TWEAK-Independent Fn14 Self-Association and NF-κB Activation Is Mediated by the C-Terminal Region of the Fn14 Cytoplasmic Domain

The tumor necrosis factor (TNF) superfamily member TNF-like weak inducer of apoptosis (TWEAK) is a pro-inflammatory and pro-angiogenic cytokine implicated in physiological tissue regeneration and wound repair. TWEAK binds to a 102-amino acid type I transmembrane cell surface receptor named fibroblast growth factor-inducible 14 (Fn14). TWEAK:Fn14 engagement activates several intracellular signaling cascades, including the NF-κB pathway, and sustained Fn14 signaling has been implicated in the pathogenesis of chronic inflammatory diseases and cancer. Although several groups are developing TWEAK- or Fn14-targeted agents for therapeutic use, much more basic science research is required before we fully understand the TWEAK/Fn14 signaling axis. For example, we and others have proposed that TWEAK-independent Fn14 signaling may occur in cells when Fn14 levels are highly elevated, but this idea has never been tested directly. In this report, we first demonstrate TWEAK-independent Fn14 signaling by showing that an Fn14 deletion mutant that is unable to bind TWEAK can activate the NF-κB pathway in transfected cells. We then show that ectopically-expressed, cell surface-localized Fn14 can self-associate into Fn14 dimers, and we show that Fn14 self-association is mediated by an 18-aa region within the Fn14 cytoplasmic domain. Endogenously-expressed Fn14 as well as ectopically-overexpressed Fn14 could also be detected in dimeric form when cell lysates were subjected to SDS-PAGE under non-reducing conditions. Additional experiments revealed that Fn14 dimerization occurs during cell lysis via formation of an intermolecular disulfide bond at cysteine residue 122. These findings provide insight into the Fn14 signaling mechanism and may aid current studies to develop therapeutic agents targeting this small cell surface receptor.


Introduction
Fibroblast growth factor-inducible 14 (Fn14) was first described in 1999 as a growth factor-inducible, immediately-early gene predicted to encode a 129-aa type I transmembrane protein that would be cleaved intracellularly by signal peptidase into a mature 102-aa protein of unknown biological function [1,2]. Shortly after these initial Fn14 studies were published, Wiley et al. [3] reported that the TNF superfamily member TWEAK could bind to Fn14 with low nanomolar affinity and, as predicted from this result, that Fn14 had several structural characteristics that supported its classification as a new member of the TNF receptor (TNFR) superfamily. TWEAK, a multifunctional cytokine that can induce either cell death, proliferation, survival, or differentiation, depending on the cellular context (reviewed in [4,5]), is the only TNF superfamily member that can bind Fn14 [6]. TWEAK:Fn14 engagement has been shown to promote TNFR associated factor (TRAF) binding [7] and activation of a number of intracellular signal transduction cascades, including the ERK1/2 [8][9][10], PI3K/Akt [11], and NF-kB [8][9][10][12][13][14][15][16] pathways. Studies using TWEAK-or Fn14-deficient mice have revealed that TWEAK/ Fn14 signaling is not required for embryonic development or postnatal growth [17,18] but may be critical for wound repair following acute tissue injury [15,18,19].
The TWEAK/Fn14 axis has been implicated in various human diseases. For example, recent work using several mouse models of human chronic inflammatory disease has indicated that TWEAK activity may exacerbate disease progression (reviewed in [4,5]). Indeed, a Phase II clinical trial is in progress to test whether an anti-TWEAK monoclonal antibody may be a beneficial therapeutic agent for lupus nephritis patients (ClinicalTrials.gov Identifier NCT01499355). TWEAK and Fn14 may also be targets for cancer therapy (reviewed in [4,20,21,22]). Of particular interest, Fn14 gene expression is elevated in over a dozen different solid tumor types compared with matched adjacent normal tissue or normal tissues from non-diseased donors [12,[23][24][25][26]. TWEAK/Fn14 signaling can have anti-tumorigenic effects (reviewed in [4,22]); for example, TWEAK is a pro-apoptotic factor for some human cancer cell lines, and two companies have developed agonistic Fn14 antibodies that can kill cancer cells in vitro and inhibit xenograft tumor growth in vivo [23,[26][27][28]. TWEAK/Fn14 signaling can also have pro-tumorigenic effects (reviewed in [4,22]) [11,12,29]; accordingly, in pre-clinical studies we have demonstrated that Fn14-targeted immunotoxins have anti-tumor activity [25,30] and Hoffmann-LaRoche Inc. has initiated a Phase I clinical trial testing the effects of an anti-TWEAK monoclonal antibody in patients with advanced solid tumors (ClinicalTrials.gov Identifier NCT01383733).
A complete understanding of TWEAK/Fn14 signaling is critical for the optimal design of TWEAK/Fn14-directed therapeutic agents. One outstanding issue is whether Fn14 ''activation'' can occur not only in response to ligand binding (TWEAKdependent Fn14 signaling) but also when Fn14 expression is elevated (TWEAK-independent Fn14 signaling). This latter signaling mechanism may be particularly important in cancers where Fn14 levels are high but TWEAK levels are low (e.g., in glioblastomas [31] and melanomas (unpublished data)). The main evidence in support of TWEAK-independent signaling is that experimental manipulation of Fn14 expression levels in cancer cells cultured in vitro can regulate signal transduction pathways [32] and cellular properties; for example, cell survival, migration and invasion [12,24,[32][33][34][35][36]. However, these results do not conclusively demonstrate TWEAK-independent Fn14 signaling for two main reasons. First, the cells were grown in culture medium containing serum, a potential source of TWEAK [37,38], and second, the cells themselves could be expressing TWEAK, and in particular, they could be releasing the soluble TWEAK isoform into the medium. In this report, we directly demonstrate TWEAK-independent Fn14 signaling by showing that an Fn14 deletion mutant encoded by an Fn14 splice variant mRNA is unable to bind TWEAK but can still activate the NF-kB pathway in transfected cells. We then show that ectopically-expressed, cell surface-localized Fn14 can self-associate into Fn14 dimers, and this dimerization is mediated by a region within the Fn14 cytoplasmic tail. Finally, we present additional evidence that Fn14 monomers are self-associated in cells by demonstrating that Fn14 dimers can be detected when cell lysates are examined under non-reducing gel conditions. These dimers form when cells are lysed and are due to the formation of a single intermolecular disulfide bond at Fn14 cysteine residue 122 in the cytoplasmic tail.

RT-PCR and Construction of Human Fn14-myc Expression Plasmids
The expression plasmids encoding full-length, wild-type human Fn14 or an Fn14 mutant missing 35 amino acid residues in the extracellular domain (the DEC mutant) with N-terminal myc epitope tags were constructed as follows. First, RNA was isolated from human U87-luc and MDA-MB-231 cells using the RNeasy kit (Qiagen) and cDNA synthesized using the AccuScript RT-PCR system (Stratagene) according to the manufacturer's instructions. Second, PCR was performed in the presence of 5% DMSO using a sense Fn14 primer targeting exon 1 nucleotides 150-176, an antisense primer targeting exon 4 nucleotides 30-56, and Vent Polymerase (New England Biolabs). Reaction products were separated by agarose gel electrophoresis and detected by ethidium bromide staining. Two DNA products were detected and the U87luc products were isolated and ligated into pCMVScript (Stratagene). These plasmids were used as the template to add an optimal Kozak sequence (GCCGCCACC) prior to the ATG codon and to insert DNA encoding the myc epitope peptide (E-Q-K-L-I-S-E-E-D-L) at base 81, immediately 39 to the DNA encoding the Fn14 signal peptide, by the PCR overlap extension method with Vent Polymerase. The final pCMVScript/Fn14-FL and pCMVScript/Fn14-DEC plasmids were then digested with Not1/Xho1 and the released DNA fragments ligated into Not1/ Xho1-digested pcDNA6 plasmid (Invitrogen). The expression plasmid encoding an Fn14 mutant missing the C-terminal 18 amino acids (the DCT mutant) was constructed using pCMVScript/Fn14-FL as a template and primers bracketing Fn14 amino acids 30 to 111. The sense primer contained an Asc1 site and the antisense primer contained a stop codon and an Xho1 site. The PCR product was cloned into pSecTag2A/Hygro (Invitrogen) as an Asc1/Xho1 fragment to add the Ig-k chain signal peptide in frame and then further cloned into Not1/Xho1digested pcDNA6. The expression plasmids encoding Fn14 mutants with C104S or C122S amino acid substitutions were generated using overlapping PCR methodology with pCMVScript/Fn14-FL as the template and the appropriate primers. The PCR products were isolated, cloned into pCMVScript, and then subcloned as Not1 fragments into pcDNA6. All expression constructs were subjected to DNA sequence analysis using appropriate primers and an Applied Biosystems automated sequencer.

Plasmid DNA Transfections
Cells were transiently transfected using 10 mg plasmid DNA per 10-cm plate and the X-tremeGENE HP transfection reagent according to manufacturer's instructions (Roche). Cells were generally harvested 24 hr post-transfection for subsequent analysis.

Western Blot Analysis
Cells were lysed in HNTG buffer and protein concentrations determined as previously described [16]. In some experiments the cell lysis buffer was supplemented with 10 mM iodoacetamide (Sigma). Equal amounts of protein were subjected to SDS-PAGE under either reducing (DTT in loading buffer, sample is heated) or non-reducing (no DTT in loading buffer, sample is not heated) conditions and Western blot analysis conducted as described [16] except that proteins were transferred to PVDF, not nitrocellulose, membranes. The following primary antibodies were used: (a) anti-Fn14 polyclonal antibody (pAb) #3600

Ligand Blot Analysis
HEK293 cells were transiently transfected with the pcDNA6 vector, the pcDNA6/Fn14-FL plasmid, or the pcDNA6/Fn14-DEC plasmid and harvested 24 hr post-transfection. Cells were lysed and equal amounts of protein were subjected to SDS-PAGE under non-reducing conditions as described above. The membrane was blocked and then sequentially incubated with 100 ng/ ml recombinant human Fc-TWEAK fusion protein (provided by Dr. Pascal Schneider, University of Lausanne) and HRPconjugated goat anti-human IgG/Fc secondary antibody (Millipore). The membrane was washed and immunoreactive Fc-TWEAK detected using Enhanced Chemiluminescence Plus.

NF-kB Pathway Luciferase Reporter Assays
HEK293/NF-kB-luc cells were plated in normal growth medium in 6-well plates. The pcDNA6 vector, the pcDNA6/ Fn14-FL plasmid, or the pcDNA6/Fn14-DEC plasmid were transfected in triplicate (1.5 mg/well) using X-tremeGENE HP as described above and harvested 24 hr later. Cells were lysed in Tropix Lysis Solution (Applied Biosystems). Aliquots of each cell lysate were combined with Luc-Screen System reagent (Applied Biosystems) and analysed in duplicate using an ML2250 microtiter plate luminometer (Dynatech Laboratories).

Cell Surface Protein Cross-linking Experiments
HEK293/NF-kB-luc/Fn14-HA cells were transiently transfected with the pcDNA6 vector, the pcDNA6/Fn14-FL plasmid, the pcDNA6/Fn14-DEC plasmid, or the pcDNA6/Fn14-DCT plasmid and 24 hr later they were washed in PBS and then treated with 1 mM BS3 (Pierce) in PBS for 20 min at room temperature. The cross-linking reaction was quenched with 50 mM Tris-HCl (pH 7.4) for 15 min at room temperature and then the cells were harvested by scraping. Cells were lysed in HNTG buffer, and protein concentrations were determined using the BCA protein

Co-immunoprecipitation Analysis
HEK293/NF-kB-luc/Fn14-HA cells were transiently transfected with the pcDNA6 vector, the pcDNA6/Fn14-FL plasmid, the pcDNA6/Fn14-DEC plasmid, or the pcDNA6/Fn14-DCT plasmid. Cells were harvested 24 hr post-transfection, lysed in HNTG buffer, and protein concentrations were determined using the BCA protein assay. An aliquot of each lysate was used for Western blot analysis. Co-immunoprecipitations were conducted using equal amounts of protein and the ProFound c-Myc Tag IP/Co-IP kit according to manufacturer's instructions (Thermo Scientific).
Samples were subjected to SDS-PAGE under reducing conditions and Western blot analysis was performed.

Identification of an Fn14 Splice Variant mRNA Predicted to Encode a Short Fn14 Isoform
The original goal of this study was to determine whether Fn14 protein-positive cancer cells expressed more than one Fn14 mRNA species. To test this, we performed RT-PCR analysis using RNA isolated from human MDA-MB-231 breast cancer cells and U87 glioma cells and exon 1-or exon 4-targeted oligonucleotide primers (Fig. 1A). Two major amplification products of 310-and 205-bp were detected (Fig. 1B). The U87 cell mRNA-derived DNA products were isolated, cloned and sequenced. The product A nucleotide sequence was 100% identical to the corresponding region of the human Fn14 cDNA sequence that we previously reported (GenBank Accession Number AF191148) [2]. This cDNA encodes full-length (129-aa) Fn14 (referred to here as Fn14-FL). The product B nucleotide sequence was 100% identical to the corresponding region of an unpublished human Fn14 cDNA sequence that was deposited in GenBank by S. Tanaka in 1999 (GenBank Accession Number AB035481). In comparison to product A, the product B sequence has a 105-bp deletion, and is predicted to encode a shorter Fn14 isoform missing an internal 35-aa region within the 53-aa extracellular domain (amino acids 33-67; we refer to this protein here as Fn14 extracellular domain deletion (Fn14-DEC) (Fig. 1C). Analysis of the human Fn14 exon sequences revealed that exon 2 is 105-bp in length and encodes these same 35 amino acids, suggesting that this exon is a cassette-type alternative exon (reviewed in [40]). Indeed, further analysis of the Fn14 gene exon/intron junction sequences is consistent with the proposal that alternative splicing of the Fn14 primary transcript via exon 2 skipping would generate the Fn14-DEC mRNA and protein  represented by product B (Fig. 1D). The structural domains of the Fn14-FL and Fn14-DEC proteins are shown in reference to the Fn14 coding exons in Fig. 1E. We have been unable to detect an Fn14 protein species in MDA-MB-231 or U87 cells that we can conclusively identify as the Fn14-DEC protein by Western blotting using an antibody recognizing the Fn14 cytoplasmic tail (data not shown). Thus, this smaller Fn14 isoform may either not be expressed by these cells at all, or alternatively, it could be expressed at relatively low levels and not detected by the antibody we utilized.

The Fn14-DEC Protein Cannot Bind TWEAK
The Fn14-DEC isoform should not be able to bind TWEAK since it is missing most of the extracellular domain, and in particular, it does not have the cysteine-rich region within this domain that has been previously shown to contain the three disulfide bonds critical for proper Fn14 folding and TWEAK:Fn14 interaction [41][42][43]. We confirmed that TWEAK could not bind to Fn14-DEC using a ligand blot assay. Expression plasmids encoding either myc epitope-tagged Fn14-FL or Fn14-DEC ( Fig. 2A), as well as the vector itself, were transfected into HEK293 cells. Cell lysates were prepared and protein was subjected to SDS-PAGE under non-reducing conditions (for ligand blot analysis) or reducing conditions (for Western blot analysis). TWEAK interaction with the Fn14-FL protein, but not the Fn14-DEC protein, was detected using the ligand blot assay, although Western blot analysis indicated that both proteins were expressed to a similar degree in the transfected cells (Fig. 2B).

Fn14-DEC Overexpression in Transfected HEK293 Cells can Activate the NF-kB Signaling Pathway
There is substantial experimental evidence supporting the proposal that when cellular Fn14 levels are high, Fn14 can activate downstream signaling events without ligand engagement (a process referred to as TWEAK-independent Fn14 signaling (reviewed in [4]). However, the studies performed to date have not ruled out the possibility that the Fn14-triggered cellular effects observed, for example, NF-kB pathway activation [32,44,45], could be mediated by a low level of soluble TWEAK in the cell culture medium (either present in FBS [37,38] or secreted by the cells under investigation). The cloning of Fn14-DEC afforded us the opportunity to directly test whether ectopic Fn14 overexpression promoted TWEAK-independent Fn14 signaling. As our readout for Fn14 signaling we selected NF-kB pathway activation, since TWEAK:Fn14 engagement in diverse cell types has been shown to stimulate this intracellular pathway (reviewed in [4]). HEK293/NF-kB-luc reporter cells were transiently transfected with vector or expression plasmids encoding either Fn14-FL (our positive control), Fn14-DEC, or Fn14-DCT, an Fn14 mutant that is missing the C-terminal 18-aa residues of the 28-aa Fn14 cytoplasmic tail (Fig. 3A) and therefore cannot bind TRAFs and activate signaling pathways (our negative control) [44]. All Fn14 proteins had an N-terminal Myc epitope tag. Cells were harvested 24 h later and processed for Western blot analysis and luciferase reporter assays. The three Fn14 proteins were expressed at similar levels (Fig. 3B). We noted previously that Fn14 does not migrate at its predicted molecular mass when analyzed by SDS-PAGE under reducing conditions [1] and this is again illustrated on this particular blot. For example, the predicted molecular mass (MM) of the mature, 112-aa myc-tagged Fn14-FL protein is 12.1-kDa, but its apparent MM on this gel is ,23-kDa. Also, the 101-aa Fn14-DCT mutant (11.0-kDa predicted MM) is running at a lower apparent MM than the much smaller, 77-aa Fn14-DEC mutant (8.4-kDa predicted MM). The basis for these findings is not known but protein amino acid composition and isoelectric point are two interrelated factors that can influence the mobility of proteins on SDS polyacrylamide gels. For small proteins like Fn14, the effects of these factors on electrophoretic mobility are even more apparent. One possible explanation for the anomalous migration of the three Fn14 species relative to one another is that the Fn14-FL and Fn14-DEC proteins are acidic (predicted pIs of 5.8 and 5.3, respectively) while the Fn14-DCT mutant is basic (predicted pI of 8.8). Transient overexpression of the Fn14-FL, Fn14-DEC, or Fn14-DCT proteins resulted in a 2.14-, 3.38-and 1.08-fold induction of the NF-kB pathway reporter gene, respectively (Fig. 3C). This finding demonstrates that TWEAK-independent Fn14 signaling can occur when cellular Fn14 levels are elevated. Although we used an ectopic overexpression approach to detect this signaling mechanism, it may also occur when Fn14 is naturally highly expressed, since we have reported previously that endogenous Fn14 down-regulation in cancer cells via RNA interference can alter cellular processes such as cell migration [24] and invasion [12,24,32].
Fn14 Overexpression in Transfected HEK293 Cells Promotes Fn14 Self-association, which is Mediated by the C-terminal 18-aa Segment of the Fn14 Cytoplasmic Tail The classic model of TNF:TNFR family signaling is that the binding of homotrimeric ligands promotes receptor trimerization which in turn leads to recruitment of signaling complexes to the cytoplasmic domains of the receptors (reviewed in [46]). However, it should be noted that the TNFR family member DR5 forms dimers, trimers and large multimers after ligand engagement [47]. Also, some TNFR family members (e.g., TNFR1 and CD40 [48,49]) contain a pre-ligand assembly domain (PLAD) in their extracellular region that mediates non-covalent receptor association in the absence of ligand binding (reviewed in [50]). In the case of TNFR1, there is crystallographic evidence for receptor dimerization prior to ligand engagement [51]. The Fn14 extracellular region, only 53-aa in length, does not contain a PLAD domain, and there have been no studies reporting the structural organization of Fn14 either prior to ligand engagement, after ligand engagement, or when overexpressed in transfected cells. We previously proposed that the most likely explanation for TWEAK-independent Fn14 signaling is that when Fn14 is expressed at high levels in cells it will spontaneously self-associate, and this ''receptor clustering'' would promote TRAF association, downstream signaling and cellular responses (reviewed in [4]). To test if high Fn14 levels induce Fn14 self-association, our approach was to transfect cells and then cross-link all cell surface proteins using BS3, a membrane-impermeable, amine-to-amine crosslinking agent. Prior to conducting these experiments, we first determined if the Fn14-FL, Fn14-DEC, and Fn14-DCT proteins were all present on the HEK293 cell surface following transient transfection by performing FACS analysis with an anti-myc epitope tag antibody. We found that each protein was expressed on the cell surface to a similar degree (Fig. 4A). For the crosslinking experiments, cells were transiently transfected with vector or the Fn14-FL, Fn14-DEC, or Fn14-DCT expression plasmids and then either left untreated or treated with BS3. The crosslinking reaction was quenched and the cells were lysed and processed for Western blot analysis. We were able to detect the presence of cross-linked Fn14-FL and Fn14-DEC species at the approximate molecular weight of an Fn14 dimer in response to BS3 addition (Fig. 4B). A relatively small amount of each Fn14 protein was detected as cross-linked dimer. This indicates that only a small percentage of the Fn14 is self-associated or, alternatively, that the crosslinking reaction is inefficient (the only available primary amines in the Fn14-myc-overexpressing cells include the myc tag N-terminus and the side chain of lysine residue 3 (myc epitope tag) and lysine residue 31 (Fn14 extracellular domain)). In contrast to our findings regarding the Fn14-FL and Fn14-DEC proteins, ectopic overexpression of the Fn14-DCT protein and cross-linker addition did not generate a high molecular weight cross-linked Fn14 species. If we assume that Fn14-triggered NF-kB activation requires self-association followed by TRAF binding, then the inability of overexpressed Fn14-DCT protein to signal (Fig. 3C) may reflect ineffective self-association, not simply the absence of a TRAF binding site.
Taken together, these results indicate that (i) ectopic Fn14 overexpression can promote Fn14 self-association, (ii) Fn14 selfassociation is not ligand-dependent, and (iii) Fn14 self-association requires the most C-terminal 18-aa segment of the Fn14 cytoplasmic tail. Previous co-immunoprecipitation studies have revealed that ectopic overexpression of the TNFR superfamily members CD30 [52] and RANK [53] also results in receptor selfassociation. In the case of RANK, the self-association domain was mapped by deletion and site-specific mutagenesis to a 6-aa region in the cytoplasmic tail [53].
The Fn14-DEC and Fn14-DCT Proteins can Both Associate with the Fn14-FL Protein It has been reported that when CD40 [54] and CD30 [52] deletion mutants lacking TRAF binding sites in the cytoplasmic domain are overexpressed in cells they can act as dominant-negative inhibitors of ligand-stimulated signaling, presumably by associating with endogenous receptors and thereby forming nonfunctional receptor complexes. In a previous study, we reported that MDA-MB-231 breast cancer cell invasion capacity could be inhibited by forced Fn14-DCT expression [32], suggesting that Fn14-DCT may also be a dominant-negative inhibitor. Therefore, we wanted to determine whether Fn14-DCT could associate with the Fn14-FL protein even though it cannot associate with itself. For these studies, we used a co-immunoprecipitation approach. HEK293 cells engineered to stably overexpress HA epitope-tagged Fn14-FL (Fig. 5A) were transfected with vector or the Fn14-FL, Fn14-DEC, or Fn14-DCT expression plasmids (all proteins myctagged) and 24 hr later the cells were harvested. Western blot analysis was performed on an aliquot of each lysate using either an anti-HA or anti-myc antibody to confirm expression of all Fn14 proteins (Fig. 5B). Lysates were also subjected to immunoprecipitation analysis using an immobilized anti-myc antibody and Fn14-HA:Fn14-myc association evaluated by Western blot analysis with an anti-HA antibody. We found that all three myc-tagged Fn14 proteins were effectively immunoprecipitated with the anti-myc antibody and could interact with HA-tagged Fn14-FL (Fig. 5B).
These results indicate that the effect of Fn14-DCT in our earlier study [32] most likely reflects Fn14-DCT association with endogenous Fn14, leading to the formation of signaling-deficient heteromeric complexes. Indeed, one would predict that ectopic Fn14-DCT expression in cells would interfere with both TWEAKdependent and -independent Fn14 signaling. Our finding that Fn14-DEC can also associate with Fn14-FL suggests that if this Fn14 splice variant isoform is naturally co-expressed at significant levels in cells with the Fn14-FL protein, it could act as a dominantnegative inhibitor of TWEAK binding to Fn14 (i.e., TWEAKdependent Fn14 signaling). However, since Fn14-DEC can signal on its own when overexpressed (Fig. 3C), one would not expect it to interfere with TWEAK-independent Fn14 signaling.

Fn14 Dimers are Detected when Cell Lysates are Analyzed by SDS-PAGE Under Non-reducing Conditions
We noted that in our ligand blot experiments, in which cell lysate samples are run under non-reducing conditions (no DTT in loading buffer, sample is not heated), the overexpressed Fn14-FL protein had slower electrophoretic mobility when compared to its mobility under normal (reducing) SDS-PAGE conditions (Fig. 2B). Therefore, we examined whether endogenously-expressed Fn14 also had different electrophoretic mobility under these two conditions. Lysates were prepared from three different cancer cell lines, protein was subjected to SDS-PAGE under either reducing or non-reducing conditions, and Western blot analysis was performed. In all three cell lines, Fn14 migrated as a single species with an apparent molecular mass of ,17-kDa under reducing conditions (Fig. 6). In comparison, under non-reducing conditions, the major immunoreactive Fn14 species detected in the cell lines was ,29-kDa, with two lower molecular mass species of ,26-and ,17-kDa also present in the A375 cell lysate. These results indicate that endogenous Fn14 could be forming a disulfide-linked dimer, either in the cell itself or after cell harvest during the cell lysis procedure.

Fn14 Dimerization Occurs during Cell Lysis via Formation of an Intermolecular Disulfide Bond at Cysteine Residue 122
Fn14 contains a total of eight cysteine residues. Six of these residues are in the extracellular region, where they form the three intramolecular disulfide bonds that stabilize the cysteine-rich domain [41][42][43]. The other two cysteines, Cys104 and Cys122, are located in the cytoplasmic tail, and since disulfide bonds are formed in the lumen of the rough ER during protein synthesis, they are not expected to form intra-or intermolecular disulfide bonds in living cells. In order to determine which Fn14 cysteine residues were likely involved in the endogenous Fn14 dimerization noted above, HEK293 cells were transfected with vector or the Fn14-FL (contains all 8 cysteines), Fn14-DEC (contains only the 2 cysteines in the cytoplasmic tail), or Fn14-DCT (contains all 6 cysteines in the extracellular domain and Cys104 in the cytoplasmic tail) plasmids and 24 hr later the cells were harvested and lysed. Samples were subjected to SDS-PAGE under either reducing or non-reducing conditions and Western blot analysis was performed. Fn14-FL and Fn14-DEC dimers, but not Fn14-DCT dimers, were detected under non-reducing conditions (Fig. 7A). This result indicated that the most C-terminal Fn14 cysteine residue, Cys122, was critical for disulfide bond formation. Since this residue is within the cytoplasmic domain, this result also implied that disulfide bond formation was occurring after cell harvest. To confirm this, HEK293 cells were transfected with vector or the Fn14-FL protein and 24 hr later the cells were harvested. The Fn14-FL cells were then lysed in normal lysis buffer or lysis buffer supplemented with iodoacetamide, which reacts with free sulfhydryls on cysteine residues. Samples were subjected to SDS-PAGE under non-reducing conditions, and Western blot analysis was performed. Iodoacetamide treatment blocked Fn14-FL dimer formation (Fig. 7B), indicating that the Cys122 intermolecular disulfide bond was forming in the cell lysis buffer, a favorable redox environment for oxidation of SH groups.
We next confirmed that Cys122 within the Fn14 cytoplasmic domain was indeed the residue participating in intermolecular disulfide bond formation by site-specific mutagenesis. Expression plasmids were constructed that encoded myc-tagged Fn14 proteins with a cysteine to serine substitution at position 104 or 122 (Fig. 8A). HEK293 cells were transiently transfected with each of these plasmids, as well as the Fn14-FL (denoted wild-type (WT) in this figure) and Fn14-DCT plasmids used earlier. Cells were harvested, lysed, and samples were subjected to SDS-PAGE under either reducing or non-reducing conditions. Western blot analysis was performed. All four of the Fn14 proteins were expressed in the transfected cells and were detected as single species when analyzed under reducing conditions (Fig. 8B). Fn14-WT and Fn14-C104S dimers, but not Fn14-DCT or Fn14-C122S dimers, were detected under non-reducing conditions, confirming the importance of the C-terminal region of the cytoplasmic tail and Cys122 in particular in intermolecular disulfide bond formation (Fig. 8B).

Conclusions
In this study, we first show that there are at least two distinct Fn14 mRNA species, with one mRNA encoding the full-length 129-aa Fn14 protein and the second mRNA, generated by alternative splicing of the primary transcript, encoding a smaller 94-aa Fn14 protein missing most of the extracellular domain and therefore unable to bind the Fn14 ligand TWEAK. The cloning of the smaller protein, denoted Fn14-DEC, afforded us the opportunity to test our earlier proposal that high levels of Fn14 expression; for example, in mouse intestinal epithelial cells after tissue injury [19] or in human non-small cell lung cancer cells with EGFR-activating mutations [24], might trigger TWEAK-independent Fn14 signaling. We found that ectopic overexpression of Fn14-DEC could indeed activate the NF-kB signaling pathway. We then demonstrated that ectopic overexpression of Fn14 promoted self-association on the cell surface using a chemical cross-linking approach. Specifically, cell surface Fn14 was in a dimeric structure that was stabilized by the most distal 18-aa region of the Fn14 cytoplasmic tail. Endogenously-expressed Fn14 as well as ectopically-overexpressed Fn14 could also be detected in dimeric form when cell lysates were subjected to SDS-PAGE under non-reducing conditions. Additional experiments revealed that Fn14 dimerization occurs during cell lysis via formation of an intermolecular disulfide bond at cysteine residue 122. Taken together, our results support the model that both endogenous Fn14 and ectopically-overexpressed Fn14 monomers can selfassemble within cells into non-covalently associated dimers (Fig. 8C). These Fn14 monomers remain in such close proximity in cell lysis buffer that a disulfide bond linkage is able to form at Fn14 cysteine residue 122 in the cytoplasmic tail. These findings provide new insights into the TWEAK/Fn14 signaling pathway and will aid in the design, development, and testing of therapeutic agents targeting this pathway.