Abstract
Sortilin1 (
95 kDa) is a member of the recently discovered family of Vps10p-domain receptors2, 3, and is expressed in a variety of tissues, notably brain, spinal cord and muscle. It acts as a receptor for neurotensin4, 5, but predominates in regions of the nervous system that neither synthesize nor respond to this neuropeptide6, suggesting that sortilin has additional roles. Sortilin is expressed during embryogenesis7 in areas where nerve growth factor (NGF) and its precursor, proNGF, have well-characterized effects6, 7. These neurotrophins can be released by neuronal tissues8, 9, and they regulate neuronal development through cell survival and cell death signalling. NGF regulates cell survival and cell death via binding to two different receptors, TrkA and p75NTR (ref. 10). In contrast, proNGF selectively induces apoptosis through p75NTR but not TrkA11. However, not all p75NTR-expressing cells respond to proNGF, suggesting that additional membrane proteins are required for the induction of cell death. Here we report that proNGF creates a signalling complex by simultaneously binding to p75NTR and sortilin. Thus sortilin acts as a co-receptor and molecular switch governing the p75NTR-mediated pro-apoptotic signal induced by proNGF.
Binding of NGF was examined by surface-plasmon resonance (SPR). As demonstrated in Fig. 1a, sortilin bound mature NGF with moderate affinity (dissociation constant (Kd) 90 nM). In contrast, the affinity of NGF for p75NTR and TrkA was high (Kd 1–2 nM), in accordance with previous studies in cells11, 12, 13. As the NGF precursor (proNGF) may escape intracellular processing and be released extracellularly, we next examined binding of proNGF9, 11, 14, 15. Whereas lack of processing reduces the affinity of proNGF for p75NTR and TrkA (Kd 15–20 nM), it results in a much higher affinity (Kd 5 nM) for sortilin (Fig. 1a). This is surprising because proNGF has been reported to interact with cellular p75NTR, but not TrkA, with a higher affinity (Kd 0.2 nM) than mature NGF, and to selectively induce p75NTR-dependent apoptosis in neurons, smooth muscle cells and oligodendrocytes8, 11. Our data may reflect the participation of a p75NTR co-receptor, and this receptor might be sortilin (see below).
Figure 1: SPR analysis of ligand binding.
a, Binding of 20–1,000 nM NGF, proNGF and GST–pro to immobilized sortilin (51 fmol mm-2), p75NTR (91 fmol mm-2) and TrkA (66 fmol mm-2). Calculated Kd values are indicated. b, SDS–PAGE analysis of ligands used in a. A Coomassie-stained gel (left panel) and a western blot (anti-pro domain antibody; right panel) are shown. c, Binding of proNGF (25 nM) to sortilin (66 fmol mm-2) in the absence or presence of 10 
M neurotensin (dashed line) and 5 M GST–pro (dotted line). The SPR response obtained for the inhibitors alone has been subtracted.
To examine the structural basis of proNGF binding, we produced the pro domain of proNGF as a glutathione S-transferase fusion protein (GST–pro) (Fig. 1b). The GST–pro protein bound to sortilin with an affinity very similar to that of proNGF (Kd 8 nM), but not to p75NTR or TrkA (Fig. 1a). Additional experiments further demonstrated that binding of proNGF to sortilin was inhibited markedly (>75%) by neurotensin, and was almost abolished in the presence of GST–pro (Fig. 1c). Thus, the pro domain constitutes the structural basis for the high-affinity binding between proNGF and sortilin.
We next assessed binding of proNGF to cellular sortilin. We transfected 293 cells without endogenous sortilin with the receptor constructs indicated in Fig. 2a, and evaluated binding of proNGF at 37 °C. Control cells demonstrated no binding or uptake of ligand, whereas cells expressing wild-type sortilin exhibited significant endocytosis of proNGF (Fig. 2b), but not of mature NGF (Fig. 2c). The observed uptake was hampered strongly in the presence of excess neurotensin or GST–pro (data not shown). Furthermore, transfectants expressing a mutant sortilin protein (sortilin(mut)) that accumulates on the plasma membrane owing to disrupted motifs for endocytosis16, displayed intense surface labelling with proNGF, but little uptake.
Figure 2: Binding and uptake of proNGF and NGF in 293 cells.
a, Western blot showing the level of receptor expression in the transfected 293 cells. b, c, Untransfected cells (control) and cells transfected with the indicated receptors were incubated (37 °C, 45 min) with 50 nM proNGF (b) or NGF (c) before fixation and staining with anti-NGF antibodies.
High resolution image and legend (62K)Binding and uptake of proNGF was also investigated in cells transfected with TrkA and p75NTR, and in cells expressing each of the two receptors in combination with sortilin (Fig. 2b). TrkA transfectants exhibited a very modest uptake of proNGF, and on co-transfection with sortilin, endocytosis of proNGF was comparable to that observed in cells expressing sortilin alone. In contrast, uptake of mature NGF was efficient in TrkA-expressing cells and was unaffected by co-transfection with sortilin (Fig. 2c). In p75NTR transfectants, proNGF as well as mature NGF was almost exclusively found on the plasma membrane, indicating a slow or insignificant endocytosis (Fig. 2b, c) consistent with prior observations17, 18. However, coexpression of p75NTR with sortilin re-established uptake of proNGF, and coexpression with sortilin(mut), as well as with wild-type sortilin, induced a significant increase in surface-associated ligand, suggesting a synergistic rather than a simple additive effect of sortilin and p75NTR coexpression.
The findings demonstrate that sortilin exhibits negligible binding of NGF, but also that it conveys a significantly higher capacity for uptake of proNGF than either of the two established receptors (that is, p75NTR and TrkA). Moreover, results in double transfectants suggest that sortilin and p75NTR cooperate to promote proNGF binding.
To characterize the molecular mechanisms underlying the putative cooperativity between p75NTR and sortilin, affinity crosslinking was performed. Crosslinking of 125I-labelled proNGF to p75NTR and sortilin double transfectants produced labelled complexes of 110, 140 and 240 kDa (Fig. 3a, lane 1), but was unproductive in single transfectants, suggesting that coexpression of sortilin and p75NTR is required for efficient binding at subnanomolar concentrations of proNGF (Fig. 3a, lanes 5–6). Furthermore, immunoprecipitation with receptor antisera established that both sortilin and p75NTR were components of the crosslinked adducts (Fig. 3a, lanes 7–8). Similar experiments performed on cells coexpressing p75NTR and sortilin(mut), which has a higher surface expression than wild-type sortilin16, resulted in a quantitative increase in crosslinked complexes (data not shown). Finally, generation of the crosslinking adducts was markedly reduced in the presence of unlabelled proNGF, neurotensin or GST–pro, implying that sortilin, as well as the NGF pro domain, is critical to complex formation (Fig. 3a, lanes 2–4).
Figure 3: ProNGF-induced formation of heterotrimeric complexes comprising sortilin and p75NTR.
a–c, Crosslinking of 125I-labelled proNGF (a, b) or 125I-labelled NGF (c) to transfected 293 cells in the presence and absence of excess proNGF, NGF, GST–pro or neurotensin. Crude lysates (- ) and immunoprecipitated proteins were subjected to SDS–PAGE and labelled bands were visualized by autoradiography. Antibodies used for immunoprecipitation (IP) of sortilin (S), p75NTR (P) and TrkA (T) are indicated. d, Biolabelled cells overexpressing sortilin(mut) and p75NTR were incubated with and without proNGF (25 nM) and treated with a reducible crosslinker. Autoradiographic bands resulting from reducing SDS–PAGE of immunoprecipitates are shown. e, Scatchard plot showing binding of radiolabelled NGF (open circles) and proNGF (closed circles) to cells expressing sortilin(mut) and p75NTR.
High resolution image and legend (64K)We conclude that the 240 kDa adduct probably represents a heterotrimeric complex comprising proNGF, sortilin and p75NTR, whereas the 110 kDa and 140 kDa species constitute proNGF in association with a single receptor. Expression of both receptors is required for efficient binding of proNGF, and our results support a model in which the pro domain and the 'mature' part of proNGF simultaneously engage sortilin and p75NTR, respectively.
Corresponding experiments established that proNGF does not form stable complexes with TrkA (Fig. 3b). In fact, crosslinking with proNGF using cells expressing all three receptors (TrkA, p75NTR and sortilin) resulted in 110, 140 and 240 kDa adducts that could be precipitated with anti-sortilin (data not shown) and anti-p75NTR antibodies but not with TrkA-specific antiserum. Thus, proNGF discriminates between TrkA and p75NTR in cells that express both receptors in combination with sortilin.
In accordance with previous reports12, 13, crosslinking using mature 125I-labelled NGF yielded complexes with both p75NTR (90 and 180 kDa) and TrkA (160 kDa) (Fig. 3c). However, no additional crosslinked complexes were observed when either of the two was coexpressed with sortilin, and sortilin single transfectants did not bind NGF. These results indicate that sortilin neither interacts with mature NGF nor is part of a complex formed on binding of mature NGF to p75NTR or TrkA.
We next examined whether sortilin and p75NTR physically associate on the cell membrane. Cells expressing both receptors were biolabelled and incubated in the absence or presence of proNGF, followed by treatment with a membrane-impermeable reducible crosslinker, lysis and immunoprecipitation using anti-p75NTR antibodies. Sortilin could be crosslinked directly to p75NTR (Fig. 3d). However, in the presence of proNGF, the relative amount of crosslinked and co-precipitated sortilin increased by about fivefold (3.9% to 18.4%). Equilibrium binding studies were then designed to determine whether p75NTR and sortilin coexpression might influence the specificity and affinity of ligand binding. As demonstrated in Fig. 3e, 125I-labelled NGF bound to cells coexpressing sortilin and p75NTR with an estimated Kd of 1.0 nM. This agrees with previous results12 obtained in p75NTR-expressing cells, and as sortilin single transfectants did not bind mature NGF (data not shown), the data indicate that mature NGF binds strictly to p75NTR. In contrast, similar experiments with 125I-labelled proNGF further indicated that sortilin and p75NTR cooperate in proNGF binding. Thus, cells expressing a single receptor type—either sortilin or p75NTR—did not bind 125I-labelled proNGF (data not shown), whereas cells coexpressing these receptors did. Scatchard analysis (Fig. 3e) suggested fewer binding sites for proNGF than for NGF in the double transfectants, but also a higher affinity for proNGF (Kd 160 pM) that could not be accounted for by binding to any single receptor. Accordingly, A875 cells, which bind proNGF with high affinity11, express high levels of endogenous sortilin and p75NTR (Fig. 4a). The results support a model in which proNGF binds to and promotes the formation of a multi-component receptor complex comprising both sortilin and p75NTR.
Figure 4: Sortilin is required for the pro-apoptotic action of proNGF.
a, Receptor expression in proNGF-responsive cell types. b, Crosslinking of 125I-labelled proNGF to SM-11 cells and inhibition by unlabelled competitors. c–e, ProNGF-induced apoptosis (cell type indicated) and its inhibition by GST, GST–pro and neurotensin. The number of apoptotic and total cells counted per condition was 100/300 (d and f) and 300/1,100 (e). All values are normalized to apoptosis in the absence of any additions. f, Killing of SCG neurons from wild-type and p75NTR knockout mice. g, ProNGF-induced apoptosis of Schwann cells transiently transfected with sortilin. Columns indicate per cent of TUNEL-positive cells among cells that express (S+) or lack (S-) sortilin. Left inset shows western blot of Schwann cell lysate; right inset shows apoptotic nuclei (TUNEL-positive, green) in transfectants expressing sortilin (red staining). Asterisk indicates an untransfected cell. h, Schematic model of receptor–complex formation.
ProNGF is more efficient than NGF in inducing apoptosis in superior cervical ganglion (SCG) neurons, vascular smooth muscle (SM-11) cells and oligodendrocytes, and in promoting chemotaxis and ligand binding in A875 melanoma cells8, 11, 19. These cell types all express significant levels of sortilin and p75NTR (Fig. 4a). We found that uptake of proNGF in dissociated SCG cultures was inhibited by the GST–pro protein (data not shown), which selectively inhibits binding to sortilin. Moreover, crosslinking of 125I-labelled proNGF to SM-11 cells resulted in labelled adducts of 110, 140 and 240 kDa, similar to those obtained in the p75NTR and sortilin double transfectants (Fig. 4b). These findings suggest that the biological effects of proNGF require coexpression of sortilin and p75NTR. To assess directly whether binding of proNGF to sortilin regulates biological action, the ability of GST–pro and neurotensin to impair proNGF actions was evaluated. In order to minimize conversion of proNGF into mature NGF, which might introduce a bias by facilitating survival in TrkA-positive cells, a furin-resistant mutant of proNGF11 was used in all subsequent experiments, unless otherwise stated. In SM-11 cells co-expressing p75NTR and sortilin, furin-resistant proNGF was more effective than mature NGF in inducing cell death as assessed by a TdT-mediated dUTP nick end labelling (TUNEL) assay (Fig. 4c). In addition, wild-type proNGF induced apoptosis as effectively as furin-resistant proNGF (data not shown). Co-incubation of proNGF with an excess of neurotensin, or with excess GST–pro, but not GST alone, impaired the induction of cell death and apoptosis by more than 90% (Fig. 4c). Similar findings were obtained in cultured SCG neurons expressing sortilin and p75NTR as well as TrkA. Thus, both GST–pro and neurotensin significantly reduced proNGF-induced apoptosis in SCG neurons, whereas neither affected neuronal survival in response to mature NGF or on NGF withdrawal (Fig. 4d, e). As shown in Fig. 4f, p75NTR-deficient neurons from p75NTR knockout mice20 that express sortilin in the absence of p75NTR (data not shown) exhibit NGF-dependent survival, but are resistant to proNGF-induced killing. A similar resistance to proNGF was seen in Schwann cells expressing p75NTR but not sortilin (Fig. 4g); however, after transfection with sortilin, Schwann cells became sensitive to proNGF-induced apoptosis. Approximately 95% of the cells that expressed sortilin (18% of total) were TUNEL positive, and among all TUNEL-positive cells 96% expressed both sortilin and p75NTR. This sensitivity to proNGF-induced killing was reversed by co-incubation with GST–pro (Fig. 4g) and neurotensin, which blocked binding of proNGF to sortilin. It follows that both receptors are obligate for the induction of proNGF-mediated cell death, whereas sortilin expression has no impact on NGF responsiveness under these circumstances.
We conclude that proNGF targets and promotes formation of a signalling complex comprising endogenous sortilin and p75NTR, and that both receptors are required for proNGF-mediated apoptosis. In contrast, mature NGF preferentially binds p75NTR and/or TrkA, with sortilin having little or no bearing on NGF-initiated signalling.
Our study indicates that the neurotrophins use not two but three distinct receptor classes to dictate and regulate opposing biological responses of survival and death. We identify sortilin as a biologically important neurotrophin receptor that targets the pro domain of proNGF with high affinity. The present data suggest that sortilin is a required component for transmitting proNGF-dependent death signals via p75NTR. Together with p75NTR, sortilin facilitates the formation of a composite high-affinity binding site for proNGF (Fig. 4h). Thus, sortilin serves as a co-receptor and molecular switch, enabling neurons expressing Trk and p75NTR to respond to a pro-neurotrophin and to initiate pro-apoptotic rather than pro-survival actions. In the absence of sortilin, regulated activity of extracellular proteases may cleave proNGF to mature NGF11, promoting Trk-mediated survival signals (Fig. 4h). In conclusion, NGF-induced neuronal survival and death is far more complicated than previously appreciated, as it depends on an intricate balance between proNGF and mature NGF, as well as on the spatial and temporal expression of three distinct receptors: TrkA, p75NTR and sortilin. As sortilin is but one member of the Vps10p-domain receptor family expressed in the nervous system, future studies should show whether other pro-neurotrophins use related Vps10p-containing receptors to switch biological responsivity to neurotrophin isoforms.
Methods
Recombinant proteins and radiolabelling
Human proNGF and mature NGF generated in Escherichia coli21 were a gift from Scil Proteins GmbH. Radioiodinated ligands22, 23 (3,000 d.p.m. fmol-1) were used within 48 h of iodination and their integrity and bioactivity was assayed by SDS–polyacrylamide gel electrophoresis (PAGE), PC12 cell neuritogenesis (NGF) and apoptosis of p75NTR-expressing vascular smooth muscle cells (proNGF). Mature NGF and furin-resistant mouse proNGF were purified from media of transfected 293 cells11. The NGF pro domain (amino acids E19 to R121) was expressed in E. coli as a GST fusion protein and purified on glutathione–agarose beads. The luminal domain of sortilin was expressed and purified as described5. p75NTR-Fc and TrkA-Fc (fusion proteins) were from R&D Systems.
SPR analysis, equilibrium binding and western blotting
The SPR analysis was performed essentially as described5, 24. The receptors were immobilized (at 10–15 g ml-1) on a CM5 chip and remaining coupling sites were blocked with 1 M ethanolamine. Sample and running buffer was 10 mM HEPES, 150 mM (NH4)2SO4, 1.5 mM CaCl2, 1 mM EGTA, 0.005% Tween-20 pH 7.4. After each analytic cycle the sensor chip was regenerated in a 10 mM glycine-HCl buffer. The SPR signal was expressed in relative response units (RU); that is, the response obtained in a control flow channel was subtracted. Kinetic parameters were determined using BIAevaluation 3.1 software. Equilibrium-binding studies were performed as described12. In brief, the cells (2 
106 ml-1) were incubated (4 °C, 40 min) with radioiodonated NGF or proNGF (2–20 10-10 M) in the presence or absence of a 500-fold molar excess of NGF or neurotensin, respectively. Bound ligand was then separated from free ligand by centrifugation through calf serum. Mean values of triplicates (from two independent experiments) were evaluated using the PRISM program.
Western blotting, after reducing SDS–PAGE, was performed using a rabbit antibody (1:1,000) directed against the pro domain of NGF (amino acids 23–81) of human NGF8 and horseradish-peroxidase-conjugated swine anti-rabbit immunoglobulin (Amersham Biosciences).
Transfected cell lines
Parental 293 cells and transfectants expressing p75NTR or TrkA12 were transfected with wild-type sortilin5 or the sortilin(mut) variant impaired in endocytosis (alanine substituted for Y14, L17, L51 and L52 in the cytoplasmic tail)16, and selected using zeocin. Primary rat Schwann cells25, plated on polylysine-coated plates, were transfected with pcDNA wild-type sortilin5 or pcDNA alone using calcium phosphate.
Crosslinking, biolabelling and immunoprecipitation
Cells (2 106 ml-1) were incubated (4 °C, 2 h) with radioiodinated proNGF or NGF (400 pM), in the absence or presence of 100 nM unlabelled proNGF, 40 
M neurotensin or 200 nM GST or GST–pro, followed by crosslinking (15 min) with 4 mM 1-ethyl-3-[3-dimethylaminopropyl]carbodiimide and 25 mM DSS (Pierce). Washed cells were subsequently lysed in 1% Nonidet P40 buffer containing protease inhibitors and precipitations were performed using anti-p75NTR and anti-TrkA antisera13, and anti-sortilin antibody5.
Transfected cells were biolabelled (3–4 h) with l-[35S]cysteine and l-[35S]methionine, then incubated (20 °C, 2 h) with or without 25 nM proNGF and, finally, treated (30 min) with 5 mM of the reducible crosslinker DTSSP (Pierce) before lysis in 1% Triton-X100 buffer containing protease inhibitors5, 26. Immunoprecipitation was performed using rabbit anti-p75NTR ( number 9993), anti-Trk (Santa Cruz) and anti-sortilin5. All precipitated proteins were analysed by reducing SDS–PAGE and were detected by autoradiography.
Induction of apoptosis in various cells
A vascular smooth muscle cell line expressing human p75NTR but not TrkA27 was incubated (16 h) with 2 ng ml-1 of mature NGF or furin-resistant proNGF11 in the presence of 50 nM GST or GST–pro, or 40 M neurotensin. After fixation, cells were fixed, incubated with 4,6-diamidino-2-phenylindole (DAPI) and subjected to TUNEL analysis (Roche Molecular Biochemicals). Results represent the mean value of three independent experiments performed in triplicate. At least 300 cells per condition were counted.
Dissociated P0–P1 rat SCG neurons28 or mouse SCG neurons obtained from p75NTR knockout mice29 or wild-type littermates were plated on collagen-coated slides and maintained for 5 days in 50 ng ml-1 NGF before use. Replicate cultures were rinsed five times with NGF-free medium and treated with or without the given additives, as indicated. After 36 h SCG cultures were processed for TUNEL analysis and counterstained with anti-neuronal-specific 
-tubulin ( Tuj1, Covance)11. TUNEL-positive neurons were scored blindly by the observer and at least 100 cells were counted for each culture condition.
Transfected Schwann cells were replated on 8-well slides (NUNC) at 20,000 cells per well. At 48 h after transfection, cells were treated (18 h) with 5 ng ml-1 mature NGF, purified recombinant cleavage-resistant proNGF or diluent alone. After fixation, the cells were stained using mouse anti-sortilin antibody (Transduction Biolabs, anti-NTR3 612101) and rhodamine goat anti-mouse IgG followed by DAPI incubation, and then subjected to TUNEL analysis (Roche Molecular Biochemicals). At least 1,000 cells per condition were counted in a blinded manner, and results are representative of three independent experiments. Where appropriate, statistical significance was determined by Student's t-test.
6 
