Sortilin Facilitates Signaling of Ciliary Neurotrophic Factor and Related Helical Type 1 Cytokines Targeting the gp130/Leukemia Inhibitory Factor Receptor β Heterodimer

ABSTRACT Sortilin is a member of the Vps10p domain family of neuropeptide and neurotrophin binding neuronal receptors. The family members interact with and partly share a variety of ligands and partake in intracellular sorting and protein transport as well as in transmembrane signal transduction. Thus, sortilin mediates the transport of both neurotensin and nerve growth factor and interacts with their respective receptors to facilitate ligand-induced signaling. Here we report that ciliary neurotrophic factor (CNTF), and related ligands targeting the established CNTF receptor α, binds to sortilin with high affinity. We find that sortilin may have at least two functions: one is to provide rapid endocytosis and the removal of CNTF, something which is not provided by CNTF receptor α, and the other is to facilitate CNTF signaling through the gp130/leukemia inhibitory factor (LIF) receptor β heterodimeric complex. Interestingly, the latter function is independent of both the CNTF receptor α and ligand binding to sortilin but appears to implicate a direct interaction with LIF receptor β. Thus, sortilin facilitates the signaling of all helical type 1 cytokines, which engage the gp130/LIF receptor β complex.

Sortilin (40) is a member of the mammalian Vps10p domain (Vps10p-D) family of neuronal type 1 receptors that also includes SorLA (23) and SorCS1-3 (19,27,32). The hallmark of the family is the N-terminal Vps10p-D consisting of two minor domains and a unique, ligand-binding, 10-bladed ␤-propeller domain (41). The Vps10p-D constitutes the entire luminal part of sortilin and is followed by a single transmembrane segment and a short cytoplasmic domain containing functional motifs for endocytosis as well as intracellular trafficking, e.g., sorting between the Golgi apparatus and endosomes (34). Sortilin is synthesized as an inert proreceptor with a short N-terminal propeptide (amino acids 1 to 44) (38). The propeptide is removed by proteolytic cleavage in late Golgi compartments, and sortilin is thereby converted to the ligand binding mature receptor (38), which binds a growing number of competing ligands, including the receptor-associated protein (RAP), its own propeptide, and, notably, neuropeptides (e.g., neurotensin [NT]) and neurotrophins (e.g., the proforms of nerve growth factor ␤ [NGF-␤] and brain-derived neurotrophic factor [BDNF]) (30,35,38,40,45). In addition to soluble ligands, sortilin also interacts with other transmembrane proteins. Thus, sortilin has been shown to associate with the common neurotrophin receptor p75 NTR and to engage both p75 NTR and proneurotrophins to form a death signaling trimeric complex (24,35,45). Moreover, sortilin may bind neurotensin receptor 1 (NTS1) and was reported previously to modulate NT signaling and the endocytosis of NTS1 (29).
Sortilin binds apolipoprotein E (apoE) (22), which is structurally related to ciliary neurotrophic factor (CNTF) (31,47). They both form homodimers, but the two may also combine to form functional heterodimeric complexes (17); however, the possibility that CNTF, similarly to apoE, targets sortilin has not been explored. CNTF belongs to the family of helical type 1 cytokines, which also includes interleukin-6 (IL-6), IL-11, leukemia inhibitory factor (LIF), oncostatin M (OSM), cardiotrophin-1 (CT-1), neuropoietin, and the heterodimer cardiotrophin-like cytokine/receptor cytokine-like factor 1 (CLC/ CLF-1). These cytokines play key roles in several physiological processes, including immune regulation, host defense, reproduction, food intake, and the regulation of neuronal growth (20). CNTF was initially identified, and named, for its ability to maintain the survival of parasympathetic neurons of chicken ciliary ganglions (2). Since then, it has been shown to support the survival of many different neuronal cell types, including sensory (44) and motor (36) neurons and neurons of cerebral and hippocampal origin (21,28). CNTF lacks a classical signal peptide sequence and is subject to alternative secretion. It is generally believed to act as a lesion factor that is released mainly from tissues subjected to trauma or stress. In support of this notion, several studies have reported a marked change in the localization and expression of CNTF on nerve lesions (15,42,43). CNTF signaling is elicited by the formation of a trimeric receptor complex (7). At first, CNTF binds to nonsignaling glycosylphosphatidylinositol (GPI)-anchored CNTF receptor ␣ (CNTFR␣) (6). The binding of CNTF to the membrane-bound or soluble CNTFR␣ (sCNTFR␣) is followed by the recruitment of the signaling subunits 130-kDa glycoprotein (gp130) and LIF receptor ␤ (LIFR␤) (7). The CNTF-induced heterodimerization of gp130 and LIFR␤ leads to the activation of the Janus kinase/signal transducer and activator of transcription 3 (JAK/STAT3) pathway by the phosphorylation of STAT3, in addition to the Ras/mitogen-activated protein (MAP) kinase and phosphatidylinositol 3-kinase pathways (18,20). CNTFR␣ is the specificity-conferring subunit of the trimeric receptor complex. It is well conserved across species and is highly expressed in neuronal tissue and skeletal muscle (8). However, the receptor is not an absolute requirement for signaling because CNTF at high concentrations (40 nM and above) may directly prompt and activate the gp130/LIFR␤ heterodimer (16). CNTFR␣ is also a receptor subunit for CLC/CLF-1 and neuropoietin, which, similar to CNTF, induce a cellular response via gp130/LIFR␤. In contrast to CNTF, CLC/CLF-1 completely depends on the initial binding of CNTFR␣ for signal transduction (12).
The importance of CNTFR␣ and the associated ligands is reflected by studies of gene deficiency. Thus, the lack of either CNTFR␣ or CLC/CLF-1 results in a severe loss of motor neurons and perinatal death due to a suckling defect (3,9,49). The fact that no particular phenotype has been ascribed to CNTF deficiency may, on the other hand, result from a significant degree of redundancy (13).
In the present study we demonstrate that CNTF, and related ligands targeting CNTFR␣, binds to sortilin with high affinity. We show that a high-affinity site in the C terminus of CNTF interacts with the sortilin ␤-propeller domain and demonstrate that sortilin provides rapid endocytosis and the removal of CNTF. Importantly, we find that sortilin facilitates CNTF signaling and mediates CNTF-dependent proliferation through the gp130/LIFR␤ heterodimeric complex. Interestingly, this function is independent of both CNTFR␣ and ligand binding to sortilin but appears to implicate a direct interaction with LIFR␤. Thus, we find that sortilin facilitates the signaling of all helical type 1 cytokines that engage the gp130/LIFR␤ complex but not of IL-6, which signals through the gp130 homodimeric complex.
Human CNTF cDNA was obtained from RZPD Deutsches Ressourcenzentrum für Genomforschung GmbH. The full-length cDNA and cDNA encoding a version of CNTF containing a deletion of the 13 C-terminal residues (CNTF-tr) were PCR amplified with primers CNTF 5Ј (5Ј-ATA ATA CCA TGG CTT TCA CAG AGC ATT-3Ј) (forward), CNTF 3Ј (5Ј-CAC GAT ATC CTA CAT TTT CTT GTT GTT AGC-3Ј) (reverse), and CNTF-tr 3Ј (5Ј-CAC GAT ATC CTA TGG GAT CCC AGT CTG ATG-3Ј) (reverse). The PCR products were precipitated and cleaved with NcoI followed by DpnI and inserted into the pET11d vector. Before ligation, pET11d was BamHI cleaved with T4 polymerase, cleaved a second time with NcoI, and subsequently dephosphorylated to avoid selfligation. Both full-length CNTF and truncated CNTF were expressed in BL21(DE3) cells.
The cells were resuspended in 20 mM Tris-HCl-10 mM NaCl (pH 7.5) and sonicated, followed by ultracentrifugation. The lysate was run over a Source 30Q column with a total volume (V t ) of 13 ml (GE Healthcare, Denmark) at 4°C. The flowthrough was collected and precipitated in 30% saturated ammonium sulfate overnight at 4°C. The pellet was resuspended in a solution containing 20 mM Tris-HCl, 100 mM NaCl, 6% glycerol, and 0.5 mM EDTA (pH 7.7) and loaded onto a Superdex75 10/300 GL gel filtration column (GE Healthcare, Denmark) at 4°C. Fractions containing CNTF were pooled and concentrated to 15 to 20 g/l using VIVAspin 6 columns with 10,000-molecular-weight-cutoff (MWCO) polyethersulfone (PES) membranes from Sartorius Stedim Biotech.
Proliferation assays. BA/F3 cells were washed three times, resuspended in cytokine-free medium, and seeded in 96-well plates at 2 ϫ 10 4 cells per well. Serial dilutions of CNTF were added in triplicates, and after a 72-h incubation period, the cells were pulse-labeled with 0. Fluorescence-activated cell sorter (FACS) analysis and antibodies. Cells were stained with appropriate concentrations of phycoerythrin (PE)-conjugated human anti-gp130 and PE-conjugated human anti-LIFR␤ antibodies (R&D Systems, MN) and Alexa 488-conjugated antisortilin antibody. The antisortilin antibody was labeled with the commercially available Alexa Fluor 488 monoclonal antibody labeling kit (Invitrogen, OR) according to the manufacturer's instructions. After staining, cells were analyzed immediately with a BD FACSCalibur flow cytometer (Becton Dickinson and Company, NJ), with logarithmic recording of fluorescence and linear recording of scatter parameters. The data were analyzed by using FlowJo (version 4.6) software (TreeStar Inc., OR).
SPR. Surface plasmon resonance (SPR) analysis was performed with a Biacore 3000 instrument (Biacore, Sweden) equipped with CM5 sensor chips activated as described previously (38). Soluble sortilin and s-prosortilin were immobilized to densities of 66 to 86 fmol/mm 2 , and samples for binding (40 l) were injected at 5 l/min at 25°C in a solution containing 10 mM HEPES, 150 mM NaCl, 1.5 mM CaCl 2 , 1 mM EGTA, and 0.005% Tween 20 (pH 7.4). Binding was expressed as the difference between the response obtained from the flow cell with the immobilized receptor minus the response obtained when using an activated but uncoupled chip. The overall K d (dissociation constant) values were determined with BIAevaluation 4.1 software using a Langmuir 1:1 binding model and simultaneous fitting to all curves in the concentration range considered (global fitting).
Immunocytochemistry. Untransfected and transfected HEK293 cells were cultured on cover slides and incubated with 50 nM CNTF for 10 to 45 min in the absence and presence of 20 M NT or 9 M RAP. The cells were then washed and fixed in 4% formaldehyde (pH 7), permeabilized with 0.5% saponin (Sigma, MO), incubated with mouse antisortilin and rabbit anti-CNTF, and finally stained with Alexa 488-conjugated goat anti-mouse and Alexa 633-conjugated goat anti-rabbit antibodies (Molecular Probes, Netherlands) and mounted with Vectashield containing DAPI (4Ј,6-diamidino-2-phenylindole) (Vector Laboratories, CA). Staining was analyzed by confocal microscopy (LSM510; Carl Zeiss, Germany). The BA/F3 cells were centrifuged onto cover slides prior to fixation and incubation with antisortilin followed by staining with Alexa 488-conjugated goat anti-mouse antibody and DAPI.
Duolink. Cells were washed and fixed in 4% formaldehyde (pH 7), permeabilized with 0.5% saponin, and incubated with either mouse antisortilin and rabbit anti-LIFR␤ or mouse anti-gp130 and rabbit antisortilin overnight. The next day, the antibodies were labeled with the commercially available Duolink kit (Olink Bioscience, Sweden) according to the manufacturer's instructions. Staining was analyzed by confocal microscopy (LSM710; Carl Zeiss, Germany).

SDS-PAGE and Western blotting.
Proteins were subjected to SDS-PAGE by using a 4% stacking gel and a 4%-to-16% gradient separation gel. For Western blotting, nitrocellulose membranes (Hybond-C; Amersham Biosciences, NJ) were blocked in a solution containing 0.01 M Tris-HCl (pH 7.6), 0.15 M NaCl, 0.1% Tween 20 (TBS-T), and 5% skim-milk powder before incubation with antibodies in the same buffer. Membranes were washed in TBS-T containing 0.5% skim-milk powder. Quantification of Western blots was done by using ImageJ software.
STAT3 phosphorylation analysis. TF-1 cells were starved overnight in rhGM-CSF-free medium, while BA/F3 cells were washed three times and starved in cytokine-free medium for 4 h. TF-1 and BA/F3 cells were counted by using a Beckman Coulter Z2 analyzer (Beckman Coulter, CA) and seeded into 24-well plates at 6 ϫ 10 5 and 1.2 ϫ 10 6 cells per well, respectively. After cytokine stimulation for 15 min, cells were lysed at 4°C in 1% Triton X-100 (0.15 mM Tris-HCl, 10 mM EDTA [pH 8.0]) supplemented with a proteinase inhibitor cocktail (CompleteMini; Boehringer, Germany) and 1% phosphatase inhibitor cocktails 1 and 2 (Sigma, MO). Supernatants containing whole-cell extracts were analyzed for protein content by using a protein assay (Bio-Rad, CA) and mixed with a reducing sample buffer. The samples were subjected to immunoblot analysis with antibodies specific for STAT3 phosphorylated at Tyr705 and ␤-actin (loading control).

RESULTS
Because CNTF exhibits structural similarity to apoE (31,47) and forms heterodimeric complexes with apoE (17), we speculated whether CNTF, similar to apoE, targets sortilin for binding (22). To clarify this, we examined the binding of CNTF to the immobilized ectodomain of sortilin (s-sortilin) using SPR analysis. As demonstrated in Fig. 1A, CNTF bound ssortilin in a concentration-dependent manner and with an estimated K d of about 25 nM. The binding was completely inhibited in the presence of excess NT (Fig. 1B) or RAP (Fig.  1C), and as apparent from Fig. 1D, CNTF did not interact with the immobilized sortilin precursor construct s-prosortilin, which carries an uncleavable propeptide. This demonstrates the specificity of the binding and that CNTF targets the ␤-propeller domain of the Vps10p-D. Interestingly, CNTFR␣ did not itself interact with sortilin, and sortilin did not bind to a preformed complex of sCNTFR␣ and CNTF (Fig. 1E), signifying that CNTF is unable to bind both receptors simultaneously.
CNTF binds sortilin via a C-terminal site. We next examined the binding of CNTF to full-length constructs of sortilin in transfected HEK293 cells. The cells were incubated with 50 nM CNTF in warm (37°C) medium, and following fixation, their uptake of CNTF was determined by immunofluorescence. No staining was observed for untransfected cells (Fig.  1F). In contrast, wild-type (wt) sortilin transfectants displayed a significant, predominantly intracellular, staining signifying a considerable uptake of ligand (Fig. 1G). This uptake was almost abolished when cells were incubated in the presence of excess NT (Fig. 1H) or RAP (not shown), and a similar lack of uptake was seen for transfectants expressing prosortilin (Fig.  1J). Finally, cells expressing a mutant sortilin incapable of endocytosis due to disrupted endocytosis motifs (Y 758 A/L 761 A and L 795 L 796 ) (34) displayed staining limited to the surface membrane, indicating binding but nearly no internalization of CNTF (Fig. 1I). As shown in Fig. 2, CNTF bound to sortilin transfectants at 4°C was translocated to intracellular vesicles within 10 min of incubation at 37°C, demonstrating that sortilin mediates the rapid internalization of the ligand.
The interaction of NT with sortilin is known to be mediated by its C terminus (41). To determine if CNTF contains a similarly situated binding site for sortilin, we generated a 13amino-acid peptide (A 188 RGSHYIANNKKM 200 [C-term]) covering the C-terminal sequence of CNTF and a truncated CNTF construct (CNTF-tr) missing the corresponding segment. As determined by SPR analysis, immobilized s-sortilin did not bind the CNTF-tr construct (Fig. 3A), but the binding of full-length CNTF was completely inhibited in the presence of excess C-terminal peptide (Fig. 3B). Accordingly, HEK293 transfectants expressing wt sortilin showed no binding of CNTF-tr, and cellular uptake of full-length CNTF was absent in the presence of excess C-term peptide (Fig. 3C to E). In contrast, both CNTF and CNTF-tr bound to CNTFR␣ with a K d of 150 to 200 nM ( Fig. 3F and G).
Taken together, these data demonstrate that CNTF has a higher affinity for sortilin than for CNTFR␣, that it interacts with sortilin via a high-affinity C-terminal site that differs from its binding site for CNTFR␣, and that sortilin conveys cellular binding and endocytosis of CNTF. Sortilin facilitates CNTF-induced phosphorylation of STAT3 and MAP kinase. To determine if sortilin might influence CNTF signaling, we initially tested the human TF-1 erythroleukemia cell line, which endogenously expresses gp130 and LIFR␤ but not CNTFR␣. The cells were stably transfected with sortilin, and the surface expression of gp130 and LIFR␤, the absence of CNTFR␣, and the expression of sortilin were confirmed by FACS analysis and Western blotting (Fig. 4A). Wild-type and transfected TF-1 cells were then stimulated with CNTF at a concentration (40 nM for 15 min) that is known to induce a cellular response even in the absence of CNTFR␣ (16). In agreement, both cell lines responded with an increase in levels of phospho-STAT3 (relative to unstimulated cells). The level of phosphorylation, however, was significantly higher (2.5-fold) in the sortilin transfectants than in wt TF-1 cells ( Fig.   4B and C), suggesting that the expression of sortilin served to facilitate CNTF signaling. A similar increase in phospho-STAT3 levels was obtained for cells transfected with a sortilin mutant lacking the cytoplasmic domain, signifying that the enhanced signaling did not depend on the sortilin tail ( Fig. 4D  and E).
To confirm and elaborate on this finding, we next performed a series of experiments with the murine BA/F3 cell line, which expresses neither sortilin, gp130, LIFR␤, nor CNTFR␣. The cells were stably transfected with different combinations of these receptors (Fig. 5A and B), and their response in terms of the content of phospho-STAT3 was subsequently determined before and after stimulation with CNTF (40 nM for 15 min). As apparent from Fig. 5C, wt BA/F3 cells and cells expressing sortilin and/or gp130 showed no response to CNTF, and only a   (Fig. 5D). In agreement with these results, sortilin was also found to increase MAP kinase activation, which is an established downstream event in gp130/LIFR␤ signaling (Fig. 5E).
A time course of CNTF-mediated phospho-STAT3 induction in BA/F3-[gp130/LIFR␤/sortilin] cells is shown in Fig. 5F. It is important that the high-level response in BA/F3-[gp130/ LIFR␤/sortilin] cells compared with that in BA/F3-[gp130/ LIFR␤] cells did not appear to result from a relative increase in gp130/LIFR␤ expression levels. Thus, the simultaneous detection of STAT3 phosphorylation and the expression of sur-FIG. 2. Bound CNTF is rapidly internalized by HEK293 transfectants. Untransfected HEK293 cells (A and B) and stably transfected HEK293 cells expressing either mutant sortilin containing disrupted internalization motifs (C and D) or wt sortilin (E and F) were incubated (4°C) with 50 nM CNTF for 2 h prior to washing and reincubation in warm (37°C) medium. After 10 min the cells were fixed and stained using rabbit anti-CNTF and mouse antisortilin as primary antibodies and Alexa 488-conjugated goat anti-rabbit Ig (green) and Alexa 633-conjugated goat anti-mouse Ig (red) as secondary antibodies. Nuclei were stained with DAPI (blue). Double staining is shown in the insets.  Fig. 6A and B). Finally, CNTF induction of phospho-STAT3 was assessed in the presence of soluble CNTFR␣, which is known to promote CNTF signaling in gp130/LIFR␤-expressing cells. BA/F3-[gp130/LIFR␤] and BA/F3-[gp130/LIFR␤/sortilin] cells were therefore incubated with either CNTF (40 nM), sCNTFR␣ (4 nM), or both prior to the detection of phospho-STAT3. As expected, sCNTFR␣ had no effect on its own (not shown), whereas it strongly upregulated the response to CNTF in both cell types. However, even upon combined stimulation, the level of phospho-STAT3 remained higher in the sortilin transfectants (Fig. 6C).
Evidently, these results suggest that sortilin and CNTFR␣ have mutually independent but additive and facilitating effects on CNTF signaling.
Sortilin induces CNTF-dependent proliferation. To further substantiate that sortilin promotes the biological activity of CNTF, we examined the proliferation of various BA/F3 transfectants in response to increasing concentrations of CNTF. As apparent from Fig. 7, stimulation with as much as 4 nM CNTF resulted in little or no proliferation of wt BA/F3 cells, of transfectants expressing either sortilin or gp130, and of cells coexpressing gp130 and LIFR␤. In contrast, enhanced proliferation was already detectable in BA/F3-[gp130/LIFR␤/sortilin] cells at 0.4 nM CNTF, and at 4 nM, this response was increased by as much as 5-to 6-fold. As expected, the stimulation of BA/F3-[gp130/LIFR␤] cells in the presence of sCNTFR␣ proved even more efficient (Fig. 7, inset), but the results confirm that sortilin significantly facilitates CNTF bioactivity by a CNTFR␣-independent mechanism.
Sortilin binds CLC/CLF-1 and neuropoietin and facilitates their signaling. Apart from CNTF, CNTFR␣ is also the primary receptor for neuropoietin and the heterodimeric CLC/ CLF-1 (10, 12), and they both interact with sortilin (Fig. 8A). The affinity of CLC/CLF-1 for immobilized s-sortilin (K d of ϳ10 nM) appeared to be even more pronounced than that of CNTF, whereas neuropoietin bound with a considerably lower affinity (K d of ϳ100 nM). In each case, binding was abolished in the presence of excess NT (Fig. 8A), and as exemplified in Fig. 8B, all three ligands exhibited some degree of competition for binding. In agreement with this, HEK293 cells transfected The results confirm that the facilitating effect of sortilin is independent of CNTFR␣ and further suggest that this function is conditioned by an interaction between the respective ligands and the gp130/LIFR␤ heterodimeric complex. Also, the appar- ent lack of a connection between sortilin's affinity for a particular ligand and its ability to promote signaling by the same ligand seems to reflect that in the present context, ligand binding and the facilitation of signaling are two separate functions in sortilin. Ligand binding does not appear to be a requirement for the signal-facilitating function of sortilin. To clarify this, we first examined if the signal-facilitating effect of sortilin was abrogated in the presence of alternative ligands that target sortilin but not the gp130/LIFR␤ heterodimer. To that end, BA/F3-[gp130/LIFR␤] and BA/F3-[gp130/LIFR␤/sortilin] cells were subjected to CNTF (40 nM) treatment in the absence and presence of 40 M NT. Neurotensin completely prevents CNTF from binding to sortilin, but as depicted in Fig. 9A, it did not affect the sortilin-mediated increase in STAT3 phosphorylation and was incapable of signal induction on its own. Similar results were obtained for TF-1 cells, which endogenously express gp130/LIFR␤, and using RAP and the sortilinpropeptide instead of NT (Fig. 9B). To elaborate on this apparent paradox, we next examined phospho-STAT3 induction by the CNTF-tr construct, which binds to CNTFR␣ but not to sortilin. Findings with TF-1 (Fig. 10A) and BA/F3 (Fig. 10B) transfectants clearly demonstrated that the expression of either CNTFR␣ or sortilin in combination with gp130/LIFR␤ strongly upregulated the response to the truncated cytokine. Moreover, the sortilin binding C-terminal peptide of CNTF was unable to alter CNTF signaling in BA/F3-[gp130/LIFR␤/ sortilin] cells (Fig. 10C).
It can be concluded that sortilin promotes signaling without having to engage in ligand binding.
Sortilin promotes gp130/LIFR␤-mediated signal transduction. Given the findings described above, it seemed plausible that sortilin could promote the cellular response to other CNTF-related helical type 1 cytokines that target the gp130/ LIFR␤ dimer for signaling. We therefore tested STAT3 phosphorylation in TF-1 and BA/F3 cells stimulated with CT-1, LIF, and OSM. These ligands are considered not to bind the CNTFR␣ (20) and exhibit weak binding to s-sortilin (Table 1). However, in agreement with our assumption, cells expressing gp130/LIFR␤ responded to all three ligands, and in each case, the level of phospho-STAT3 was further enhanced upon coex-  11A and B). Considering that signal induction by hIL-6 in TF-1 cells (Fig.  11C) and in BA/F3-[gp130] cells (Fig. 11D) was unaltered upon transfection with sortilin (sortilin transfected-to-untransfected ratio of 0.9 [SD, Ϯ0.1] [n ϭ 4] for both cell lines), it appears that the facilitating effect of sortilin is restricted to the gp130/LIFR␤ heterodimer, with particular reference to the function of the LIFR␤ chain.

Sortilin and LIFR␤ interact in cells.
To elaborate on this idea, experiments were set up to provide evidence of a possible direct interaction between gp130/LIFR␤ and sortilin. SPR analysis demonstrated that while the extracellular (C-terminally tagged) domain of gp130 did not bind to immobilized s-sortilin, the corresponding domain of LIFR␤ (sLIFR␤) did (Fig. 12A). The interaction was not inhibited by a peptide (V 819 GPEKSMYVVTKENS) covering the C-terminal sequence of sLIFR␤ and therefore did not originate from the "artificial" C terminus generated by receptor truncation (not shown). Notably, even a large surplus of NT or CNTF could inhibit binding only by about 30% (not shown), which seems in good agreement with the inability of NT to abrogate sortilin's facilitating effect on signaling ( Fig. 9A and B).
As determined by the subcellular fractionation of untransfected HEK293 cells that express both LIFR␤ and sortilin, the localizations of the two receptors overlapped (Fig. 12B). Notably, both were found (in CNTF-stimulated and unstimulated cells) in fractions containing flotillin-1, a marker for lipid rafts, which was suggested previously to be a functional site in gp130/ LIFR␤ signaling (4).
Further evidence of a possible interaction between sortilin and LIFR␤ at the cellular level was next obtained by using fluorescence microscopy and Duolink, a method that visualizes interactions and/or close colocalization of single molecules. As apparent from Fig. 12C, staining with a combination of antisortilin and anti-gp130 did not provide (elicit) a detectable (fluorescence) signal in either untransfected HEK293 cells or in sortilin transfectants. In contrast, staining with antisortilin and anti-LIFR␤ generated a strong signal in the transfectants as well as a significant (although comparatively weaker) signal in untransfected cells carrying endogenous levels of both sortilin and LIFR␤. Similar results confirming the close colocalization of the two receptors were obtained with BA/F3 cells, yet attempts to cross-link and coimmunoprecipitate sortilin and LIFR␤ proved unproductive (not shown).
Taken together, the above-described results support a model in which sortilin facilitates gp130/LIFR␤-mediated signaling by interacting with LIFR␤ and, e.g., increasing its affinity for ligands and promoting the assembly of the gp130/LIFR␤ heterodimer.

DISCUSSION
It is well known that CNTF binds CNTFR␣ and subsequently recruits the gp130 and LIFR␤ receptors to form a heterotrimeric signaling complex (6,7). Here we report that a high-affinity site in the C terminus of CNTF interacts with the sortilin ␤-propeller domain and that sortilin may serve at least two functions in relation to CNTF. In the first place, it provides rapid cellular uptake and endocytosis of the ligand, and second, it facilitates the CNTF-mediated induction of gp130/ LIFR␤ signaling. Our findings further demonstrate that other ligands of CNTFR␣, notably CLC/CLF-1, also interact with sortilin and that sortilin, independent of ligand binding, facil-itates the signaling (STAT3 phosphorylation) of all helical type 1 cytokines targeting the gp130/LIFR␤ heterodimer. Based on analysis by SPR and immunofluorescence, we propose that the latter is brought about by a direct interaction between sortilin and LIFR␤.
CNTF is internalized by cellular sortilin and targets sortilin and CNTFR␣ via separate sites. The binding of CNTF to sortilin was inhibited by other sortilin ligands and completely abolished by the tridecapeptide NT and a 13-residue peptide covering the C-terminal sequence of CNTF itself. In agreement with this, truncated CNTF missing the C-terminal peptide showed no binding activity. Thus, CNTF interacts with the ␤-propeller of sortilin via a site close to and possibly incorporating its own carboxy terminus. This is very similar to the binding mode of NT (41), and in fact, preliminary data on the crystal structure of the C-term peptide in complex with sortilin indicate that NT and CNTF may target the very same site within the tunnel of the ␤-propeller (M. Hansen et al., unpublished data).
CNTF has been reported to bind CNTFR␣ via residues located in helix A, the AB loop, helix B, and the C-terminal residues of helix D (37); hence, CNTF binds sortilin and CNTFR␣ via separate binding sites. Nevertheless, CNTFR␣ completely inhibits the binding of CNTF to s-sortilin, and CNTF is unable to bind both receptors simultaneously. Considering this and the fact that CNTFR␣ itself does not bind to sortilin, it is highly unlikely that sortilin in any way partakes in the sorting of isolated CNTFR␣ or the CNTFR␣-CNTF complex. Since CNTFR␣ itself is not an endocytic receptor and the uptake via gp130/LIFR␤ concerns only the CNTFR␣-CNTF complex, there is no doubt that sortilin may contribute significantly to the uptake and clearance of extracellular CNTF (11,46,48). This may also be case with CLC/CLF-1, which binds sortilin with a similar high affinity. However, unlike CNTF, CLC/CLF-1 is secreted via the classical secretory pathway, and it can so far not be excluded that sortilin, once its propeptide has been cleaved, can target and transport this cytokine in Golgi compartments as well as at the cell surface (34,38).
Sortilin enhances CNTF signal transduction. Following the assembly of the trimeric CNTF receptor complex, information is relayed to the nucleus by a number of signaling molecules, including STAT3. The present work demonstrates that sortilin facilitates CNTF signaling as determined by STAT3 phosphorylation as well as by the proliferation of responding cells. Sortilin's contribution to signaling is clearly less significant than that of CNTFR␣, which serves to concentrate CNTF on the  cell surface and at the same time, in complex with CNTF, binds and assembles the gp130/LIFR␤. On the other hand, the effect of sortilin seemingly depends neither on the presence of CNTFR␣ nor on the binding of ligand; i.e., enhanced signaling was seen with both wt CNTF and a C-terminally truncated CNTF that does not bind sortilin. Thus, sortilin-mediated signaling appears to be conditioned by only two things: the expression of gp130/LIFR␤ and the presence of a ligand that can bind this heterodimer. It follows that sortilin cannot replace CNTFR␣ but adds to signal transduction by a separate mechanism (9). Sortilin may promote signaling by helical type 1 cytokines through an interaction with the LIFR␤. The above-described conclusion is underscored and expanded by our findings for the related cytokines CT-1, LIF, OSM, and IL-6. Thus, sortilin enhanced STAT3 phosphorylation in gp130/LIFR␤expressing cells responding to CT-1, LIF, and OSM, although all three are independent of CNTFR␣ and exhibit little or no binding to sortilin. In contrast, signaling by IL-6 in cells expressing gp130 or gp130/LIFR␤ was unaltered by the presence of sortilin.
The latter observation is particularly informative because it strongly indicates that LIFR␤, rather than gp130, is directly implicated in sortilin function. This notion was supported by our SPR analysis of the interaction between s-sortilin and the ectodomains of the two signaling receptor chains. Sortilin was previously shown to bind other transmembrane proteins (29,35), and whereas gp130 did not bind, LIFR␤ bound with medium affinity, and saturating concentrations of neither NT nor CNTF could reduce binding by more than 30%. Soluble sortilin was not seen to facilitate signaling (not shown), yet the results imply that full-length sortilin and LIFR␤ may interact on the cell membrane and thereby promote gp130/LIFR␤mediated signal transduction. In support of this, sortilin and LIFR␤ are both found in flotillin-1-containing cell fractions and exhibit distinct colocalization in cells. Thus, it is conceivable that the binding to sortilin may cause, e.g., a conformational change that increases LIFR␤'s affinity for cytokine ligands or, perhaps, even gp130. This implies that the effect of sortilin could be important in tissues with no or low levels of expression of CNTFR␣ and under conditions of low concentrations of CNTFR␣-CNTF or CNTF alone in circulation. However, our findings obviously do not allow definitive conclusions. The fact that sortilin facilitated signaling in the absence of its cytoplasmic tail suggests that its effect is related to the ectodomain and/or transmembrane domain and events on the plasma membrane, but different/alternative mechanisms, including receptor translocation, may be involved, and although it seems unlikely, it cannot be entirely excluded that changes due to transfection may play a role.
In conclusion, we demonstrate that sortilin mediates the cellular uptake of CNTF and related helical type 1 cytokines targeting CNTFR␣, in addition to being a facilitator of cytokines that signal through the gp130/LIFR␤ heterodimer. The latter function is independent of both CNTFR␣ and ligand binding to sortilin and seems to implicate a direct interaction with LIFR␤. Even though the contribution from sortilin is perhaps modest, it is nonetheless clear, and it may implicate