The regulated secretion and vectorial targeting of neurotrophins in neuroendocrine and epithelial cells.

The varied roles that neurotrophins play in the development and activity-dependent plasticity of the nervous system presumably require that the sites and quantity of neurotrophin release be precisely regulated. As a step toward understanding how different neurotrophins are sorted and secreted by neurons, we expressed nerve growth factor (NGF), brain-derived neurotrophic factor, and neurotrophin-3 in cell lines used as models for neuronal protein sorting. All three neurotrophins were secreted by a regulated pathway in transfected AtT-20 and PC12 neuroendocrine cells, with a 3-6-fold increase in neurotrophin release in response to 8-bromo-cAMP or depolarization, respectively. To determine if the propeptide directs the intracellular sorting of mature NGF, we examined mutants in which regions spanning the propeptide were deleted. These mutants underwent regulated release in every case in which expression could be detected. Similarly, NGF sorting was not significantly altered by mutations which specifically abolished N-glycosylation or proteolytic processing sites within the NGF precursor. Finally, we found that all three neurotrophins were secreted 65-75% basolaterally by polarized Madin-Darby canine kidney epithelial cells. These findings suggest that the determinants of regulated neurotrophin secretion lie within the mature neurotrophin moiety and that NGF, brain-derived neurotrophic factor, and neurotrophin-3 are likely to be sorted similarly and released in a regulated manner by neurons.

that neurons compete for limiting amounts of factors secreted by the target tissue, with only the successful competitors surviving and establishing functional connections (4). In the periphery, (non-neuronal) target tissues are believed to secrete neurotrophins constitutively and regulate neurotrophin mRNA independent of neuronal input (5,6). Recent studies, however, have unveiled additional complexities in how neurotrophins function. For example, neurotrophins have been implicated in a variety of activity-dependent processes in the central nervous system (reviewed in Refs. 7 an 8) including ocular dominance column formation (9), activity-dependent survival of cortical neurons (10), axonal sprouting, and the long term enhancement of synaptic transmission (11). These studies suggest that the expression and/or release of neurotrophins may be regulated by activity. Furthermore, autocrine or paracrine, as opposed to target-derived, interactions appear to promote the survival of some mature neurons (10,12,13), and neurotrophins may also be transported in anterograde direction and participate in neuron to target signaling (14 -16). Since neurons are polarized cells that can target proteins to at least two distinct domains (the axon and somatodendritic domain) and can secrete proteins in either a regulated or constitutive manner, understanding how neurotrophins function requires an understanding of how the neurotrophins are intracellularly sorted and secreted.
The neurotrophins are initially synthesized as precursors containing an amino-terminal propeptide that is proteolytically processed to release the mature neurotrophin. In the case of NGF, this propeptide is glycosylated (17,18), contains multiple potential cleavage sites, and is required for the proper folding of mature NGF (18). Because neurotrophins are normally expressed at extremely low levels, little is known about how they are sorted and secreted in vivo by neurons and other cell types. When heterologously expressed in AtT-20 neuroendocrine cells, in human insuloma (HIT) cells (17), or in hippocampal neurons (19), NGF was found to be secreted via a regulated secretory pathway. Despite these advances, however, fundamental questions about neurotrophin sorting and secretion remain unanswered. For example: are all members of the neurotrophin family released via a regulated secretory pathway? Are there sorting differences between members of the neurotrophin family which are functionally significant, as there are in the platlet-derived growth factor and fibroblast growth factor families (20,21)? Finally, what regions or motifs within the neurotrophin precursor play a role in directing the neurotrophins to specific secretory pathways or subcellular domains?
To address these questions, we have analyzed neurotrophin secretion in cell lines that are well established model systems for the study of neuronal secretion and targeting. The rat pheochromocytoma PC12 cell line and the mouse pituitary AtT-20 cell line sort proteins into the regulated secretory pathway (22) and possess dense core granules that are similar to neuronal large dense core vesicles in terms of size, morphology, and composition (reviewed in Refs. 23 and 24). PC12 cells, but apparently not AtT-20 cells, also contain synaptic-like microvesicles which resembles neuronal synaptic vesicles (25). Both cell lines share a number of other characteristics with neurons such as electrically excitable membranes and the ability to extend neurites with growth cones (26,27). In order to investigate the vectorial targeting of neurotrophins in polarized cells, MDCK cells were used. This is an epithelial cell line which targets surface-bound proteins to one of two distinct domains, the apical and basolateral membranes, by a mechanism which is likely to be, in at least some cases, similar to that used by neurons to sort proteins to axons and dendrites, respectively (28,29).
In this study we have used transient and stable transfection of NGF, BDNF, and NT-3 to compare the intracellular sorting of the different neurotrophins in AtT-20 and PC12 cells. We have also used site-directed mutagenesis to investigate the role of the NGF propeptide and to determine if mechanisms that are known to influence protein sorting such as N-linked glycosylation (30), proteolytic processing (31,32), or alternative splicing (33,34) play a role in neurotrophin sorting. We show that NGF, BDNF, and NT-3 are all sorted into the regulated pathway in both types of neuroendocrine cells. Furthermore, this targeting does not require N-glycosylation or proteolytic processing of the NGF propeptide and is not affected by deletions scanning most of the propeptide. These findings suggest that different neurotrophins are sorted and secreted similarly, and information required for this sorting is likely to lie within the mature neurotrophin rather than the propeptide.

MATERIALS AND METHODS
Expression Plasmids-The cDNA for mouse NGF, rat BDNF, and rat NT-3 were all subcloned into pBJ-5, an SR␣-based expression plasmid (35), as described previously (18,36). NGF precursor mutants were constructed by the polymerase chain reaction-based overlap extension method, and the coding region of all mutants was sequenced by the dideoxynucleotide method to ensure that errors were not introduced during the polymerase chain reaction process. The construction of the R(Ϫ4)Q, N(Ϫ8)Q, and deletion mutants, as well as their expression in COS cells, have been previously described (18,36).
Cell Culture and Transfection-COS-7 cells and PC12 cells were grown in DMEM supplemented with 6% horse serum, 6% bovine calf serum (Hyclone). AtT-20-F2 cells (obtained from A. Lowe) and MDCK type II cells (J clone, obtained from W. J. Nelson) were maintained in DMEM supplemented with 10% fetal calf serum (Life Technologies, Inc.). Penicillin and streptomycin (Life Technologies, Inc.) were added to both media. COS cells were transfected and metabolically labeled for 3 h with 100 Ci of [ 35 S]cysteine (Amersham Corp.) as described previously (18). AtT-20 were stably transfected as described previously (22) with the following modifications: 80 g of expression plasmid and 20 g of SV2-neo were used to transfect 1 ϫ 10 6 cells by the calcium phosphate method (Specialty Media transfection kit). Neurotrophin-expressing clones were selected using cloning cylinders and screened by ELISA.
Transient transfections of AtT-20, PC12, and MDCK cells were performed using LipofectAMINE (Life Technologies, Inc.) as indicated in manufacturer's instructions. The following amounts of expression plasmid DNA and LipofectAMINE were used to transfect 2-3 ϫ 10 6 cells/ 10-cm plate: for AtT-20 cells, 25 g of DNA and 100 l of Lipo-fectAMINE; for PC12 cells, 12 g of DNA and 75 l of reagent; for MDCK cells, 5 g of DNA and 15 l of LipofectAMINE. COS cells were transiently transfected using the DEAE-dextran/chloroquine method as described previously (18).
AtT-20 and PC12 Secretion Studies-AtT-20 cells were removed from a single 10-cm plate by trypsinization 24 h post-transfection, equally distributed to the wells of a six-well plate, and allowed to grow for an additional 40 -48 h in full medium. Cells were then washed twice with DMEM, and three wells were incubated in DMEM, 0.2% BSA (unstimulated controls), while the other three were stimulated by incubation in DMEM, 0.2% BSA with 5 mM 8-Br-cAMP (Sigma) added. Conditioned medium was removed after 6 h and cleared of cellular debris by centrifugation; in some experiments the cells were washed with cold phosphate-buffered saline and lysed in phosphate-buffered saline buffer containing 150 mM NaCl, 1% Nonidet P-40, 1% deoxycholate, 0.2% BSA, 1 mM EDTA, and protease inhibitors. Neurotrophin content of conditioned medium or lysate was measured by ELISA as described previously (36), and the results from the three stimulated or unstimulated samples were averaged. Control experiments using purified NGF standards indicated that the use of lysis buffer did not appreciably alter the amount of NGF detectable by ELISA. For cycloheximide (CHX) and actinomycin D experiments, the transfected cells were washed and incubated in DMEM, 0.2% BSA containing 100 g/ml CHX or 10 mM actinomycin D for 1.5 h to inhibit new protein synthesis. Cells were then washed again and incubated for 4 h in the same medium as in the pretreatment. Media and cell lysates were subsequently collected and assayed as described above.
Neurotrophin secretion by transiently transfected PC12 cells was analyzed in a similar manner with the following modifications. Since PC12 cells differentiate in the presence of NGF, 10 ng/ml purified mouse 2.5S NGF (Harlan Bioproducts) was added the medium for all cells 8 h post-transfection; this caused mock-transfected or NT-3-expressing cells to extend neurites to a similar extent as NGF-expressing cells. 72 h post-transfection the NGF-containing medium was removed, and the cells were washed three times with DMEM. Depolarizationinduced secretion was then stimulated by adding 50 mM KCl to DMEM, 0.2% BSA medium during the 15-min incubation. Similar results were obtained when modified Hanks' buffer (125 mM NaCl, 5 mM KCl, 1.2 mM NaH 2 PO 4 , 1.2 mM MgCl 2 , 1 mM CaCl 2 , 1 m ZnCl 2 , 10 mM glucose, 25 mM HEPES, 0.2% BSA, pH 7.4) instead of DMEM was used, and depolarization was induced in modified Hanks' buffer containing 50 mM KCl and 80 mM NaCl.
MDCK Secretion Studies-2.5 ϫ 10 6 MDCK cells were seeded and grown on 24.5-mm collagen-coated Transwell filters (Costar; 0.4-m pore size) as described previously (37,38) with changes of media every day. In general, for each neurotrophin-expressing cell line studied, four filters were seeded in parallel. Three days after seeding, one filter was tested for the presence of functional tight junctions by measuring the diffusion of [ 3 H]inulin (DuPont NEN) from the apical to basolateral compartment for 1 h; filters were used only if the monolayer inhibited greater than 99% of the inulin diffusion. Medium from the other three filters was changed, collected after an additional 24-h incubation, and assayed by ELISA. The percent of neurotrophin secreted into the apical and basolateral compartments was determined for each individual filter, and results from the three filters were averaged. Filter-grown cells were metabolically labeled by incubating for 1 h in labeling medium (DMEM lacking cysteine with 1% dialyzed fetal calf serum added), then adding 250 Ci of [ 35 S]cysteine in labeling medium to the basolateral compartment. After 3-h continuous labeling, the conditioned medium was withdrawn, immunoprecipitated with NGF antiserum, and analyzed by SDS-PAGE. gp81 (clusterin) (39, 40) was detected as described previously (38).
Transient transfection experiments were performed as follows: 3 ϫ 10 6 cells in 1 ml of serum-free DMEM was combined with a 200-l DNA/LipofectAMINE mixture (see above) and seeded on filters. After 6 -8 h, the medium was replaced with DMEM supplemented with 10% fetal calf serum. 48 h after seeding on filters, monolayers were tested for tightness using [ 3 H]inulin and conditioned medium was assayed by ELISA after an additional 24 h. Control experiments indicated that monolayers were polarized by this time as judged by immunostaining using monoclonal antibodies to GP135 (gift of Dr. G. Ojakian, SUNY, Brooklyn), an apical marker, and E-cadherin (DECMA monoclonal, purchased from Sigma), a basolateral marker.
Immunoprecipitations and SDS-PAGE-Immunoprecipitations and 12.5% or 15% SDS-PAGE were performed as described previously (36) with the following modifications. Samples were incubated with rabbit NGF antisera (Sigma) overnight followed by a 2-h incubation with anti-rabbit IgG coupled to Sepharose (Sigma). Conditioned medium from metabolically labeled MDCK cells was analyzed in a similar manner.  Cells-In order to examine the sorting of NGF, BDNF, and NT-3 into the regulated and constitutive secretory pathways, we expressed these neurotrophins in AtT-20 cells by stable and transient transfection. When the cells were transiently transfected with expression plasmids for NGF, BDNF, or NT-3, stimulation with a known secretagogue, 8-Br-cAMP, led to a 3-6-fold increase in the amount of all three neurotrophins detected within the condi-tioned medium as compared to unstimulated controls (Fig. 1A). Note that mock-transfected cells did not secrete significant levels of NGF, BDNF, or NT-3, indicating that there is little endogenous production of neurotrophins by AtT-20 cells.

Secretion of Neurotrophins by
Similar results were obtained when cell lines that stably expressed either NGF or NT-3 were analyzed (Fig. 1B) Four cell lines that expressed detectable levels of NGF, and three that expressed NT-3 were examined. In each case, stimulation with 8-Br-cAMP led to a 2.5-6-fold increase in the amount of neurotrophin released. There was a somewhat smaller increase in the release of neurotrophin after stimulation in NT-3 expressing lines as compared to those expressing NGF, suggesting that there may be differences in the efficiency with which NGF and NT-3 are sorted to the regulated pathway. This difference, however, was generally not observed in transiently transfected cells.
To verify that the 8-Br-cAMP-induced neurotrophin secretion was due to an increase in release from intracellular stores, as opposed to an increase in neurotrophin synthesis, we treated the cells with 100 g/ml cycloheximide or 10 mM actinomycin D for 1.5 h prior to stimulation with 8-Br-cAMP and continued the treatment throughout the entire period of stimulation. Treatment of AtT-20 cells with this concentration of cycloheximide has been shown to inhibit 97% of new protein synthesis within 30 min (41). As shown in Fig. 2, 8-Br-cAMP induces a 4-fold increase in NGF secretion in the presence of cycloheximide. Similar results were obtained after treatment with actinomycin D, demonstrating that new mRNA or protein synthesis is not required for the secretagogue-induced NGF secretion. Interestingly, cycloheximide or actinomycin D treatment decreased NGF secretion by 8-Br-cAMP-stimulated cells but had little effect on unstimulated controls. This is consistent with the finding that cycloheximide treatment impaired regulated secretion to a greater extent than constitutive secretion in AtT-20 cells (41). Treatment with 8-Br-cAMP also led to significant decrease in the amount of NGF detected within cell lysates (Fig. 2). Together, these results provide evidence that the neurotrophins are released via the regulated secretory pathway by AtT-20 cells.
Regulated Neurotrophin Secretion by PC12 Cells-To determine if the neurotrophins were also released via the regulated secretory pathway in PC12 cells, cells were transiently transfected and analyzed as described for AtT-20 cells except that regulated secretion was stimulated by depolarization with 50 mM potassium chloride for 15 min. Similar conditions have previously been used to study the secretion of catecholamines, human growth hormone, and other molecules targeted to the regulated pathway in PC12 cells (42,43). As seen in Fig. 3, depolarization with 50 mM KCl led to a 2.5-4-fold increase in the amount of NGF or NT-3 detected within the conditioned medium. This depolarization-induced release of NGF or NT-3 was not inhibited by cycloheximide or actinomycin D at concentrations that inhibit new protein or mRNA synthesis (44), indicating that this release was not due to new protein synthesis and supporting the conclusion that neurotrophin was being released from preexisting intracellular stores.
Sorting of NGF Propeptide Deletion Mutants-To investigate the mechanism of neurotrophin sorting and how it might be regulated, we used site-directed mutagenesis to examine regions or motifs within the NGF precursor that might be expected to influence the regulated secretion of NGF. Our analysis focused on the NGF propeptide for several reasons. First, this region contains potential sites for proteolytic processing and N-glycosylation, post-translational modifications that have been shown to play a role in the targeting of other proteins to specific secretory pathways (30,32). Second, propeptides have themselves been demonstrated to carry critical sorting information (45,46). Finally, based on our earlier studies, it was observed that mature, biologically active NGF could be produced even after deletion of more than 75% of the propeptide; by contrast, we find that relatively small deletions within the mature NGF moiety can drastically destabilize the protein. 2 Three groups of proNGF mutants were analyzed (Fig. 4). In the first group, the propeptide was divided into five regions and each of these deleted individually. Expression of these mutants in COS cells has already been described (18). In the second group, three potential processing sites were altered. In the third group, two potential N-glycosylation sites were mutated either alone or in combination. A mutant lacking residues in the amino terminus of mature NGF (NGF⌬3-8) (47) was also analyzed.
When the propeptide deletion mutants were expressed in AtT-20 cells NGF was consistently detected in the conditioned medium of cells transfected with the ⌬1, ⌬2, ⌬4, ⌬5, and NGF(⌬3-8) mutants (although the relative expression levels of these mutants varied) but never after transfection with the ⌬3 mutant. When regulated secretion was assessed as described above, stimulation with 8-Br-cAMP led to at least a 3-fold increase in NGF secretion for all of the detectable mutants (Fig.  6A). The lower levels of NGF secretion for some mutants was not due to their selective retention within the cell, as NGF levels were lower within the cell lysates of these mutants as well (Fig. 6A). These data suggest that information required for regulated secretion is not contained within regions 1, 2, 4, or 5 of the NGF propeptide or within the amino terminus of mature NGF. Similar results were obtained for a mutant lacking the three carboxyl-terminal residues of mature NGF (data not shown).
The lack of expression of the ⌬3 mutant is consistent with our earlier demonstration that only region 3 of the propeptide was strictly required for NGF production in COS cells. When this region was deleted, NGF precursor appeared to be rapidly degraded within the endoplasmic reticulum (18), 3 suggesting that it is involved in the proper folding or dimerization of the nascent NGF precursor.
The Role of Proteolytic Processing in Regulated Secretion-The NGF propeptide contains one tetrabasic and two dibasic sites, each of which appears to be cleaved in vitro (48). Furin, and possibly other prohormone convertases, cleave at the third of these sites to liberate mature NGF (49,50). We used sitedirected mutagenesis to eliminate each of these sites individually (PS1, PS2, and CS mutants; see Fig. 4). We also analyzed a mutant (R(Ϫ4)Q) in which the amino-terminal arginine within the RX(K/R)R consensus recognition sequence was altered but the dibasic site left intact (36). When expressed in COS cells, the CS and R(Ϫ4)Q mutants were secreted as uncleaved proNGF (Fig. 5A), which was confirmed by immunoprecipitation with an antibody directed against a region near the amino terminus of the NGF propeptide (51). The PS1 and PS2 mutants were expressed and processed to mature NGF in a manner indistinguishable from wild type NGF (data not shown). When these mutants were transiently expressed in AtT-20 cells, NGF secretion was stimulated by 8-Br-cAMP to a similar extent as in the case of wild-type NGF (Fig. 6B) suggesting that proteolytic processing at dibasic or tetrabasic sites within the NGF precursor is not required for regulated secretion. Because expression of these mutants was too low to detect by metabolic labeling and SDS-PAGE, however, we were not able to rule out the possibility that cleavage occurs at other sites within the propeptide when the dibasic sites are mutated.
The Role of N-linked Glycosylation in Regulated NGF Secretion-The NGF propeptide contains two consensus N-glycosylation sequences. One of these sequences, located 8 amino acids upstream of mature NGF, is conserved in all known neurotrophin precursors. Earlier work demonstrated that the precursor was N-glycosylated at least at this conserved site (17,18). To study the role of N-glycosylation in NGF sorting, we analyzed mutants in which either or both of the acceptor asparagines was changed to glutamine (Fig. 4). When these mutants were expressed in COS cells, mature, biologically active NGF was secreted at levels consistently lower than that of wild-type NGF (Fig. 5A). To verify that both of these sites were used, the lysate of transfected, metabolically labeled cells was immunoprecipitated with NGF antisera followed by SDS-PAGE; as seen in Fig. 5B, the apparent size of the precursor was reduced when either site is mutated, and is further reduced when both are mutated, providing evidence that both sites were used. When these mutants were transiently transfected into AtT-20 cells, NGF release was stimulated 3-6-fold by 8-Br-cAMP (Fig.  6B) demonstrating that N-glycosylation of the NGF precursor is not required for sorting to the regulated secretory pathway.
Neurotrophin Secretion by MDCK Cells-To investigate neurotrophin targeting in polarized cells, stably or transiently transfected MDCK cells were grown on semipermeable filters as described previously (37). After 3-4 days the filters were washed and fresh media added to both the apical and basolateral compartments. Medium was then collected after 24 h and assayed for neurotrophin content by ELISA. Control experiments demonstrated that, within 48 h after seeding, both transiently and stably transfected cells were able to form tight monolayers and target proteins correctly as demonstrated by the polarized staining of GP114, an apical marker, and Ecadherin, a basolateral marker (see "Materials and Methods" for details). As seen in Fig. 7A, when two stable cell lines were analyzed, MNGF.2 and MNGF.5, 65-69% of the total secreted NGF was detected on the basolateral side. A similar percentage was also detected in transiently transfected MDCK cells and in a pool of more than 40 stable clones, indicating that these results were not due to clonal variation (data not shown). To examine the possibility that the targeting of NGF could be regulated by alternative splicing of the NGF gene, we analyzed stable cell lines (MLNGF. 4   Note that mature NGF is not drawn to scale. Constructs were created by polymerase chain reaction double overlap mutagenesis as described under "Materials and Methods." Construction and expression of R(Ϫ4)Q and N(Ϫ8)Q mutants has been described previously (18,36). cDNA encoding the "long" NGF precursor, an isoform arising from alternative splicing of the NGF gene (52). The polarity of NGF secretion in these cell lines did not differ significantly from other NGF-expressing cells (Fig. 7A).
To verify the polarity of NGF secretion, the cell line MNGF.5 was metabolically labeled for 3 h and media from the apical and basolateral compartments were immunoprecipitated with NGF antisera followed by SDS-PAGE (Fig. 7B). MNGF.5, but not the parental MDCK line, secreted a product which appears to be mature NGF based on its electrophoretic mobility and ability to be immunoprecipitated by NGF antisera. 71% of this product was detected basolaterally as determined by densitometry, in agreement with the ELISA results. MNGF.5 retained its ability to target proteins to the apical membrane as evidenced by the secretion of gp81, an endogenous glycoprotein (39,40). BDNF, which was tested only by transient transfection, was secreted 71 Ϯ 0.5% basolaterally (n ϭ 3). Similarly, three cell lines stably expressing NT-3 all secreted 65-70% basolaterally. This percentage was reached by the 3rd day on filters and remained stable through the 8th day (Fig. 7C). Thus it appears that NGF, BDNF, and NT-3 are secreted with a similar polarity by MDCK cells. DISCUSSION In this study we have expressed NGF, BDNF, and NT-3 in neuroendocrine cells containing a regulated secretory pathway and in polarized epithelial cells in order to assess how the neurotrophins are sorted and secreted and whether there are differences between the neurotrophins that could be physiologically significant. Furthermore, we have used site-directed mutagenesis to examine the molecular determinants of NGF sort-ing to the regulated secretory pathway. We observed that NGF, BDNF, and NT-3 were all secreted by the regulated pathway when expressed in AtT-20 and PC12 cells, with a 3-6-fold increase consistently detected in the conditioned media after stimulation. Regulated neurotrophin secretion was observed after transient transfection as well as in stable cell lines with widely varying levels of neurotrophin expression and is therefore unlikely to be an artifact caused by clonal variation or expression level. The NGF findings are consistent with earlier studies (17,19), and, taken together with our data for BDNF and NT-3, suggest that the neurotrophins are sorted similarly by a mechanism which functions in diverse cell types.
For some precursors it is known that sorting information may be contained within the propeptide and the specific cleavage sites used within the precursor can regulate this sorting (31,32). To determine if the NGF propeptide influenced sorting, we analyzed mutants containing deletions which scanned the entire propeptide. Each of these mutants except one (⌬3 mutant) was expressed at detectable levels and underwent regulated release. Mutants in which the consensus N-glycosylation acceptor sites or potential proteolytic processing sites were abolished were also secreted by the regulated pathway. Therefore, for NGF, and presumably the other neurotrophins as well, the signals required for sorting to the regulated pathway are likely to lie within the mature neurotrophin. This is reasonable in light of the fact that mature neurotrophins are highly conserved, sharing ϳ50% amino acid identity, while their propeptides are much less related (53,54). The nature of the sorting "signal" remains to be determined; it may be a linear stretch of amino acids, a structural motif, or a property of the protein such as its net charge (see below).
What is the mechanism by which neurotrophins are directed to the regulated pathway? According to two current models, regulated and constitutive proteins may be segregated on the basis of their affinity for a receptor or, alternatively, by their ability to multimerize or aggregate in the conditions within the trans-Gogi network (55,56). Our data do not directly discriminate between these models. However, while it is known that PC12 cells have both trkA receptors (which bind NGF and to a lesser extent NT-3) and p75 (which binds all known neurotrophins), but not trkB or trkC, none of the known neurotrophin receptors were detected in AtT-20 cells by immunoblotting (data not shown). Since both cell types sort the neurotrophins to the regulated pathway, we conclude that the known neuro-trophin receptors are not required for this sorting. It is possible that the sorting is mediated by low affinity interactions between neurotrophins and molecules within the regulated secretory pathway. For example, the strong positive charge on the neurotrophins (all have pI values greater than 9.0) may promote interactions with negatively charged molecules such as secretogranins, a family of acidic proteins that are believed to act as helper proteins in the sorting of peptide precursors to the regulated secretory pathway. One member of this family, chromogranin B, increases the efficiency of ACTH sorting when overexpressed in AtT-20 cells (57). Interestingly, when either chromogranin B or secretogranin II is coexpressed with NGF in AtT-20 cells the processing of NGF is altered, suggesting that the secretogranins interact with either NGF or a processing enzyme (50). It would be interesting to determine if the efficiency of NGF sorting to the regulated pathway is increased by secretogranin coexpression. Finally, while it is not known if the neurotrophins undergo a milieu-induced multimerization under trans-Gogi network conditions, sedimentation equilibrium data for NGF, BDNF, and NT-3 suggest that all three form some higher order oligomers at pH 7.1 (58).
Neurotrophin Secretion by MDCK Cells-We find that NGF, BDNF, and NT-3 were all secreted with a similar polarity by MDCK cells, with 65-75% of the total neurotrophin detected in the basolateral compartment. The alternatively spliced "long" NGF precursor was secreted identically as the shorter form, a finding consistent with the earlier observation that both forms were sorted to the regulated pathway in AtT-20 cells (17). The secretion of the neurotrophins is considerably less polarized than the greater than 90% basolateral secretion observed for the endogenous basement membrane components laminin and heparin sulfate proteoglycan (59) and only slightly more than the 60 -65% observed for other transfected proteins such as growth hormone, chicken oviduct lysozyme, and cystatin C (40,60), which are all sorted to the regulated pathway in AtT-20 cells but are believed to lack signals for polarized sorting and be secreted by the bulk flow "default" pathway in MDCK cells. The neurotrophin results could therefore be explained by either active but relatively inefficient basolateral sorting or by the absence of epithelial targeting signals.
Implications for Neuronal Sorting of Neurotrophins-Neurons are unique in their morphological and functional complexity and it is unlikely that any cell line can provide a exact model for all the diverse mechanisms that neurons employ to target proteins. In fact, there are substantial differences even between different types of neurons (61). Nevertheless, the existing data suggest that neurons mainly rely on mechanisms also found in non-neuronal cells (24). In particular, there appear to be strong similarities in the exocytotic machinery and sorting mechanisms that neurons, neuroendocrine, and endocrine cells employ for regulated protein and peptide release (23). The earlier findings that NGF undergoes regulated secretion when heterologously expressed in endocrine and hippocampal neurons support this conclusion (17,19). Based on these earlier studies and our own data, we predict that BDNF and NT-3, in addition to NGF, undergo regulated secretion in neurons.
Our findings in MDCK cells suggest that the secretion of NGF, BDNF, and NT-3 is more somatodendric than axonal but not highly polarized. This is consistent with immunocytochemical studies of BDNF localization in the hippocampus and cortex (13,62). However, the MDCK data should be interpreted cautiously. The model that the axonal and dendritic domains of neurons are equivalent to the apical and basolateral domains, respectively, of epithelial cells is based on the localization of transmembrane proteins such as transferrin receptor, vesicular stomatitis virus G protein, viral hemagglutinin protein, and . Cells were then grown as confluent monolayers on 24.5-mm collagen-coated Transwell filters (0.4-m pore size) for 3-4 days. The monolayers were washed, and fresh medium was added. Medium was collected from apical and basolateral compartments after 24 h and assayed for NGF content by ELISA. For each cell line three filters were analyzed; data represent average of three filters Ϯ S.E. B, parental MDCK cells or MNGF.5 cell line were grown on filters as described in A, then metabolically labeled for 3 h with [ 35 S]cysteine (top panel). Medium was immunoprecipitated with NGF antisera (top panel) and analyzed by 12.5% SDS-PAGE. gp81, an apically secreted endogenous protein, was analyzed under nonreducing conditions as described previously (38). C, NT-3-expressing stable cell lines were generated and analyzed as described in A, except that medium was collected every 24 h and assayed. Three filters were grown for each of the three cell lines. Each data point represents the average of three filters Ϯ S.E. For points with no error bars, S.E. was less than 1%. the ␥-aminobutyric acid transporter GAT-1, as well as glycosyl phosphatidylinositol-linked proteins such as Thy-1 (reviewed in Refs. 61 and 63); the parallel between neuronal and epithelial sorting has not proven to be absolute (61,63) and has not been established for secreted proteins. A critical question which remains unresolved is whether protein targeting to different subcellular locations in the neuron (i.e. axon versus dendrite) and sorting to the regulated versus constitutive pathways are determined independently. In some cell types regulated secretion is confined to a single domain (i.e. the apical membrane of exocrine cells). Indeed, it has been suggested that regulated and constitutive neurotrophin secretion occur at the dendrites and cell body, respectively (8). In general terms, protein targeting can be divided into two steps: the sorting of proteins into particular organelles, and the targeting of these organelles to their cellular destinations (24). It may prove to be the case that the first step is well conserved for neurotrophins among diverse cell types, while the second step is more dependent on factors such as the developmental stage, cytoarchitecture, and environment of a given cell.
In conclusion, our data suggest that all the neurotrophins are capable of undergoing regulated secretion and suggest that the signal(s) for sorting to this pathway lie within the mature neurotrophin moiety. Important questions about neurotrophin secretion remain to be answered. For example, can regulated neurotrophin secretion occur locally, i.e. at a single synapse or dendrite, and therefore serve as a means to specifically reward active connections? Can a neuron regulate the targeting of neurotrophins by other mechanisms such as mRNA targeting or by expressing accessory molecules which direct these proteins to specific subcellular locations or secretory pathways? Finally, in what type of vesicles are neurotrophins stored and what stimuli trigger their release? Understanding these and other issues surrounding neurotrophin targeting will help clarify the role of the neurotrophins in the development and plasticity of the nervous system.