To investigate the effect of mutations in the tumor necrosis factor receptor superfamily 1A (TNFRSF1A) gene on the conformation and behavior of the TNFRSF1A protein. Mutations in TNFRSF1A cause the autosomal-dominant, autoinflammatory TNFR-associated periodic syndrome (TRAPS).
The expression of recombinant TNFRSF1A was compared in SK–HEp-1 endothelial cells and HEK 293 epithelial cells stably transfected with full-length R347A or Δsig constructs of wild-type or TRAPS-associated mutant TNFRSF1A. TNF binding was assessed in HEK 293 cell lines expressing R347A wild-type or mutant TNFRSF1A. Homology modeling of the 3-dimensional structure of the ectodomains of wild-type and mutant TNFRSF1A was performed.
TRAPS-associated mutant and wild-type TNFRSF1A behaved differently and had different localization properties within the cell, as a direct result of mutations in the ectodomains of TNFRSF1A. From a structural perspective, mutants with a predicted structure similar to that of the wild-type protein (e.g., R92Q) behaved similarly to wild-type TNFRSF1A, whereas forms of TNFRSF1A with mutations predicted to drastically destabilize the protein structure (e.g., cysteine mutations) showed defects in cell surface expression and TNF binding.
The results obtained from the in vitro experiments, in combination with the modeled structures, indicate that the phenotype and clinical differences between different TRAPS-associated mutants of TNFRSF1A result from different conformations of the TNFRSF1A ectodomains.
Tumor necrosis factor receptor–associated periodic syndrome (TRAPS; OMIM 142680) is an autoinflammatory periodic fever syndrome. It is caused by autosomal-dominant mutations in TNFRSF1A, the gene that encodes tumor necrosis factor receptor superfamily 1A (TNFRSF1A) (1, 2), the main cell surface receptor for TNF (3). Diagnostic indicators of TRAPS include periodic bouts of inflammation (often lasting >5 days), with symptoms that include fever, chest and abdominal pain, myalgia with migratory erythematous macular rashes, conjunctivitis and periorbital edema, arthralgia, monarticular synovitis, and, in ∼15% of patients, amyloidosis (4).
TNFRSF1A is a transmembrane glycoprotein, with 4 tandem-repeat cysteine-rich domains (CRD1–4), each of which contains 3 cysteine–cysteine disulfide bonds, a transmembrane region, and a conserved death domain (DD) in the intracellular compartment that are responsible for most of the signaling properties of the receptor (5). TNFRSF1A is usually associated with silencer of death domains (SODDs) until TNF ligation occurs at the surface (6). CRD1, also known as the preligand assembly domain, has been reported to undergo homologous interactions to form TNFRSF1A homotrimers (7). TNF binds as a trimer to the CRD2 and CRD3 regions of a TNFRSF1A trimer, which then induces interaction of its intracellular DD with TRADD, leading to downstream signaling events that may result in apoptosis (via caspase cascade activation) or cytokine production via NF-κB activation (8–13). More than 60 different mutations of TNFRSF1A have been reported (14) (online at http://fmf.igh.cnrs.fr/infevers/), and ∼60% of these mutations are associated with TRAPS. The majority of TRAPS-related mutations are point mutations that result in single amino acid substitutions in CRD1, CRD2, or CRD3 of the ectodomain of the mature protein. Approximately half of these mutations involve substitutions of the cysteine residues that form disulfide bonds (15).
Although it is well established that TRAPS is caused by mutations in the ectodomains of TNFRSF1A, to date the exact mechanism by which the periodic fevers are triggered has not been identified. It was originally proposed that TRAPS was a consequence of impaired down-regulation of membrane TNFRSF1A and diminished shedding of soluble receptor (1), but this has not been observed with all TRAPS-associated TNFRSF1A mutations and also appears to be a function of cell type and other features of the genetic makeup. Thus, other pathophysiologic mechanisms are probably also involved (16, 17).
More recently, reports suggested that, in some patients, TRAPS might be a result of defective TNF-induced signaling and apoptosis (18, 19), leading to reduced NF-κB signaling (20). Our own findings indicate that TRAPS-associated mutations in the ectodomains result in the receptor displaying defective behavior, but the signaling properties of the cytoplasmic DD are not defective in the mutants (17). In particular, we observed reduced TNF binding and increased intracellular retention of mutant TNFRSF1A (16, 17); the latter observation has also recently been reported by other investigators (20).
We hypothesized that in the mutant forms of TNFRSF1A, mutations in the ectodomains lead to a conformational change in the overall structure of the TNFRSF1A protein that subsequently causes the receptor to misfold and exhibit aberrant behavior, through a gain-of-function effect via ligand-independent aggregation and signaling (17) (Figure 1). Our hypothesis has recently been supported by other investigators (21). Consistent with this, the change in behavior in the TRAPS mutants in comparison with the change in wild-type TNFRSF1A is much more pronounced in mutations that may have a more drastic effect on the overall 3-dimensional (3-D) structure of the receptor (17). Several examples discussed in the literature, including cystic fibrosis, Alzheimer's disease, and Parkinson's disease, involve other protein conformational disorders that result in accumulation and aggregation of misfolded proteins, which lead to disease (22–27). Abnormal signaling by mutant proteins leading to a gain-of-function effect of the mutants when compared with their wild-type counterparts has also been reported as a mechanism leading to cancer (28), immunodeficiency diseases (29), and a subset of Alzheimer's disease–linked mutations (30).
To further develop our understanding of the effects of mutations in the TNFRSF1A ectodomains, we made recombinant constructs of both mutant and wild-type forms of TNFRSF1A with a single-point mutation, R347A, that abolishes SODD and TRADD binding without affecting the general protein structure and folding (31, 32). This approach allowed us to investigate the structural effects of TRAPS-related mutations without most of the complicating effects of the receptor signaling function, but with the advantage of still using the full-length protein instead of the deletion construct lacking the cytoplasmic region (Δsig) that was used in our previous studies (16, 17).
In order to investigate the link between receptor functionality and protein conformation, we extended the range of TNFRSF1A constructs examined by comparing the R92Q mutation with 3 other R92 mutants that were chosen for their potential ability to cause changes in the overall TNFRSF1A protein conformation. R92Q was chosen because it is a low-penetrance mutation that can be considered to have a fairly conserved structural impact on overall receptor conformation, and because it exhibits properties similar to those of wild-type TNFRSF1A in our model (17).
We produced TNFRSF1A transfectants of the SK–HEp-1 endothelial cell line as well as the HEK 293 epithelial cell line, which we used previously (17). The SK–HEp-1 line may be more relevant to the pathophysiology of TRAPS, because endothelial cells are known to be an important site of TNFRSF1A expression and function in the setting of inflammation. The levels of expression of TNFRSF1A in SK–HEp-1 cells were lower than those observed in HEK 293 cells; this allowed distinction between surface expression of the R92Q mutant and that of wild-type TNFRSF1A, which was not previously apparent.
Three-dimensional homology modeling of the extracellular domain of the wild-type and mutant forms of TNFRSF1A was generated based on the crystal structure of TNFRSF1A, to allow the potential effects of mutations in the ectodomains of TNFRSF1A to be predicted. This indicated the effects of point mutations on the entire ectodomain structure of TNFRSF1A rather than just within the vicinity of the mutations. Taken together, the results obtained from the in vitro experiments and homology modeling support our hypothesis that misfolding of mutant TNFRSF1A may have functional consequences that play a role in the pathophysiology of TRAPS.
MATERIALS AND METHODS
Production of recombinant DNA clones of wild-type and mutant TNFRSF1A.
The production of recombinant DNA clones of wild-type and mutant (R92Q, T50M, C33Y, and C52F) TNFRSF1A, encoding either the full-length receptor or lacking the cytoplasmic signaling domain (Δsig) due to a stop codon at residue 215, has been described previously (16, 17).
In the R347A constructs, TNFRSF1A DD signaling was abolished by introducing a single point mutation that leads to the substitution of arginine by alanine at residue 347 of the mature protein. Wild-type and mutant (R92Q, T50M, C33Y, and C52F) R347A TNFRSF1A DNA was generated from each respective full-length TNFRSF1A DNA construct cloned into the vector pcDNA4TO (Invitrogen, San Diego, CA). The R347A point mutation was introduced using site-directed mutagenesis (QwikChange kit; Stratagene, Amsterdam, The Netherlands) with the primer 5′-CTGGAAGGAATTCGTGGCGCGCCTAGGGCTGAG-3′ and its reverse complement.
In the R92 constructs, wild-type TNFRSF1A DNA constructs cloned into the vector pcDNA4TO were used in site-directed mutagenesis to introduce single point mutations that lead to the substitution of arginine by lysine (R92K), glutamic acid (R92E), or proline (R92P) at residue 92 of the mature protein. The following primers and their reverse complements were used for each construct: for R92E, 5′-CTTGCACAGTGGACGAAGACACCGTGTGTG-3′; for R92K, 5′-TCTTGCACAGTGGACAAGGACACCGTGTGT-3′; for R92P, 5′-TCTTGCACAGTGGACCCCGACACCGTGTGT-3′.
All products were sequenced along the full-length TNFRSF1A coding region to ensure that only the desired mutation(s) had been introduced. Sequencing was carried out in an ABI Prism 310 analyzer using the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction kit (both from Applied Biosystems, Warrington, UK). Plasmid isolation from bulk cultures for transfection into eukaryotic cells was carried out using a GenElute High Performance Plasmid Maxiprep kit (Sigma, Poole, UK).
Transfection of cell lines expressing recombinant TNFRSF1A.
Two human cell lines known to express TNFRSF1A endogenously were used for transfection: the Tetracycline-Regulated Expression (T-REx) HEK 293 cell line and the human liver adenocarcinoma SK–HEp-1 endothelial cell line. The SK–HEp-1 cells were stably transfected with the T-REx gene in the plasmid vector pcDNA6/TR, using FuGENE 6 (Roche Diagnostics, Lewes, UK) according to the manufacturer's protocols. A FuGENE 6 reagent:DNA ratio of 3:1 was used. Forty-eight hours posttransfection, the cells were switched to selective medium containing 5 μg/ml blasticidin HCl (Invitrogen Life Technologies, Renfrew, UK) for 3 weeks.
The HEK 293 cell line has been reported to express low levels of TNFRSF1A (12); however, expression of endogenous TNFRSF1A was not detected by flow cytometry. Cells were stably transfected with wild-type and mutant TNFRSF1A constructs using the transfection reagent FuGENE 6 (Roche Diagnostics) according to the manufacturer's protocol, with a ratio of 3 μl of reagent to 1 μg of DNA. Forty-eight hours posttransfection, cells were switched to selective medium (Dulbecco's modified Eagle's medium containing 10% fetal calf serum, 100 units/ml of penicillin, 10 mg/ml of streptomycin, and 5 μg/ml of blasticidin [Invitrogen Life Technologies], 250 μg/ml of zeocin [Invitrogen Life Technologies], 1 mML-glutamine, and 10 mM HEPES buffer). Transfected cells were initially selected for 28 days, after which the cells were induced to express receptor. Cultured cells were split 1:2, and 1 μg/ml of doxycycline (a derivative of tetracycline) (Sigma) was added; noninduced duplicate cells were used as controls. Eighteen hours postinduction, the cells were assayed by flow cytometry for surface-expressed TNFRSF1A, using mouse anti-human TNFRSF1A conjugated with phycoerythrin (PE) (R&D Systems, Abindgon, UK) or PE-conjugated mouse IgG1 negative control (Dako, High Wycombe, UK), and assayed on an Epics XL flow cytometer (Beckman Coulter, High Wycombe, UK). Following successful transfection, cell lines were routinely cultured in 25-cm2 flasks and grown at 37°C in a humidified 5% CO2 atmosphere until confluency. For the Δsig HEK 293 cells, clones were produced and selected as described previously (17).
Induction and detection of surface and intracellular TNFRSF1A expression.
Wild-type and mutant cell lines were induced to express TNFRSF1A, as described above. Cells were harvested, washed, and resuspended in PBA (phosphate buffered saline containing 0.5% [volume/volume] bovine serum albumin and 0.1% [v/v] sodium azide). Cell suspensions were mixed with mouse anti-human TNFRSF1A–PE (R&D Systems) or mouse IgG1–PE negative control (Dako) and incubated for 30 minutes. Cells were then washed twice in PBA, resuspended in 0.5% formaldehyde fixative, and analyzed by flow cytometry on an Epics XL flow cytometer (Beckman Coulter). To determine the intracellular expression of recombinant TNFRSF1A, cell lines were induced as described above. After induction, cells were harvested, washed in PBA, and fixed in 4% formaldehyde fixative for 5 minutes at room temperature. Fixed cells were then washed in saponin buffer (PBA containing 0.1% saponin [Sigma] and 10 mM glucose) and resuspended. Subsequently, cells were mixed with mouse anti-human TNFRSF1A–PE or mouse IgG1–PE negative control and incubated for 2 hours. Finally, cells were extensively washed with saponin buffer, resuspended in 0.5% formaldehyde fixative, and analyzed by flow cytometry.
Cell cycle analysis and cell death assessment.
All cell lines used in these experiments were split 1:2 and were grown in the presence (induced cells) or absence (control, noninduced cells) of 1 μg/ml of doxycycline for 18 hours. Cells were harvested, washed with PBA, fixed with ice-cold 70% (v/v) ethanol, and incubated at −20°C for 1 hour. The cells were then washed twice with PBA, resuspended in PBA containing 100 μg/ml of RNAse A (Sigma) and 50 μg/ml of propidium iodide (PI) solution (Sigma), and were incubated for 20 minutes. Cell samples were then analyzed on an ALTRA flow cytometer (Beckman Coulter). PI fluorescence is emitted at 600–700 nm and was detectable in the PMT4 detector of the ALTRA. Plotting of the PMT4 area versus the PMT4 peak was then used to gate single cells to produce a PMT4 histogram showing cell cycle peaks and subdiploid DNA-containing cells. The percentage of subdiploid (apoptotic) cells from the displayed cell population was determined.
Analysis of TNFα binding.
HEK 293 cell lines expressing R347A TNFRSF1A were split 1:2 and grown in the presence of 1 μg/ml of doxycycline for 18 hours. Recombinant human TNFα (R&D Systems) was added at a final concentration of 50 ng/ml, and the cells were incubated at 37°C for 2 hours. Following incubation with TNFα, cells were harvested, washed in PBA, and resuspended. Cells were then stained with primary antibodies, either mouse anti-human TNFα or mouse IgG2a negative control (Dako), and incubated on ice for 30 minutes. Subsequently, cells were washed twice with PBA, resuspended, and incubated with secondary antibody, goat anti-mouse F(ab′)2–fluorescein isothiocyanate (Dako), for 30 minutes on ice. Finally, cells were washed twice in PBA, resuspended in 0.5% formaldehyde fixative, and analyzed by flow cytometry. Duplicate cells were stained for surface expression of TNFRSF1A, as described above, to confirm receptor induction by doxycycline. Controls showed no nonspecific binding of secondary antibody.
Homology modeling of TRAPS-related TNFRSF1A mutants.
The 1TNR Protein Data Bank file (online at http://www.rcsb.org/pdb/) was used. Models for the TNFRSF1A mutations were produced using the receptor monomer chain from the crystal structure 1TNR (33) as a template. Comparative protein structure modeling was performed with Modeller 8v1 (34) (online at http://salilab.org/modeller/). Mutated side chains were fitted using an automatic procedure that minimizes violations of the spatial restraints. After each single mutation was introduced, mutated and nonmutated (wild-type) protein structures were initially subjected to 2,000 steps of conjugate-gradient minimization (34–36). Output structures from Modeller were further minimized and equilibrated in order to remove any bad contacts and optimize solvent–protein interactions, using the NAMD program (37).
Structures were solvated in a rectangular water box measuring 98 × 92 × 110 angstroms (for each dimension, 10 angstroms greater than the protein size). The resulting structure was then minimized for an additional 100 cycles. Each of the solvated files was equilibrated for 5 picoseconds at a constant temperature of 310°K (37°C). Equilibrated output data were used as input for molecular dynamics simulations. Simulations were run under periodic boundary conditions, and the SHAKE algorithm (38) was used with a time step of 1 femtosecond. The particle mesh Ewald method (39) was employed to calculate the long-range electrostatic interactions, using a grid (100 × 96 × 112 angstroms). Structure validation was done with the MolProbity system (40) (online at http://kinemage.biochem.duke.edu/molprobity/index-king.html). Three-dimensional vector graphics and 3-D renderings of proteins were produced using the VMD 1.8.2 program (41) (online at http://www.ks.uiuc.edu/Research/vmd/) and POV-Ray version 3.6 (online at www.povray.org/), respectively.
DD signaling–deficient TRAPS-related mutants of TNFRSF1A.
In order to obtain an in vitro system able to model the consequences of TRAPS-related TNFRSF1A mutants as a function solely of the mutations in the TNFRSF1A ectodomains, a point mutation at position R347A of the mature TNFRSF1A protein was used. The R347A mutation has previously been reported to abolish interactions with the wild-type DD and hence abolish TRADD and SODD binding without affecting the general protein structure and folding (32). Recombinant DNA TNFRSF1A constructs with the R347A mutation were produced as described in Materials and Methods, for both wild-type and the following selected mutant forms of TNFRSF1A: R92Q, T50M, C33Y, and C52F. These constructs were in a doxycycline-inducible expression vector and were stably transfected into 2 cell types: the HEK 293 cell line and the human liver adenocarcinoma SK–HEp-1 endothelial cell line. Stable transfections with related full-length and Δsig TNFRSF1A constructs into HEK 293 cells had already been established as an in vitro system with which to study potential mechanisms related to TRAPS (16, 17). (The Δsig construct has a stop codon at residue 215, numbered from the start of the mature TNFRSF1A sequence, leaving a 10-residue cytoplasmic tail.) The SK–HEp-1 endothelial cell line was also used to generate an in vitro model that may be more representative of patients' cells, because this cell line has detectable levels of endogenously expressed TNFRSF1A. Also, endothelial cells are likely to play a central role in the tissue inflammation exhibited by patients with TRAPS.
TNFRSF1A can transduce signals for both cell death (apoptosis) and NF-κB activation (8–13). We have previously shown that HEK 293 cells expressing recombinant wild-type or mutant forms of TNFRSF1A produce interleukin-8 and undergo apoptosis (17). TNFRSF1A-mediated apoptosis was therefore investigated in transfected cells in order to determine whether the DD signaling of the recombinant R347A constructs was impaired when compared with that of their full-length counterparts. Transfected cell lines were induced to express recombinant TNFRSF1A with 1 μg/ml of doxycycline and subsequently stained with PI and analyzed by flow cytometry to detect the cellular DNA content.
Noninduced appropriately matched controls were also analyzed. Apoptotic cells were defined in cell cycle analysis profiles as those with a subdiploid DNA content. The percentage of apoptotic R347A TNFRSF1A–transfected HEK 293 and SK–HEp-1 cells was very similar for both induced and noninduced cells (Table 1); in contrast, the percentage of apoptotic full-length TNFRSF1A-transfected HEK 293 and SK–HEp-1 cells was much higher for induced cells than for noninduced cells (Table 1). HEK 293 cells were more prone to apoptosis than were SK–HEp-1 cells. The observed differences between the full-length and R347A TNFRSF1A–transfected cells is attributable to the absence of DD signaling rather than the lack of recombinant TNFRSF1A expression.
Table 1. Percentage of apoptotic cells induced or not induced by doxycycline to express recombinant wild-type or mutant forms of TNFRSF1A*
TNFRSF1A recombinant constructs behave differently from each other and have different properties within different cell lines.
Expression profiles for transfected HEK 293 cell lines were compared for full-length, Δsig, and R347A TNFRSF1A constructs (Figures 2A–C). R347A constructs differ from full-length constructs only by the point mutation at aa 347. In contrast, Δsig constructs have a more drastic structural change, because almost the entire cytoplasmic signaling domain is missing. Results with the full-length and Δsig forms of recombinant TNFRSF1A were consistent with those previously reported (4, 16, 17). Thus, the full-length wild-type and R92Q TNFRSF1A showed significant cell surface expression, with a proportion also being intracellular, whereas full-length T50M, C33Y, and C52F were retained in the cytoplasm, with essentially no surface expression (Figure 2A). A large proportion of Δsig wild-type and R92Q TNFRSF1A were expressed on the cell surface, but also, for the Δsig constructs, a lower proportion of T50M, C33Y, and C52F were surface-expressed (Figure 2B). The pattern of expression of R347A TNFRSF1A was very similar to that seen with Δsig TNFRSF1A, i.e., most wild-type and R92Q was surface-expressed, with lower surface expression of T50M, C33Y, and C52F (Figure 2C). This indicates that the influence of the cytoplasmic region in promoting intracellular retention of most of the mutant forms of TNFRSF1A is primarily dependent on properties of the DD (42).
Compared with the HEK 293 transfectants, the SK–HEp-1 transfectants showed less surface expression of recombinant TNFRSF1A (Figure 3). However, the qualitative pattern of expression was similar between the 2 cell lines. Thus, in SK–HEp-1 transfectants, a small proportion of full-length wild-type and R92Q TNFRSF1A was surface-expressed, although there was slightly less surface expression of full-length R92Q than full-length wild-type TNFRSF1A despite similar (or slightly higher) total expression of the R92Q protein. Essentially, no full-length T50M or C33Y TNFRSF1A was surface-expressed (Figures 3A and B). For the R347A constructs, a proportion of wild-type and R92Q TNFRSF1A was surface-expressed, as was a very small amount of T50M and C33Y TNRSF1A, with the majority being retained intracellularly (Figures 3C and D). Expression levels for recombinant R347A TNFRSF1A in SK–HEp-1 cells were ∼10-fold lower than the high expression levels seen in the HEK 293 cell line (Figures 2 and 3).
Differences in the ability of TNFRSF1A R347A mutants to bind TNFα.
TNFα binds to the ectodomain of TNFRSF1A. Because R347A recombinant receptors are expressed on the cell surface in the HEK 293 cell line, the ability of the wild-type and different mutant recombinant R347A receptors to bind TNFα was investigated. Only wild-type and R92Q, but not T50M or the cysteine mutants, bound TNFα (Figure 2D). The inability to bind TNFα was related to the effect of the ectodomain mutations in the receptor's binding site for TNFα rather than lack of cell surface expression of the receptor. Thus, even for the T50M mutant, which had a relatively high level of surface expression, no binding of TNFα was observed (Figure 2D).
R92 mutations and the link between structural change and behavior of mutant TNFRSF1A.
The R92Q mutation is a low-penetrance mutation, and recombinant R92Q TNFRSF1A behaves in a manner similar to that of wild-type TNFRSF1A in our transfection model. In the R92Q TNFRSF1A protein, the arginine basic side chain at position 92 is replaced by glutamine (which is an uncharged polar side chain), with no predicted major impact on the overall protein structure. To further investigate the link between receptor functionality and protein conformation in this system, 3 single amino acid changes (R92E [E = glutamic acid, an acidic side chain], R92K [K = lysine, a basic side chain], and R92P [P = proline, a hydrophobic neutral side chain]) were introduced by site-directed mutagenesis at residue 92 of the TNFRSF1A protein. The 3 additional mutations were chosen for their potential ability to cause conformational changes to the overall TNFRSF1A protein conformation when compared with the wild-type TNFRSF1A protein. Of these mutations, R92P has been reported to be associated with TRAPS, while R92K and R92E have not been reported in patients with TRAPS.
Recombinant DNA full-length TNFRSF1A constructs with the above-mentioned R92 mutations were produced as described in Materials and Methods. Recombinant TNFRSF1A expression was determined in both doxycycline-induced and noninduced cells and analyzed by flow cytometry. The doxycycline-induced HEK 293 wild-type and all R92 cell lines (except R92P) expressed recombinant TNFRSF1A at the cell surface (Figure 4). Also, total levels of expressed R92P TNFRSF1A were much lower than those of wild-type or the other R92 mutant TNFRSF1A proteins. This seems to be a direct effect of the R92 mutation itself, because 3 independent rounds of stable transfections were carried out, and all constructs expressed similarly low levels of recombinant R92P TNFRSF1A protein (data not shown). It is also possible that the particular anti-TNFRSF1A antibody used does not bind well to R92P. Levels of transgene expression were again lower in the SK–HEp-1 cells, and very little of the mutant recombinant TNFRSF1A was expressed at the cell surface (Figure 4). R92P was not detected at all in SK–HEp-1 cells.
Model of the ectodomains of wild-type and mutant forms of TNFRSF1A.
Homology modeling of TRAPS-associated TNFRSF1A mutants was carried out to test the hypothesis that the abnormal behavior observed in vitro of most TRAPS-associated mutants is attributable to the effect that the mutations have on overall protein conformation. The structure of the extracellular domain of the TNFRSF1A receptor has been thoroughly investigated and, as a result, 3 crystal structures are available in the Protein Data Bank depository: 1EXT (43), 1TNR (33), and 1NCF (44). The 1TNR file is the only heterodimer between ligand (TNFβ) and TNFRSF1A and also contains data to model the TNFRSF1A trimer, which is the complex responsible for the signaling properties of TNFRSF1A. The 1TNR molecule was therefore chosen as a template for homology modeling. In order to simplify the model and more accurately look at the impact of single mutations in the ectodomains of TNFRSF1A on overall protein structure, only the monomer of the TNFRSF1A extracellular domain was used. A monomer Protein Data Bank file was produced containing only the chain R of the original 1TNR Protein Data Bank file, thus corresponding to the extracellular domain of TNFRSF1A.
Structural differences of wild-type and mutant TNFRSF1A proteins.
Model structures of the different R92 mutant proteins showed that even a single amino acid substitution at a residue not involved in receptor–ligand interaction can have an overall impact on protein folding (Figure 5). The backbone and side chains around the mutated residues are different for the wild-type and the R92 molecules (Figure 5A). However, the change extends beyond the CRD bearing the mutation. Following mutation and energy minimization, superimposition of all the molecules showed that the whole 3-D structure was altered as a result of the amino acid substitutions at position 92 (Figure 5B); in many respects, this was most marked for R92P (shown in red in Figure 5B), which also appeared most abnormal in the R92 mutant transfectants (Figure 4).
The structural differences resulting from the single amino acid changes occurred across the 4 ectodomains, as seen by the changes to the protein surface (Figure 5C; buried residues are shown in yellow, and surface residues are shown in blue). As opposed to wild-type TNFRSF1A, in TNFRSF1A mutants different residues are buried or exposed to the surface across the whole protein and are not just localized to the ectodomains affected by the mutation (Figure 5C). Examples of residues that change markedly across wild-type and mutant TNFRSF1A include Gly-47, Ile-28, Cys-88, and Gly-142 (Figure 5C). The mutants T50M, C33Y, and C52F all have buried residues in one of the receptor's ligand-binding pockets at residues 54–61, which in the wild-type (and also the R92Q) molecule are surface-exposed (Figure 5C), offering a possible explanation for the fact that T50M, C33Y, and C52F do not bind TNF.
Another indication of strain in the structural organization of mutant versus wild-type TNFRSF1A is the root-mean-square distance (RMSD) values, which are a numeric measure of the difference between 2 structures. All of the TNFRSF1A mutant proteins showed a higher RMSD value compared with the wild-type protein. The increasing RMSD values corresponded to the wild-type and mutant proteins as follows: 0.912 (wild-type), 0.925 (R92E), 0.929 (R92K), 0.933 (R92Q), 0.934 (T50M), 0.938 (R92P), 0.938 (C88Y), and 0.947 (C33Y and C52F).
Cysteine mutations affect areas involved in the disulfide bonds and therefore are expected to destabilize the protein more than the noncysteine mutations. This is reflected by the average RMSD values as well as the RMSD for the whole protein, which gives a visual indication of structural tension in the 3-D structure caused by the mutations (Figure 5D; the most stable residues are shown in blue, and the least stable residues with the greatest mobility are shown in red). The molecule was quite “mobile” through the equilibration steps. As shown in Figures 5C and D, in the wild-type protein, some of the blue segments (i.e., more rigid and therefore less mobile) correspond to areas of cysteine bonds (e.g., Cys-88, which changes in the T50M, C33Y, and C52F compared with wild-type). The changes seen in residue mobility in the mutant TNFRSF1A proteins in relation to the wild-type illustrate how the overall structural organization of the protein is strained as a result of the single amino acid mutations (Figure 5D). Interestingly, rigid (stable) residues in the wild-type protein encompass certain areas of the ligand-binding pocket (Glu-56, Phe-60, Cys-73, and Gly-81) or areas near it (Glu-54, Cys-55, Ser-74, Lys-78, and Cys-88) (Figure 5D) (7). These areas are also changed in the mutants when compared with wild-type TNFRSF1A (Figure 5C).
The causal relationship between mutations in the ectodomains of TNFRSF1A and TRAPS is well established (1). However, a causal link between TRAPS-associated TNFRSF1A mutations and disease progression that might explain the mechanism by which the periodic fevers are triggered has not been clearly identified (16–18). The fact that TNFRSF1A signaling functions are quite complex, with intracellular domains reported to signal through a myriad of different cellular pathways (45), presents a problem when trying to correlate the structural effects of TRAPS-associated ectodomain mutations with receptor function.
We addressed this problem by creating an in vitro system with 2 different transfected cell lines expressing TNFRSF1A with the R347A mutation, which abolishes signaling through the DD of the receptor (31, 32). This allowed investigation of the structural effects of TRAPS-related mutations exclusive of the complicating effects of the receptor signaling function, but without losing the cytoplasmic structure as occurred with the Δsig constructs we used previously (16, 17). Therefore, the differences observed in the behavior of the different cell lines transfected with wild-type and selected mutant forms of TNFRSF1A are a direct reflection of the effect of the ectodomain mutations in this system. Enhanced apoptosis and cell death due to receptor self-aggregation and spontaneous signaling through the DD have previously been reported in systems in which high levels of receptor expression overload the cell (46). It has been shown previously that HEK 293 cells transfected with full-length versions of the wild-type or mutant forms of TNFRSF1A undergo apoptosis (17). The fact that this does not occur with the R347A mutants confirms that the signaling ability of the DD has been abolished.
A direct comparison between the 3 types of constructs (i.e., full-length, R347A, and Δsig) has allowed us to identify the DD as the region responsible for intracellular retention in this model, which is consistent with previously published results (42). When the DD is either impaired (R347A mutants) or is totally missing (Δsig mutants), recombinant TNFRSF1A is expressed at the cell surface to a higher extent than that in the full-length mutants in HEK 293 cells.
The use of 2 different cell types makes the in vitro model more representative of a biologic system. The behavior of SK–HEp-1 endothelial cells is of particular interest as a parallel to what might happen in the endothelial cells of patients. This cell line proved to be particularly sensitive to the TRAPS-associated TNFRSF1A mutations; most mutant forms of TNFRSF1A are trapped intracellularly. It is possible that this cell type is particularly sensitive to misfolded proteins and therefore targets them to intracellular degradation pathways. SK–HEp-1 cells also demonstrate lower levels of transgene expression than those achieved in HEK 293 cells. In particular, slightly lower surface expression of full-length R92Q mutant TNFRSF1A (compared with wild-type) was observed in SK–HEp-1 cells, indicating that there are subtle abnormalities in the behavior of this mutant receptor.
The different behavior seen in the mutant (compared with wild-type) TNFRSF1A R347A cell lines, i.e., reduced surface expression and TNFα binding, is much more pronounced for mutations that have a more drastic effect on the overall 3-D conformational structure of the receptor. The R92 mutations, which have a low impact on the protein backbone (R92Q, R92E, and R92K), can be seen as fairly conservative in terms of structural changes to the overall receptor conformation. These mutations exhibit properties similar to those of wild-type TNFRSF1A. In contrast, the cysteine mutations (C33Y and C52F) cause disruption in the highly conserved disulfide bonds that are essential in determining overall protein structure and have a more radical change in behavior: despite reaching the cell surface in the HEK 293 R347A–transfected cell lines, they do not have the ability to bind TNFα.
The R92P protein is predicted from homology modeling to have a high impact on overall protein conformation. Indeed, it has the highest (most abnormal) RMSD value of all the noncysteine mutations we analyzed. It is detected only at very low levels intracellularly in the HEK 293 cell line and not at all on the cell surface. It is possible that the R92P mutant protein is marked for degradation immediately after production due to having a 3-D structure that is recognized by the cell as being unable to fold properly. It is also possible that the epitope recognized by the detection of monoclonal antibody is partially destroyed. However, we have observed that a polyclonal anti-TNFRSF1A antibody also demonstrates low staining of R92P TNFRSF1A in the HEK 293 transfectants, again suggesting low levels of expression. In view of the more profound effects of the R92P mutation compared with those of R92Q on the conformation of TNFRSF1A, it is noteworthy that both R92Q and R92P occur in patients with TRAPS, but only R92Q has also been identified in healthy individuals.
Comparative modeling between the wild-type and mutant forms of TNFRSF1A show that mutations in the ectodomains affect the overall structure of the TNFRSF1A protein, altering its overall conformation. This may cause the receptor to misfold and exhibit aberrant behavior. In a recent study (47), a similar strategy linked a mutation in death receptor 3 with rheumatoid arthritis. A growing number of protein conformational disorders, e.g., cystic fibrosis, Alzheimer's disease, and Parkinson's disease, have been reported (25); such disorders result in misfolding of the mutant proteins, causing abnormal signaling when compared with their properly folded wild-type counterparts, leading to disease (22–27, 48, 49). Also, as changes occur across the entire extracellular domain, it is possible that, in some mutants, the site for TNFRSF1A cleavage is altered, leading to shedding abnormalities of those mutants (50). Furthermore, changes in conformation within the ligand-binding pockets (Figure 5C) may account for the lack of TNF binding by most of the receptor mutants.
In summary, the phenotype changes observed in the R347A ectodomain mutants of TNFRSF1A and in the overall structure predicted by homology modeling of the mutant protein indicate that structural abnormalities of mutated TNFRSF1A are important in the pathophysiology of TRAPS.
We thank Duncan Hopkins for his valuable help with computer modeling.