Cytokine-Like Factor 1, an Essential Facilitator of Cardiotrophin-Like Cytokine:Ciliary Neurotrophic Factor Receptor α Signaling and sorLA-Mediated Turnover

Cardiotrophin-like cytokine:cytokine-like factor-1 (CLC:CLF-1) is a heterodimeric neurotropic cytokine that plays a crucial role during neuronal development. Mice lacking CLC:CLF-1 die soon after birth due to a suckling defect and show reduced numbers of motor neurons. Humans carrying mutations in CLC:CLF-1 develop similar disorders, known as Sohar-Crisponi or cold-induced sweating syndrome, and have a high risk of early death. It is well known that CLC binds the ciliary neurotrophic factor receptor α (CNTFRα) and is a prerequisite for signaling through the gp130/leukemia inhibitory factor receptor β (LIFRβ) heterodimer, whereas CLF-1 serves to promote the cellular release of CLC. However, the precise role of CLF-1 is unclear. Here, we report that CLF-1, based on its binding site for CLC and on two additional and independent sites for CNTFRα and sorLA, is a key player in CLC and CNTFRα signaling and turnover. The site for CNTFRα enables CLF-1 to promote CLC:CNTFRα complex formation and signaling. The second site establishes a link between the endocytic receptor sorLA and the tripartite CLC:CLF-1:CNTFRα complex and allows sorLA to downregulate the CNTFRα pool in stimulated cells. Finally, sorLA may bind and concentrate the tripartite soluble CLC:CLF-1:CNTFRα complex on cell membranes and thus facilitate its signaling through gp130/LIFRβ.

T he heterodimeric cytokine cardiotrophin-like cytokine:cytokine-like factor-1 (CLC:CLF-1) is a member of the human interleukin-6 (IL-6) family of cytokines, which also includes ciliary neurotrophic factor (CNTF), cardiotrophin, leukemia inhibitory factor (LIF), oncostatin M, IL-6, IL-11, and IL-31 (1). The family members are structurally related, and following binding to separate but mutually related (cytokine type 1) membrane receptors, they recruit the transmembrane glycoprotein 130 (gp130) and, in most cases, the LIF receptor ␤ (LIFR␤), for signal transduction (1,2). CLC:CLF-1 is composed of two independently expressed proteins, CLC (alias, novel neurotrophin-1 or B cell-stimulating factor-3) and CLF-1 (alias, cytokine receptor-like factor 1) (3)(4)(5). Mature CLC consists of 198 amino acids with ϳ28% sequence identity to CNTF and forms a four-helical bundle. It is synthesized with a signal peptide and is expressed in the biosynthetic pathway of a variety of cell types but is poorly secreted. Like CNTF, CLC binds to the CNTF receptor ␣ (CNTFR␣), either the glycosylphosphatidylinositol (GPI)-anchored or the soluble form, and elicits signaling via interaction with gp130/LIFR␤ heterodimers and subsequent activation of the Janus kinase (JAK) and the signal transducer and activator of transcription-3 (STAT3) pathway (6). Human CLF-1 is a glycosylated soluble 385-amino-acid protein with significant similarities to cytokine type 1 receptors like the IL-6 receptor ␣ (IL-6R␣). On its own, CLF-1 does not appear to induce signaling, but several lines of evidence have established that it plays a crucial role in the cellular and physiological functions of CLC by binding CLC in a stable complex (4,7,8). When coexpressed, CLF-1 and CLC interact in the biosynthetic pathway, and CLF-1 serves to facilitate rapid transport and secretion of the complex (4). Alternatively, CLF-1 may target free CLC released from cells or bound to plasma membranes via the CNTFR␣. In any case, the CLC:CLF-1 heterodimer binds CNTFR␣, and compared to CLC:CNTFR␣, the tripartite complex significantly enhances gp130/LIFR␤ activation and STAT3 phosphorylation in cells (4,6). Thus, CLF-1 promotes not only the cellular secretion of CLC but also its induction of transmembrane signaling.
The close functional connection between CNTFR␣, CLC, and CLF-1 is strongly reflected by findings in vivo showing that mice deficient in any of the three proteins display the same phenotype (7,9,10). The mice all fail to suckle, die soon after birth, and show reduced numbers of motor neurons in the facial nucleus and lumbar ventral horns (7,(9)(10)(11). Moreover, human patients carrying mutations (single homozygotes or compound heterozygotes) in CLC or CLF-1 (8,12,13) develop a series of similar manifestations known as Sohar-Crisponi or cold-induced sweating syndrome (CISS) (14)(15)(16), which includes feeding difficulties and risk of early death. However, despite the apparent overlap between CLC and CLF-1 functions, the precise role of CLF-1 in the tripartite complex with CNTFR␣ and CLC is insufficiently understood, and recent findings even suggest that CLF-1 may have separate functions and alternative partners (17,18). Thus, CLF-1 has been reported to complex with other cytokine components, and it is still an open question if it (on its own) may target hitherto unidentified receptors (19).
In humans CNTFR␣ is the common primary receptor for CNTF and CLC:CLF-1, and we have previously reported that both ligands also bind the membrane receptor sortilin, which mediates their uptake and appears to promote their induction of gp130/ LIFR␤ and JAK/STAT3 signaling in transfected cells (20). Preliminary unpublished findings further indicate that CLC:CLF-1, but not CNTF, engages the sortilin-related multiligand receptor sorLA, which is widely expressed both inside and outside the nervous system.
Along with sorCS1 to sorCS3, sortilin and sorLA constitute the Vps10p-domain (Vps10p-D) family of type 1 membrane receptors (21). The family hallmark is the N-terminal Vps10p-D comprising a unique neuropeptide and protein-binding 10-bladed ␤-propeller supported by two minor domains (22). The Vps10p-D constitutes the entire luminal part of sortilin, whereas sorLA contains a series of other structural elements, notably, a small ␤-propeller domain with an associated epidermal growth factor class B-like motif, followed by a cluster of 11 low-density lipoprotein receptor (LDLR) class A repeats (23,24). Similar structural elements are typical for members of the LDLR family, and accordingly ligands like receptor-associated protein (RAP), platelet-derived growth factor bb, lipoprotein lipase, apolipoprotein E, and proteins of the plasminogen activator system that target members of this family also target sorLA (25,26). The cytoplasmic tail of sorLA is short but contains several functional motifs for binding of cytoplasmic adaptor proteins (e.g., AP-1 and -2, GGA1 to GGA3, and elements of the retromer complex) involved in receptor trafficking (27)(28)(29). Accordingly, sorLA conveys rapid endocytosis and cellular uptake of ligands bound at the surface membrane as well as transport between trans-Golgi compartments and endosomes of newly synthesized protein bound in the biosynthetic pathway (27,30). Interestingly, sorLA targets not only soluble ligands but also membrane-associated and transmembrane proteins. It is well documented that complex formation with the amyloid precursor protein (APP) affects APP processing and the generation of A␤-amyloid peptide, and recent findings provide evidence that sorLA's interaction with glia cell line-derived neurotrophic factor (GDNF) and its primary receptor, GDNF family receptor ␣, may impact receptor tyrosine kinase RET-mediated signaling in vivo (31,32). It appears that sorLA is a multiligand receptor displaying a diversity of cellular functions which, depending on the local expression pattern of ligands and receptors, may involve tissues both inside and outside the nervous system.
In the present report we have examined interactions between the elements of the tripartite complex CLC:CLF-1:CNTFR␣, the capacity of the individual subunits for binding to sorLA, and the functional implications of these interactions in neuronal and nonneuronal cells. Our findings reveal that CLF-1 promotes CLC's association with CNTFR␣, that CLF-1, in complex or on its own, exhibits a high-affinity site for binding to sorLA, and that its binding to sorLA has a profound effect on the turnover and signaling of CLC:CNTFR␣.
cDNA construct, protein expression, and purification. Human sorLA glutathione S-transferase (GST)-propeptide and RAP were expressed in Escherichia coli and purified as previously described (26,33). Generated human cDNA constructs encoding soluble sorLA and sorLA Vps10p-D (26) were inserted into pcDNA3.1 expression vectors and expressed in eukaryotic Chinese hamster ovary (CHO) cells as described previously (26). For generation of CNTFR␣-Myc, a Myc-tag was inserted between the signal peptide and the gene coding for the mature human CNTFR␣ protein. The cDNA fragment was inserted in the pcDNA3.1/ hyg(Ϫ) expression vector and synthesized by MWG Biotech. Human CLF-1 cDNA was purchased from ImaGenes. CLF-1-coding cDNA was amplified by PCR, and a 6ϫHis tag was added at the C terminus using the forward primer 5=-ACA ACA TCT AGA ATG GCC GCC GGC CGC CGG GGC CCC-3= and the reverse primer 5=-CGC TCT CCA GGA CGG TCT GTA GTG GTA GTG GTA GTG ATT GAG CTC CAC-3=. The PCRamplified CLF-1 was cleaved with XbaI and XhoI (New England BioLabs) and ligated into pcDNA3.1/zeo(Ϫ). For CLF-1-His expression and purification, HEK293 cells transfected with CLF-1-His were grown in Dulbecco modified Eagle's medium (DMEM) (Lonza) until cells were confluent. Afterwards, cells were changed to CCM5 serum free medium (HyClone; Thermo Scientific) for approximately 5 days (until cells started to fall off). The medium was harvested and centrifuged for 30 min at 4,000 ϫ g at 4°C. The supernatant was collected and preserved in 0.5% NaN 3 . The spun medium was prepared for purification by the addition of 50 mM Na 2 HPO 4 , 0.3 M NaCl, 10 nM imidazole, and 0.02% Triton X-100 and adjusted to pH 7. The purification was done overnight on a Talon His tag purification resin (Clontech)-prepared XK column (Amersham Biosciences), with a total column volume of 13 ml at 4°C. Before purification from medium, the column was washed in 50 mM Na 2 HPO 4 and 300 mM NaCl, pH 7.0. The column was then eluted in 50 mM sodium acetate (NaOAc) and 300 mM NaCl, pH 5.0, and collected in 1-ml fractions, where 3 M Tris-HCl, pH 8.8, was added to raise the pH. The concentration of fractions containing CLF-1-His was determined by absorption at 280 nm, and fractions were concentrated on a Centriprep YM-10 instrument (Millipore) to a concentration of 1 mg/ml. Cell lines and transfection. Human embryonic kidney (HEK293) cells were cultured in DMEM supplemented with 10% fetal bovine serum (FBS), 100 U/ml penicillin, and 100 g/ml streptomycin. HEK293 cells were transfected with pcDNA3.1/zeo(Ϫ) and/or pcDNA3.1/hyg(Ϫ) using FuGENE 6 (Roche), and stably selected clones were selected in medium containing 150 g/ml zeocin and/or 500 g/ml hygromycin B. SH-SY5Y cells were cultured in DMEM-F-12 medium supplemented with 10% FBS, 100 U/ml penicillin, and 100 g/ml streptomycin and transfected with pcDNA3.1/zeo(Ϫ) using FuGENE 6 (Roche). Stably selected clones were selected in medium containing 300 g/ml zeocin. Murine bone marrow-derived pro-B cells (Ba/F3) containing gp130 and LIFR␤ (34), were grown in DMEM supplemented with 10% FBS, 100 U/ml penicillin, 100 g/ml streptomycin, and 10 ng/ml LIF. The cells were transfected with sorLA by electroporation as described previously (20). Stably selected clones were selected in medium containing 150 g/ml zeocin and 10 ng/ml LIF.
Immunocytochemistry and quantitative analysis. Untransfected and sorLA transfected HEK293 cells were cultured on cover slides and incubated with 10 nM CLC, CLF-1, or CLC:CLF-1 for 25 min at 37°C. The cells were then washed in phosphate-buffered saline (PBS), pH 7.4, fixed in 4% paraformaldehyde (PFA), pH 7, and permeabilized with 0.5% saponin (Sigma). The cells were then incubated with rabbit anti-CLC and/or mouse anti-His and/or mouse anti-sorLA and stained with the matching Alexa Fluor-conjugated secondary antibodies. In some experiments cells were incubated with 0.5 g/ml bisbenzimide (Hoechst; Sigma) for nucleus staining. All stainings were analyzed by confocal microscopy (LSM710 or LSM780 instrument; Carl Zeiss).
Murine wild-type (wt) and sorLA knockout (KO) (32) astrocytes and hippocampal neurons (all procedures involving animal objects were performed in compliance with Danish and European regulations, and breeding was approved by The Danish Animal Experiments Inspectorate [permit number 2012-15-2935-00007]) were cultured on cover slides and incubated with or without 40 nM CLC:CLF-1 (containing a His tag) for 25 min at 37°C and then washed in PBS, fixed in 4% PFA, permeabilized with 0.5% saponin or Triton X-100, and blocked with FBS. The cells were then incubated with mouse anti-His and rabbit anti-sorLA and stained with appropriate conjugated antibodies.
Untransfected and sorLA transfected SH-SY5Y cells were cultured on cover slides and incubated with or without 10 nM CLC or CLC:CLF-1 (containing a His tag) for 25 min at 37°C. Cells were then washed in PBS, fixed in 4% PFA, permeabilized with 0.5% saponin, and labeled with mouse anti-His and/or goat anti-CNTFR␣ and/or mouse anti-sorLA, followed by staining with the matching secondary antibodies.
HEK293 cells stably transfected with CNTFR␣-Myc or sorLA and CNTFR␣-Myc in combination were labeled with mouse anti-Myc for 10 min at 37°C, washed, and then incubated for 0 or 15 min at 37°C in medium with or without 10 nM CLC or CLC:CLF-1. Subsequently, the cells were washed, fixed, and permeabilized, followed by staining using the appropriate secondary antibody.
Quantitative analysis of SH-SY5Y cells was performed on an Olympus ScanR imaging station as described in Klinger et al. (30). In short, all images were acquired with a 40ϫ dry objective, a triple band filter for Hoechst, Alexa Fluor 488, and Alexa Fluor 568, and a Hamamatsu (C8484-05G) camera. The analysis was done on z-stack (3-stack) projected images that were background subtracted. An edge detection algorithm was used for segmentation of nuclei and CNTFR␣-stained vesicles. Objects were independently selected using only the criteria that they should be able to be encircled in a closed circle (edge) of 3 m in diameter. The enclosed circles were counted as CNTFR␣ vesicles. The mean number of CNTFR␣ vesicles was determined by calculating the ratio of vesicles to nuclei.
SPR. Surface plasmon resonance (SPR) analysis was performed on a Biacore 3000 instrument (Biacore, Sweden) equipped with CM5 sensor chips activated as described previously (35). The entire luminal part of sorLA (soluble sorLA), the sorLA Vps10p-D, and sCNTFR␣ 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 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 immobilized receptor and the response obtained when an activated but uncoupled chip was used. The overall K d (dissociation constant) values were determined by BIAevaluation, version 4.1, software using a Langmuir 1:1 binding model and simultaneous fitting to all curves in the concentration range considered (global fitting).
Cross-linking and coimmunoprecipitation. Transfected HEK293 cells were washed with PBS, followed by cross-linking in 45 min with 2 nM dithiobis(succinimidyl propionate) (DSP). The reaction was stopped at 100 mM Tris, pH 7.5, and cells were washed twice with PBS. Cells were subsequently lysed at 4°C in 1% Triton X-100 lysis buffer (20 mM Tris-HCl, 10 mM EDTA, pH 8.0) supplemented with proteinase inhibitor cocktail (Complete Mini; Roche), and precipitations were performed using rabbit anti-sorLA and mouse anti-Myc antibodies. The samples were subjected to immunoblot analysis using mouse anti-sorLA and goat anti-CNTFR␣ antibodies.
Downregulation of CNTFR␣. HEK293 cells stably transfected with CNTFR␣-Myc or sorLA and CNTFR␣-Myc in combination were incubated (5 h at 37°C) in medium with or without 10 nM CLC:CLF-1. Medium was recovered, and the cells were lysed in 1% Triton X-100 buffer as described above. Supernatants containing whole-cell extracts were analyzed for protein content using a protein assay (Bio-Rad), and the samples were subjected to immunoblot analysis with antibodies specific for sorLA, ␤-actin, and Myc tag.
For inhibition of lysosomal hydrolases, the cells were preincubated at 50 g/ml leupeptin and pepstatin A (replaced every 6 h) for 24 h prior to continued incubation (37°C) in the presence of 10 nM CLC:CLF-1. The cells were subsequently either lysed as described above or fixed in PFA.
Analysis of STAT3 phosphorylation. Ba/F3 cells were starved for 3 h in cytokine-free medium prior to ligand stimulation. The cells were counted and seeded in 24-well plates with 1,200,000 cells per well. HEK293 and SH-SY5Y cells were seeded in 24-well plates and starved for 3 h prior to stimulation. Cells were incubated (15 min at 37°C) with or without CLC, CLF-1, CLC:CLF-1, CNTF, or CLC-sCNTFR␣ alone or with CLC:CLF-1 and sCNTFR␣ in combination at a concentration of 5 or 10 nM. Cells were lysed at 4°C in 1% Triton X-100 lysis buffer, as described above, supplemented with a phosphatase inhibitor cocktail (PhosSTOP; Roche). Supernatants containing whole-cell extracts were analyzed for protein content, and the samples were subjected to immunoblot analysis with antibodies specific for STAT3, pSTAT3, and ␤-actin.

RESULTS
CLC:CLF-1 targets sorLA. We have previously reported that the Vps10p-D receptor sortilin binds the two CNTFR␣ ligands CNTF and CLC:CLF-1 (20). Since sortilin and sorLA are members of the same receptor family and are known to share several ligands, we initially examined whether sorLA also exhibited affinity for the two ligands. We found that sorLA, unlike sortilin, did not bind to CNTF, whereas the heterodimeric CLC:CLF-1 bound to sCNTFR␣ as well as to the ectodomain of sorLA ( Fig. 1A and C). Analysis by SPR demonstrated that binding to immobilized soluble sorLA, i.e., the luminal part of full-length sorLA, exhibited high affinity (K d of 5 Ϯ 2.9 nM; n ϭ 3), as determined by binding at different concentrations of CLC:CLF-1 (Fig. 1C). The binding was completely abolished in the presence of a surplus of sorLA's own propeptide, demonstrating that CLC:CLF-1 selectively targets the Vps10p-D ␤-propeller (Fig. 1D). Binding was also efficiently inhibited by receptor-associated protein (RAP) but almost unchanged by neurotensin (NT) (data not shown).
sorLA selectively binds CLF-1, while CNTFR␣ binds both CLF-1 and CLC. The affinity of CLC:CLF-1 for immobilized sCNTFR␣ was even higher, approximately 10-fold, than its affinity for sorLA, and the K d was estimated to be 0.3 Ϯ 0.3 nM (mean Ϯ standard deviation [SD]; n ϭ 3) (Fig. 1A). To determine if one or possibly both elements of the heterodimer contribute to receptor binding, we first examined the activity of CLC. It appears (Fig. 1E) that CLC on its own was incapable of engaging sorLA and that its affinity for sCNTFR␣ was surprisingly modest (Fig. 1B), having a K d of 66 Ϯ 39 nM (mean Ϯ SD; n ϭ 3), which is more than a 100-fold lower than that of the CLC:CLF-1 complex (Fig.  1A). Since CLC and the CLC:CLF-1 heterodimer were prepared from different expression systems (bacteria and insect cells, respectively), a direct comparison between the estimated K d values should be considered with some caution. However, similar results were obtained with CLC from several different batches, strongly indicating that CLF-1 could be a key player in the binding of CLC:CLF-1 to each of the two receptors. To test CLF-1 itself, we used a His-tagged CLF-1 construct purified from the medium of stably transfected CHO cells (Fig. 1B, inset). The subsequent SPR experiments showed that CLF-1 binds to sorLA with an affinity (K d ϭ 4 Ϯ 0.8 nM; n ϭ 3) (Fig. 1E) very similar to that of the CLC:CLF-1 heterodimer (Fig. 1C) and that the interaction was inhibited by the sorLA propeptide (Fig. 1F) as well as by RAP (data not shown), whereas inhibition by NT was negligible (data not shown). It can be concluded that CLC:CLF-1 targets the Vps10p-D in sorLA by a single site residing in the CLF-1 subunit. Notably, CLF-1 also bound sCNTFR␣ (K d ϭ 4 Ϯ 1.3 nM; n ϭ 3) (Fig. 1B) and, seemingly, with an affinity significantly higher than that of CLC. The latter finding suggests that the comparatively much higher affinity exhibited by the assembled CLC:CLF-1 heterodimer is owed to a bonus effect generated by two "minor" binding sites, one in each subunit.
We next examined CLF-1's contribution to CLC-and CNTFR␣-mediated signaling. To that effect, CNTFR␣-expressing SH-SY5Y cells were stimulated for 15 min with a 10 nM concentration of either CLC, CLF-1, or the CLC:CLF-1 complex, and the magnitude of induced signaling was subsequently measured by determining the resulting level of phosphorylated STAT3 (pSTAT3). The results (Fig. 1G and H) showed that whereas CLC: CLF-1 induced a strong signal, very similar to the response to 10 nM CNTF, stimulation with CLC resulted in only a modest increase in pSTAT3, and the pSTAT3 level upon CLF-1 stimulation did not differ from that of unstimulated controls.
Thus, CLF-1's role in signaling seems to reflect its contribution to the binding between CLC and CNTFR␣. In other words, via complex formation with CLF-1, CLC gains a higher affinity for CNTFR␣ which enhances its accumulation on the cell membrane and formation of the tripartite signaling complex (CLC:CLF-1: CNTFR␣) and, hence, its ability to promote engagement of gp130/LIFR␤ for signaling.
CNTFR␣ and sorLA connect via CLC:CLF-1. An additional SPR experiment was set up to determine if the CLC:CLF-1 heterodimer could in fact interact with sorLA and CNTFR␣ simultaneously. To that end, we first tested binding of a complex of sCNTFR␣ and CLC, in which the two were connected with a covalent linker peptide to prevent dissociation (CLC:sCNTFR␣). As demonstrated by the results shown in Fig. 2A, this construct exhibited no binding activity toward immobilized sorLA, while a significant binding of the construct was seen after pretreatment of sorLA with CLF-1 (Fig. 2B). It can be concluded that CLC:CLF-1 may target both receptors at the same time and thereby constitute a connecting bridge between the two. sorLA mediates cellular uptake of CLF-1 and CLC:CLF-1. Binding to full-length sorLA was studied in cells. Wild-type HEK293 cells and cells stably transfected with sorLA were incubated at 37°C in medium containing a 10 nM concentration of either CLC, His-tagged CLF-1, or CLC:CLF-1. After 25 min the cells were fixed, permeabilized, and stained with anti-CLC or anti-His tag antibodies to visualize internalized ligands. No staining was observed in untransfected cells, but in good agreement with the SPR results, CLF-1 and CLC:CLF-1, but not CLC, had been taken up (and internalized) by the sorLA transfectants (Fig. 3A). Double staining (for CLC:CLF-1 and sorLA) of cells from similar experiments demonstrated that the internalized ligand colocalized with sorLA (Fig. 3B). Similar results were obtained in sorLA transfected SH-SY5Y cells (data not shown), but in these cells, which express endogenous CNTFR␣, internalized CLC:CLF-1 was furthermore seen to colocalize with CNTFR␣ (Fig. 3C).
Uptake of CLC:CLF-1 was confirmed in astrocytes, which exhibit endogenous expression of sorLA but little or no CNTFR␣ as determined by Western blotting and lack of response to CLC: CLF-1 stimulation (data not shown). Cultured astrocytes isolated from wt and sorLA-deficient (KO) mice were incubated (25 min) in the absence or presence of 40 nM His-tagged CLC:CLF-1 prior to staining with anti-His and anti-sorLA antibodies. As demon-strated in Fig. 4A, sorLA-deficient cells showed little or no uptake of CLC:CLF-1, whereas wt astrocytes presented a significant vesicular uptake, which to a large extent colocalized with sorLA. Thus, as determined by manual counting, wt astrocytes contained 11.44 Ϯ 4.02 (mean Ϯ SD; n ϭ 16) CLC:CLF-1-positive vesicles per cell in contrast to the 1.25 Ϯ 1.29 (n ϭ 16) positive vesicles per cell found in sorLA-deficient astrocytes. A similar accumulation of ligand colocalizing with sorLA was observed in isolated wt hippocampal (mouse) neurons expressing sorLA but not in sorLAdeficient hippocampal neurons (Fig. 4B).
sorLA mediates internalization of CNTFR␣ via binding to CLC:CLF-1. The SPR data (Fig. 2B) and the findings in SH-SY5Y cells suggested that the CLC:CLF-1 heterodimer might serve as a molecular link between sorLA and CNTFR␣, and this prompted us to examine if sorLA, via binding to CLC:CLF-1, also conveyed uptake of CNTFR␣. HEK293 cells expressing CNTFR␣-Myc alone or both receptors in combination were fixed and stained with an anti-Myc antibody before and after incubation (37°C) in unsupplemented medium or in medium supplemented with 10 nM CLC or CLC:CLF-1. It appears from the subsequent fluorescence microscopy (Fig. 5A) that in the absence of sorLA nearly all CNTF receptors were localized to the surface membrane both before and after incubation with ligands. In contrast, distinct intracellular staining, signifying internalized CNTFR␣, was seen in the double transfectants. Staining was particularly strong in the cells treated with CLC:CLF-1, but minor staining was also seen upon stimulation with CLC and, surprisingly, in cells not subjected to either of the two ligands.
Similar results were obtained by automated counting of CNTFR␣-positive vesicles in sorLA transfected and in untransfected SH-SY5Y cells, which constitutively express CNTFR␣ ( Fig.  5B and C).
As shown in Fig. 5B, internalization of CNTFR␣ in the absence of ligands was more pronounced in sorLA transfected than in wt SH-SY5Y cells, and this difference was strongly enhanced in the presence of CLC:CLF-1. However, in cells subjected to medium containing CLC, sorLA's promoting effect on internalization of CNTFR␣ appeared to be eliminated or, compared to internalization in sorLA transfectants not exposed to ligands, even reversed (Fig. 5C). sorLA and CNTFR␣ may interact directly. Whereas the former result confirmed that CLC:CLF-1 may bind both receptors simultaneously and thus promote sorLA-mediated endocytosis of CNTFR␣, the latter seemed to suggest that the mere presence of sorLA was enough to facilitate a significant degree of endocytosis and that this effect was blocked by CLC, which binds CNTFR␣ but not sorLA (Fig. 1E). To clarify this, we first performed SPR analysis to determine if the ectodomains of the two receptors interacted. Soluble CNTFR␣ bound to immobilized sorLA and competed with RAP for binding ( Fig. 6A and B). RAP interacts with both the complement-type repeats and the Vps10p-D of sorLA, but as further analysis demonstrated specific binding of sCNTFR␣ to the purified Vps10p-D (Fig. 6C), this domain must harbor the binding site. Since sCNTFR␣ in complex with CLC did not associated with sorLA ( Fig. 2A), it can be deduced that the two receptors can interact directly in the absence of ligands or interact indirectly (by proxy) in the presence of CLC:CLF-1 and that CLC, when complexed to CNTFR␣, inhibits/blocks the interaction.
CNTFR␣ forms a complex with full-length sorLA in cells. Direct interaction between full-length receptors in cells was next pursued by cross-linking. HEK293 cells expressing either CNTFR␣-Myc or sorLA or both in combination were incubated with the chem-ical cross-linker DSP at room temperature. After 15 min, the reaction was stopped by addition of Tris. The cells were then washed and dissolved in lysis buffer, and following removal of debris by centrifugation, supernatants were subjected to immunoprecipitation using anti-Myc and anti-sorLA Ig and finally analyzed by native and reducing SDS-PAGE (Fig. 6D). The analysis showed that only one band, representing the expressed receptor, could be precipitated from single transfectants and only by the relevant receptor-specific antibody (data not shown). In double transfectants, however, both antibodies precipitated high-molecular-weight adducts which, in the presence of dithiothreitol (DTT) (cleaving the cross-linker), resolved in two bands representing CNTFR␣ and sorLA. In view of this result, the above SPR analysis, and the findings using fluorescence microscopy (Fig. 5C), we conclude that sorLA and the CNTFR␣ interact on cell membranes.
Following binding of CLC:CLF-1, CNTFR␣ is downregulated by sorLA. To determine if sorLA-mediated internalization of the CLC:CLF-1:CNTFR␣ complex was accompanied by downregulation (degradation) of CNTFR␣, we initially tried to detect downregulation using biolabeled receptors. However, as labeling was poor due to a very modest synthesis of CNTFR␣ at steady state, we instead decided to measure the total pool of CNTFR␣ in HEK293 transfectants before and after incubation with CLC: CLF-1. Single transfectants expressing CNTFR␣-Myc and double transfectants expressing both CNTFR␣-Myc and sorLA were incubated at 10 nM CLC:CLF-1 or in unsupplemented medium. After 0 and 5 h of incubation, the cells were lysed, and their content of CNTFR␣ was quantified by Western blotting. The results are shown in Fig. 7, and it appears (Fig. 7B) that in the double transfectants the cellular pool of CNTFR␣ was reduced by more than 50% upon 5 h of exposure to CLC:CLF-1. In contrast, the bulk of CNTFR␣ in the single transfectants (subjected to CLC: CLF-1 or not) remained virtually unchanged (showing a slight but insignificant increase) upon incubation with CLC:CLF-1 (Fig.  7A). Furthermore, downregulation of CNTFR␣ was almost absent in cells treated with the lysosomal inhibitors leupeptin and pepstatin A (Fig. 7C) and was instead seen to accumulate in LAMP-1-positive vesicles (Fig. 7D). Thus, our data demonstrate that CNTFR␣, following endocytosis in complex with CLC:CLF-1 and sorLA, is directed toward degradation in lysosomes rather than recycled to the surface membrane. In accordance with the above (downregulation of CNTFR␣), we found that sorLA lowered the response to renewed stimulation in cells that had been preexposed to CLC:CLF-1. This was seen in wt HEK293 cells, which have modest endogenous expression levels of both CNTFR␣ and sorLA. The cells were first preincubated in the absence or presence of CLC:CLF-1 (5 h, 10 nM), then washed and incubated in unsupplemented medium (1.5 h), and finally restimulated with CLC:CLF-1 (5 nM for 15 min). Signal induction, in terms of pSTAT3, was subsequently measured by Western blotting of cell lysates. The results showed that cells which had been preincubated with CLC:CLF-1 presented a much weaker response to restimulation than control cells preincubated in blank medium. Thus, in three experiments, the level of pSTAT3 in preexposed cells was only 57.17% Ϯ 9.51% (mean Ϯ standard error of the mean [SEM]) of that in control cells (Fig. 7E, left  panel), and, notably, the pSTAT3 level in transfected HEK293 cells overexpressing sorLA (and tested in parallel with the wt cells) was even lower than that in wt cells (although not significantly lower; P ϭ 0.13), constituting as little as 36.83% Ϯ 5.35% (mean Ϯ SEM) of the level in controls (Fig. 7E, right panels).
sorLA may facilitate signaling in cells not expressing CNTFR␣. To determine if sorLA impacts cellular signaling via its ability to interact with CLC:CLF-1 bound to CNTFR␣, we first examined STAT3 phosphorylation in HEK293 cells with endoge-nous expression of CNTFR␣. The cells were untransfected or transfected with sorLA and were stimulated for 15 min with either 5 nM CLC:CLF-1 or a 5 nM concentration of the fusion protein CLC-sCNTFR␣. The pSTAT3 level was measured by Western blotting, and the results are shown (Fig. 8A) relative to the level obtained in untransfected HEK293 cells responding to CLC: CLF-1. As can be seen, expression of sorLA did not significantly alter the response. A similar experiment was next performed in CNTFR␣-deficient Ba/F3 cells (expressing gp130/LIFR␤), but in this case stimulation with CLC:CLF-1 was performed in the presence of sCNTFR␣ (Fig. 8B). Again the responses to stimulation with the fusion protein, which does not bind sorLA, did not differ significantly between untransfected cells and sorLA transfectants. However, in cells subjected to CLC:CLF-1 and sCNTFR␣, the resulting level of pSTAT3 was about 4-fold higher in the sorLA transfectants than in the untransfected Ba/F3 cells. Taken together, the two experiments strongly indicate that sorLA may serve to bind and concentrate circulating tripartite complexes of CLC:CLF-1 and sCNTFR␣ on the cell membrane and thereby promote their interaction with gp130/LIFR␤ and signaling. This effect appears to be insignificant on cells that express CNTFR␣ but may have a significant effect on cells without endogenous expression of CNTFR␣.
CLC and CLF-1 are (co)expressed in a large number of cell types both inside and outside the nervous system (39), and, like CNTF, the CLC:CLF-1 heterodimer as well as the individual CLC subunit promotes growth and survival of CNTFR␣-expressing neuronal cells (5,11,40). In the presence of CNTFR␣ (membrane anchored or soluble), CLC is sufficient to induce signaling, whereas CLF-1, as exemplified in the results shown in Fig. 1G and H, has no independent capacity for signal induction (41).
Similar observations have been made in humans. Thus, apart from an insignificant (and seemingly harmless) decrease in motor neurons in old age, lack of CNTF appears to be innocuous (43), while compound mutations causing insufficient expression of CLC and/or CLF-1 result in a series of defects manifested in the form of human CISS/Sohar-Crisponi syndrome (8,(12)(13)(14)(15)(16)44). Some of the cardinal symptoms in CISS, e.g., the risk of early death (originating from motor neuron defects and the resulting inability to suckle) is very similar to findings in the KO animals.
As demonstrated by results in both humans and mouse, each of the three subunits in the signaling tripartite complex CLC:CLF-1:CNTFR␣ is indispensable in vivo; i.e., a functional deficiency in either one of the three subunits is enough to cause neonatal death in mice and full-blown CISS in humans. With regard to CLC and CNTFR␣ this is not surprising, as both are needed for interaction with gp130/LIFR␤ and the subsequent signaling events. However, CLF-1, alone or in combination with CNTFR␣, has no direct impact on signaling, and so far its only know function is to facilitate transport and secretion of CLC. As previously described, CLC is expressed in the biosynthetic pathway but seems to depend on coexpression and complex formation with CLF-1 (or possibly CNTFR␣) for cellular release (4,39,45). The clinical manifestations of CLF-1 deficiency have therefore been ascribed to the resulting decrease in CLC secretion. However, a recent study (41) has provided compelling evidence for a more complex function. Thus, it appears that some CLF-1 mutations may cause disease without seriously affecting the cellular secretion of CLC, which implies that CLF-1 is more than just a facilitator of CLC secretion.
CLF-1 is as a key player in CLC:CNTFR␣ signaling and turnover. In this study, we show that, apart from facilitating the cellular release of CLC, CLF-1 has at least two other functions, i.e., one that has to do with CLC:CNTFR␣ complex formation and signal-  ing and one that implicates a new receptor with a significant impact on the turnover of CLF-1 itself as well as of its binding partners CLC and CNTFR␣.
CLF-1 sustains CLC binding to CNTFR␣ and thereby promotes CLC signaling. Despite the fact that CLC, unlike CLF-1, is a prerequisite for the recruitment of gp130/LIFR␤ and signaling (41), the CLC:CLF-1 complex is a much stronger inducer of STAT3 phosphorylation than the individual CLC subunit. The reason for this is revealed by the present binding experiments which show that, compared to CLC:CLF-1, CLC binds poorly to CNTFR␣. Thus, CLC:CLF-1 binds with a K d which is about a 100-fold lower than that of CLC, and in effect CLC seems to gain a much higher affinity via complex formation with CLF-1. Since CLC and CLF-1 (as determined by gel filtration) tend to form polymers and even aggregates and as CLC:CLF-1 may dissociate into subunits, the reported affinities should, of course, be considered with some caution; the observed gain of affinity might, for instance, originate from a more stable and optimal conformation of CLC upon binding to CLF-1. Another factor that might explain or contribute to the shift in the K d is, as previously mentioned, the different expression systems used to generate the two reagents. Thus, the possibility that CLC generated in bacteria has a poor fold and a low binding ability compared to those of CLC and CLC: CLF-1 purified from cell cultures (or lysates) cannot be excluded (46). Yet our present findings suggest an alternative and perhaps more likely explanation. As demonstrated by the SPR analysis, CLF-1 comprises independent sites for binding of both CLC and CNTFR␣. Also, it exhibits a similar or perhaps even higher affinity than CLC for CNTFR␣, and, notably, it can engage both targets at the same time. In other words, whereas CLC comprises a single receptor binding site and has a relatively modest affinity, the CLC: CLF-1 complex contains two sites which in combination create a bonus effect and a much higher affinity for CNTFR␣.
It can be concluded that in addition to its established function as a chaperone that facilitates transport and secretion of CLC in the biosynthetic pathway, CLF-1 may also play a decisive role in extracellular events involving the CNTFR␣ binding and signaling capacity of CLC. Previous findings have shown that a CLC-sCNTFR␣ fusion protein in which a linker peptide connects the two subunits provides a much stronger cellular stimulant than a 1:1 mixture of the two unlinked subunits (47). With that in mind, CLF-1 can be regarded as a sort of in vivo linker protein inasmuch as it binds both CLC and CNTFR␣ and thereby promotes their mutual interaction, their subsequent mobilization of the gp130/ LIFR␤ heterodimer, and ultimately their signal induction in responding cells. It has so far been generally accepted that the CLC subunit alone accounts for the binding of CLC:CLF-1 to the CNTFR␣; the present findings do not exclude this possibility, but they do substantiate a likely and alternative mode of binding which includes the participation of the CLF-1 subunit.
Functional aspects of CLF-1's interaction with the Vps10p-D receptor sorLA. Since CLF-1 and CLC are not always coexpressed, it could be speculated that CLF-1 have CLC-independent functions and perhaps target alternative ligands and receptors. Previous findings suggest that the p28 subunit of IL-27 may represent such an alternative ligand (19), but, apart from CNTFR␣ (as shown in this study) potential CLF-1-specific receptors have not been reported until now. CLF-1 binds to sorLA's Vps10p-D with an affinity similar to or higher than its affinity for CNTFR␣. Notably, sorLA does not interact with CLC but binds the CLC:CLF-1 heterodimer just as well as CLF-1, signifying that CLF-1 comprises at least three binding sites: one for CLC, one for CNTFR␣, and one for sorLA. The sites are functionally independent, and complex formation with one partner does not prevent or compromise simultaneous interaction with either of the two others. Ultimately, CLF-1 can bind all its targets at the same time and in this way link both CLC and CNTFR␣ to sorLA.
sorLA is expressed both on the cell membrane and in intracellular compartments, and its cytoplasmic tail contains functional motifs for adaptors (e.g., GGAs and AP-1 and -2) involved in endocytosis as well as Golgi compartment-endosome trafficking (27,29,48). We have previously demonstrated that sorLA may sort certain newly synthesized proteins from the Golgi compartment to lysosomes and thereby regulate their cellular release (30). This is not the case with CLF-1. CLF-1 binds to the Vps10p-D of sorLA, and this domain contains a propeptide, which prevents domain-specific ligand binding in the Golgi compartment (26). To engage ligands targeting the Vps10p-D, sorLA must be subjected to propeptide cleavage, and as this event takes place after passage through the Golgi compartment, secretion of CLF-1 and/or CLC:CLF-1 is not affected by the presence of sorLA. However, sorLA mediates highly efficient uptake and endocytosis of CLF-1 and CLC:CLF-1 and appears to be a major player in the turnover of both. In accordance, we found rapid and significant cellular internalization of both ligands in sorLA transfected cell lines and in wt astrocytes and hippocampal neurons but nearly no internalized ligand in sorLA knockouts. Evidently, sorLA is a potent clearance receptor for removal of CLF-1 and the CLC:CLF-1 complex, but via its affinity for CLF-1 it also has a considerable impact on the turnover and expression of the CNTFR␣.
sorLA conveys CLF-1-dependent downregulation of CNTFR␣. Our cross-linking data and SPR analysis demonstrate direct binding between sorLA and CNTFR␣, and our findings in cells suggest that this mechanism may account for a minor fraction of internalized CNTFR␣. The direct interaction is abrogated when CNTFR␣ is in complex with a ligand (CLC), but when complexed with CLC:CLF-1, the CLF-1 subunit takes over and reestablishes the binding to sorLA, which in turn promotes endocytosis of the entire CLC:CLF-1:CNTFR␣ complex. As a result, sorLA-expressing cells subjected to prolonged stimulation with CLC:CLF-1 exhibit substantial downregulation of CNTFR␣ due to lysosomal degradation of internalized receptors and a low level of CNTFR␣-synthesis.
Following preexposure to CLC:CLF-1 (and sorLA-mediated downregulation of CNTFR␣), the cellular response to renewed CLC:CLF-1 stimulation was significantly reduced in wt cells with endogenous expression of CNTFR␣ and sorLA and even more so in transfectants overexpressing sorLA. Thus, sorLA may have a profound effect on the cellular response under conditions of constant exposure to ligand (CLC:CLF-1), but apart from that, sorLA did not directly influence signaling; i.e., it neither enhanced nor hampered interaction with the gp130/LIFR␤ heterodimer in CNTFR␣-expressing cells. However, in cells expressing gp130/ LIFR␤ but not CNTFR␣, expression of sorLA enhanced signaling (pSTAT3) in response to sCNTFR␣ in complex with CLC:CLF-1 but not to the CLC-sCNTFR␣ fusion protein complex, which does not bind sorLA. These results suggest that in cells without membrane-expressed CNTFR␣, sorLA may bind and concentrate the tripartite soluble complex (CLC:CLF-1:sCNTFR␣) on the cell membrane and in this way promote interaction with gp130/ LIFR␤ and thereby signaling. A previous study based on molecular modeling reported an overlap between CLC's binding sites for CLF-1 and LIFR␤, suggesting that CLF-1 has to be dissociated/ removed to allow signaling, i.e., interaction between CLC and LIFR␤. However, we have recently generated a CLC-linker peptide-CLF-1 fusion protein (unpublished data) which binds both CNTFR␣ and sorLA and which is fully capable of signal induction. Since this indicates that CLF-1 may well be part of the signaling complex, we have included it in our proposed model (Fig. 9C).
Similar functions in different settings? Our study has revealed new functional aspects of CLF-1 and sorLA in the regulation of CLC:CNTFR␣ signaling and turnover, but it also brings a new perspective to previous observations concerning (i) the sorLArelated receptor sortilin and (ii) the IL-27 subunit p28. Thus, sortilin exhibits high affinity for CLC:CLF-1 (20), and CLF-1 may interact with p28 to form a cytokine complex which, in the presence of IL-6R␣, activates NK and T cells (19). Whether the findings reflect that sortilin plays a CLF-1-dependent role similar to that of sorLA and/or that CLF-1 implicates sorLA (and possibly sortilin) in p28 functions are among the obvious subjects for future research.