Results
To identify candidate Trx1 target proteins on the cell surface of lymphoid cells, we applied a trapping technique based on the reaction mechanism. This approach makes use of the finding that mutant thiol-dependent oxidoreductases lacking the C-terminal cysteine of the CXXC active site motif form long-lived mixed disulfide intermediates with target proteins. Thus, target proteins remain covalently linked to the mutant oxidoreductase and become amenable to isolation and analysis (principle shown in Figure 1A). Kinetic trapping has been applied previously to identify interaction partners of Trx family members in plants (Motohashi et al, 2001) and in the secretory pathway of human lymphocytes (Dick and Cresswell, 2002; Dick et al, 2002). In these studies, the CXXC-based trapping technique identified both established and novel target proteins, subsequently confirmed by independent techniques, demonstrating the competence of this technique to identify bona fide interaction partners.

Mechanism-based kinetic trapping can be applied to human Trx1
To determine whether kinetic trapping can be applied to human Trx1, we created recombinant wild-type and mutant Trx proteins, each equipped with a C-terminal dual affinity tag composed of a streptavidin-binding peptide (SBP) and a hexahistidine tag. To create a trapping mutant, the second cysteine of the 32CXXC35 motif was replaced by serine (C35S). Trx1 harbors three additional cysteine residues distal to the active site (cysteines 62, 69 and 73). As these residues are dispensable for catalytic activity but may cause oxidative inactivation by either intra- or intermolecular disulfide bond formation (Casagrande et al, 2002; Watson et al, 2003), we also created mutants containing amino-acid substitutions for those additional cysteines (Figure 1B; CCCCC, CCAAA, CSCCC, CSAAA and SSAAA annotate the identity of residues 32, 35, 62, 69 and 73).
To test whether Trx1(C35S)-based trapping is capable of identifying known Trx1 target proteins, Trx1(CSAAA) was allowed to react with cytosolic proteins released from digitonin-permeabilized cells. Incubation led to the formation of a reproducible pattern of distinct mixed disulfide conjugates as visualized by silver staining of a SDS–PAGE gel under non-reducing conditions (Figure 1C, lane 7). In accordance with the trapping mechanism, conjugation strictly depended on the availability of the N-terminal thiol (Cys-32) and the concurrent absence of the C-terminal thiol (Cys-35), as wild-type or cysteine-free Trx1 did not capture any proteins (Figure 1C, lanes 3, 5 and 9). The pattern of trapped proteins was not significantly influenced by the presence or absence of the non-catalytic cysteines (data not shown). The main cytosolic interaction partner of Trx1(CSAAA) was identified as peroxiredoxin-1 (Prx1) by liquid chromatography tandem mass spectrometry (LC-MS/MS). The Trx1–Prx1 association was further confirmed by immunoblotting (data not shown). The Trx1–Prx1 disulfide-linked conjugate is maintained under non-reducing conditions (Figure 1C, lane 7) and cleaved into its monomer constituents under reducing conditions (Figure 1C, lane 8).
To test whether Trx1(CSAAA) would also undergo authentic interactions under conditions more typical of an extracellular environment, we allowed Trx1(CSAAA) to react with human plasma proteins. To avoid nonspecific absorbance to high-abundance serum proteins, we applied Trx1(CSAAA) to a <30 kDa plasma ultrafiltrate, leading to the capture of a small number of proteins, as visualized by colloidal Coomassie staining (Figure 1D). Using LC-MS/MS, the principal interaction partner from the plasma ultrafiltrate was identified as peroxiredoxin-2 (Prx2), a well-established target protein of Trx1, recently found to be present in human plasma (Chen et al, 2004). These experiments provided proof-of-principle evidence that kinetic trapping is capable of capturing and identifying proven target proteins of human Trx1 from both intra- and extracellular environments.

Trx1 kinetic trapping is mediated by specific protein–protein interactions
To further investigate whether the capture of proteins by Trx1(CSAAA) is Trx-specific, we directly compared Trx1(CSAAA) with the corresponding trapping mutant of another member of the Trx superfamily, glutaredoxin-1 Grx1(CSAAA) (CSAAA annotates the identity of residues 22, 25, 7, 78 and 82). Grx1, like Trx1, uses its active site thiol to act as a disulfide reductase in the cytosolic environment. However, in contrast to Trx1, Grx1 is specialized in the recognition and reduction of protein–glutathione mixed disulfide bonds and forms mixed disulfide intermediates with glutathione rather than with proteins (Yang et al, 1998; Peltoniemi et al, 2006). As expected, we did not detect trapping of peroxiredoxins (or other Trx1-interacting proteins) by Grx1(CSAAA), neither on silver gels (Figure 1E, lanes 3 and 4) nor by immunoblotting (data not shown). The activity and thiol reactivity of the Grx1 trapping mutant was confirmed in independent experiments (data not shown), thus demonstrating that the mere availability of an active site thiol does not explain the profile of proteins captured by the Trx1 trapping mutant. Instead, our results support the notion that Trx-mediated reducing activity is steered toward distinct target disulfide bonds by specific protein–protein interactions.

Kinetic trapping can be applied to detect Trx1 interactions on the cell surface
Having established the Trx1 kinetic trapping approach for soluble target proteins, we asked whether the kinetic trapping technique can also be applied to the surface of intact cells in culture. Given previous indications of disulfide bond exchange between CD4 and wild-type Trx1 (Matthias et al, 2002), we asked whether kinetic trapping would enable us to detect this interaction on the surface of the CD4 positive promyelocytic cell line U937. In brief, we allowed mutant Trx1 to interact with the surface of live cells, removed unreacted oxidoreductase by washing and collected disulfide-linked Trx1 complexes from cellular lysates by streptavidin (SAv) affinity purification. We found that cell surface CD4 forms a mixed disulfide with exogenously added Trx1(CSAAA) (Supplementary Figure S1, left panel), which could be dissociated by DTT treatment (Supplementary Figure S1, right panel). This result confirmed that kinetic trapping can indeed be used to identify specific Trx1-reactive cell surface proteins and should therefore allow de novo identification of previously unknown cell surface target proteins.

Kinetic trapping on the surface of lymphoid cell lines identifies a prominent Trx1 interaction partner
Human Trx1 was first identified as an autocrine factor secreted by and acting on virally transformed lymphoid cell lines. We therefore applied cell surface trapping to a human EBV-transformed lymphoblastoid B-cell line (LCL-721.220) derived from the same parental cell line (LCL-721) as the 3B6 cell line, originally used in the description of the costimulatory factor ‘3B6-IL1', later identified as Trx1 (Wakasugi et al, 1990). Mixed disulfide complexes, which formed on the surface of LCL-721.220 cells were isolated and analyzed by Trx1-specific immunoblotting to visualize overall mixed disulfide conjugates. Interestingly, we found that Trx1 predominantly engages a single protein on the lymphoblastoid surface, suggesting a highly selective interaction (Figure 2A, lane 3). As expected, the trapping product, a mixed disulfide conjugate of about 160 kDa, was susceptible to reduction (Figure 2A, lane 4). The 160 kDa conjugate did not form on the surface of the EBV-negative Burkitts lymphoma cell line BL-41 (Figure 2B). In contrast, a conjugate of the same size was found to be formed on the surface of CCRF-CEM T cells (Figure 2C) and YT large granular lymphoma cells (data not shown). Pretreatment of the cell surface with the alkylating agent iodoacetamide (IAA) did not interfere with the formation of the 160 kDa mixed disulfide conjugate, thus confirming that the conjugate was formed by the expected disulfide exchange mechanism (rather than by de novo disulfide bond formation between two thiol groups).

CD30/TNFRSF8 is the principal Trx1 sensitive receptor on various lymphoid cell lines
To identify the unknown Trx1 target protein, we performed cell surface trapping on a larger scale (5 × 109 LCL-721.220 cells), purified the Trx1-interacting surface protein by SAv affinity purification and visualized the protein by colloidal Coomassie staining (Figure 3A, left panel). The 160 kDa band was absent in the control precipitation with Trx1(CCAAA). Corresponding bands from non-reducing and reducing lanes were subjected to tryptic digestion and LC-MS/MS analysis. From both samples the unknown protein was identified as TNFRSF8 (CD30), a member of the TNFR superfamily. To validate the direct covalent interaction between Trx1(CSAAA) and CD30, an aliquot of trapped complexes from the same experiment was separated under non-reducing and reducing conditions and subjected to immunoblotting analysis with anti-Trx1 (Figure 3A, middle panel) and anti-CD30 antibodies (Figure 3A, right panel), respectively. The observed mobility difference between non-reducing (NR) and reducing (R) lanes demonstrated the formation of a mixed disulfide conjugate (Figure 3A, right panel). Additional immunoblotting experiments demonstrated that low nanomolar concentrations of Trx1(CSAAA) are sufficient to detect the interaction with CD30 (Figure 3B) and also confirmed that trapping of CD30 depends on the N-terminal cysteine of the CXXC motif (Figure 3C, lanes 1–8). Application of the Grx1 trapping mutant under the same conditions did not lead to its conjugation to CD30 (Figure 3C, lanes 9 and 10). Immunoblotting and flow cytometry experiments confirmed that BL-41 cells, unlike the other lymphocytic cell lines, do not express CD30 (Supplementary Figure S2), thus explaining the absence of the 160 kDa conjugate band (Figure 2B).

Trx1 discriminates between CD30 and other members of the TNFR superfamily
Ectodomains of TNFR superfamily proteins typically are composed of one to four cysteine-rich domains (CRDs), each normally harboring three disulfide bonds (Bodmer et al, 2002). To exclude the possibility that Trx1 interacts with CRDs uniformly, we tested whether Trx1 discriminates between distinct members of the TNFR superfamily. As shown in Figure 4A, CD95 (TNFRSF6) did not form a mixed disulfide conjugate with Trx1(CSAAA) on the same cells under identical conditions. The same result was obtained for the epidermal growth factor receptor (EGFR), which contains a total of 25 ectodomain disulfide bonds, and is expressed at substantial levels on A431 cells (Gill and Lazar, 1981) (Figure 4B). These findings indicate that Trx1 reactivity of cell surface receptors is selective and is not determined by the presence of CRDs or the number of ectodomain disulfide bonds. To further exclude that Trx1 reactivity of receptors is determined or limited by surface expression levels, we ectopically overexpressed CD30 or other TNFR superfamily members under control of the same promoter in HeLa cells and analyzed Trx1 cell surface trapping by indirect immunofluorescence. While mock-transfected HeLa cells did not capture Trx1(CSAAA) on their surface (Figure 4C, lower row), expression of CD30 led to a strong Trx1 surface association and colocalization of both proteins (Figure 4C, upper row). In contrast, expression of CD95 (Figure 4D), TNFR1 or NGFR (data not shown) did not promote Trx1 interactions with the cell surface, further strengthening the notion that Trx1 reactivity is a specific property of CD30.
The domain structure of human CD30 differs from other members of the TNFR superfamily and from its murine homologue by the presence of two additional CRDs, arising from the internal duplication of two exons (Burgess et al, 2004). We asked whether this unusual feature might confer Trx1 reactivity to human CD30 and tested whether the shorter murine CD30 could also interact with Trx1. As shown in Figure 4E, murine CD30 expressed on the Rauscher murine leukemia virus-induced T-cell lymphoma line RMA is efficiently targeted by Trx1(CSAAA), thus demonstrating that the additional CRDs in human CD30 are not required for Trx1 reactivity. The result also suggests that the enzymatic affinity of Trx1 for CD30 has been conserved during mammalian evolution.

Trx1 catalyzes disulfide bond reduction in CD30 and induces a conformational change disrupting the Ki-1 epitope
To demonstrate by an independent method that Trx1 attacks and breaks a disulfide bond in CD30, we used thiol-specific cell surface biotinylation to verify that Trx1 activity generates free thiols within the CD30 ectodomain (Supplementary Figure S3). Subsequent analysis of CD30 cell surface expression by flow cytometry and fluorescence microscopy revealed that a brief treatment of CD30+ cells with wild-type Trx1 led to the complete loss of CD30 recognition by the Ki-1 antibody (Figure 5A, upper panel and Figure 5B, second column). Similar results were obtained when another monoclonal antibody against CD30, MAB229 (R&D Systems; Clone 81337), was used to examine CD30 expression (Figure 5A, middle panel). Under the same conditions, recognition by the Ber-H2 antibody was only slightly affected (Figure 5A, lower panel and Figure 5B, third column), indicating that Trx1-mediated disulfide bond reduction (and possibly rearrangement) induces a structural alteration in the CD30 ectodomain, which disrupts or conceals the Ki-1 epitope.
Pursuant to the observation that recognition of CD30 by antibodies Ki-1 and MAB229 is affected by CD30 redox state, we used flow cytometry to analyze the response of CD30 to reduction in greater detail. Addition of oxidized Trx1 did not influence antibody reactivity of CD30 (data not shown). When reduced Trx1 was applied in the absence of a regenerating system, a conformational change in CD30 could be observed, but remained incomplete (data not shown). Complete and sustained loss of antibody reactivity required a source of reducing equivalents for the oxidoreductase. Both DTT and Trx reductase (TrxR)/NADPH were found to be effective as regenerating systems. Importantly, neither DTT nor TrxR/NADPH had an effect when applied in the absence of Trx1 (Figure 5A, middle panel and data not shown). Wild-type Grx1 (Figure 5C, left panel) and the redox-inactive mutant of Trx1 (Figure 5C, middle panel) did not alter the redox-sensitive CD30 epitopes. Other cell surface receptors, for example CD28, analyzed in parallel on the same cells were unaffected by Trx1 treatment (Figure 5C, right panel). Loss of CD30 antibody recognition typically occurred within minutes (Figure 5D, upper panel). Testing the influence of Trx1 concentration under the same conditions, consistent effects on CD30 conformation became evident at concentrations around 100 nM (Figure 5D, lower panel).

Trx1-mediated disulfide exchange interferes with CD30 receptor–ligand interactions
The observation that antibody binding to CD30 is influenced by Trx1 suggested a redox-dependent conformational change within the CD30 ectodomain. To test whether Trx1-mediated reduction of CD30 influences binding of CD30 to its ligand (CD30L), we analyzed the interaction between CD30 and recombinant soluble CD30L (sCD30L) on the cell surface by flow cytometry. A brief incubation of CD30+Hodgkin's lymphoma HDLM-2 cells with wild-type Trx1 led to a substantial loss in CD30L binding to the cell surface (Figure 6A). A similar result was obtained for the large granular lymphoma cell line YT (Supplementary Figure S4). The same effect was evident in the absence of an exogenously added reducing system (Supplementary Figure S5), thus demonstrating that under the given conditions Trx1–substrate interactions are not limited by oxidative inactivation.
As shown by fluorescence microscopy, sCD30L binds to the surface of CD30-transfected HeLa cells and colocalizes with its receptor (Figure 6B, upper row). Treatment of HeLa cells with Trx1(CCAAA) (Figure 6B, middle row) but not a redox-inactive mutant (Figure 6B, lower row) strongly interferes with CD30L binding and colocalization.

Trx1-mediated reducing activity interrupts agonist-induced CD30 signaling in YT cells
Given its influence on ligand binding, we reasoned that Trx1 might interfere with CD30-mediated signaling. To test whether Trx1-mediated conformational alteration of CD30 affects CD30-dependent cellular responses, we made use of the YT large granular lymphoma line that has previously been used to study signals emanating from CD30 and to define the genes regulated by such signals (Muta et al, 2000). Consistent with previous results (Bowen et al, 1993), we observed that stimulation of CD30 with either agonistic antibodies or sCD30L led to upregulation of the IL-2Rα chain (CD25) within 24 h (Figure 7A, lower panel, compare columns 3, 4 and 7). Treatment of YT cells with Trx1(CCAAA) (but not with redox-inactive Trx1) prevented CD25 upregulation (Figure 7A, lower panel, columns 5 and 8), concomitant with the redox change in CD30 (Figure 7A, upper panel, column 4), thus demonstrating that the redox interaction between Trx1 and CD30 has a pronounced influence on CD30-mediated gene expression.
YT cells respond to CD30 signals by downregulating the expression of cytotoxic effector molecules, including FasL, thus decreasing their cytotoxicity against Fas-expressing target cells (Bowen et al, 1993; Muta et al, 2000). To test if Trx1 influences CD30-mediated suppression of cytotoxicity, we treated YT cells with Trx1(CCAAA) or Trx1(SSAAA) before stimulation with agonistic anti-CD30 antibody and quantified cytotoxicity against Cr-labeled Raji cells. Upon CD30 stimulation, cytotoxicity was markedly reduced (Figure 7B, compare columns 1 and 4). The decrease in cytotoxicity was completely reversed by Trx1(CCAAA) but not the catalytically inactive mutant (SSAAA) (Figure 7B, columns 5 and 6). As judged by RT–PCR, changes in cytotoxicity correlated with changes in FasL mRNA expression (Figure 7B, lower panel). These results confirm that the catalytic activity of Trx1 modulates CD30-dependent changes in cellular behavior and function.