appearance of neuron-specific IGF-I receptors can be experimentally approached only when the IGF-I receptor gene is characterized.
It has been previously reported that NGF inhibits IGF-I receptor binding in developing chick brain (34). The data here show that 7 S NGF, but not 2.5s P-NGF, inhibits IGF-I binding to the rat brain IGF-I receptor, possibly indicating that either the entire 7 S complex is required or the a or y subunits are the active species. However, since P-NGF is the species that binds to its own receptor in brain (35) and since it is the species that has some structural similarity with the insulin-like family of peptides, this finding is surprising and its significance is unclear. Inhibition does not appear to result from the formation of a complex between IGF-I and 7 S NGF since 7 S NGF did not alter the mobility of IGF-I on an AcA 44 column (data not shown). Nevertheless, other artifactual explanations are possible. Although 7 S NGF and unlabeled IGF-I were equipotent on a molar basis in inhibiting the binding of 1251-IGF-I, on a mass basis, much greater amounts of NGF were necessary. Therefore, the possibility that a minor contaminant was responsible for the inhibitory activity cannot be ruled out.
Neuronal cultures secreted a binding protein(s) for IGF-I in sufficiently large amounts that no IGF-I binding could be detected in the presence of conditioned medium. Since this binding protein was much more abundant in purified astroglial cultures (in serum-free medium) than in enriched neuronal cultures, it is quite possible that glial cells which initially contaminate the enriched neuronal cultures are the cellular source for IGF-I-binding protein. IGF-I-binding proteins have been described for various cell types (36), including cultured neural cells (3,8). Such proteins may serve as storage forms for extracellular IGF-I. Cross-linking of labeled IGF-I to conditioned medium from cultured neurons revealed a major specific binding species of M, = 35,000; the predominant IGF-I binding species on intact cultured astroglia had the same apparent Mr. The ability of developing brain to synthesize an IGF-I-binding protein as well as possibly IGF-I itself (8), suggests a dual mechanism by which the effective concentration of IGF-I can be regulated. Furthermore, a possible interaction between neurons and the binding protein itself has not been ruled out.
The implications of the presence on cultured neurons of an IGF-I receptor remain to be elucidated. However, the stimulation of RNA synthesis after either brief or prolonged expo-sure to IGF-I suggests a neurotrophic role for this molecule. Investigating the mode of induction of neuronal RNA synthesis and the possible involvement of IGF-I receptor tyrosine kinase activity (20, 37) may provide useful clues to neurotrophic mechanisms. In addition, clarification of the molecular events responsible for production of a neuron-specific IGF-I receptor may provide insight into important principles of neuronal differentiation.