In Vitro Translation and Assembly of a Complete T Cell Receptor–CD3 Complex

The cDNAs encoding the HA1.7 TCR-α and -β chains were provided by Dr. M.J. Owen (Imperial Cancer Research Fund, London, UK; reference 25). We obtained the cDNAs encoding CD3-γ, -δ and -ε (2–4) and TCR-ζ (5, 26) from Dr. C. Terhorst (Beth Israel Hospital, Boston, MA). The cDNAs encoding HLA-DR1-α (27) and -β chains (28) were a gift from Dr. Trowsdale (Human Immunogenetics Group, Imperial Cancer Research Fund, London, UK). The sequences which encode the cleavable signal peptide of the TCR-α, -β, CD3-γ, -δ, -ε, TCR-ζ, and HLA-DR1–β were exchanged for the signal peptide encoding sequence of the mouse class I heavy chain H2-K^b. The CD4 encoding cDNA (29) was cloned into pSP64 (Promega Corp., Madison, WI) under the control of the SP6 promotor. All other cDNAs were cloned into pSP72 (Promega Corp.) under the control of the T7 promotor.

The following mouse mAbs were used in this study: Tü36 (30) recognizes properly conformed HLA- DR1 heterodimers, OKT4 (31, 32), which is reactive against properly conformed human CD4 and OKT3 (31), recognizes CD3 (both obtained from American Type Culture Collection, Rockville, MD). JOVI.1 and JOVI.3 were raised against human TCR-β (Vβ3Cβ1; 33) and TIA-2 recognizes preferentially human TCR-ζ homodimers (34; gift from Dr. P. Anderson, Dana-Farber Cancer Institute, Boston, MA).

The following polyclonal rabbit antisera were used: antisera raised against the cytoplasmic tails of CD3-γ, -δ and -ε were provided by Dr. J. Coligan (National Institutes of Health, Bethesda, MD) and the antisera raised against the human TCR-α and -β chain (35) were a gift from Dr. O. Acuto (Institut Pasteur, Paris, France). The rabbit polyclonal antiserum against the COOH terminus of canine calnexin was purchased from StressGen Biotechnologies Corp. (Victoria, Canada).

The HLA-DR1–presentable peptide from influenza virus HA306–318 (36) has been synthesized in the lab by 9-fluorenylmethoxycarbonyl chemistry. Cycloheximide and iodoacetamide were purchased from Sigma Chemical Co. (St. Louis, MO).

SDS-glycine electrophoresis (SDS-PAGE) was performed as described (37). SDS-glycine gels were fluorographed using DMSO–2,5-diphenyl oxazole (Sigma Chemical Co.) and exposed to XAR-5 films (Kodak, Rochester, NY).

In vitro transcriptions were performed using T7 or SP6 RNA polymerase (Promega Corp.). RNA was equipped with a G⁵′ppp⁵′ N-cap structure either co- or posttranscriptionally according to the vendor's instructions (Promega Corp.). RNA was stored in water at −80°C. The optimal amount of RNA for translation was determined empirically for each stock of RNA prepared. Before its use in translations, the RNA was heated (80°C) for 2 min and subsequently chilled on ice.

In vitro translations were previously described (38). Reaction times were indicated in the results section. Reticulocyte lysate without supplemented dithiothreitol (DTT) (Flexi™) was purchased from Promega. Microsomes were prepared from dog pancreas as described (39).

After translation, microsomes were pelleted by centrifugation (5415C centrifuge; Eppendorf North America, Inc., Madison, WI; 14,000 rpm, 10 min) and subsequently lysed for 30 min on ice in 1 ml NP-40 lysis buffer (0.5% NP-40, 50 mM Tris-HCl, pH 7.4, 5 mM MgCl₂) or digitonin (Wako, Kyoto, Japan) lysis mix (1% digitonin, 150 mM NaCl, 50 mM Tris-HCl, pH 7.4) containing 1 mM phenylmethylsulfonyl fluoride (Sigma Chemical Co.). Insoluble debris was removed by centrifugation (14,000 rpm, 10 min). Supernatants were precleared twice with either normal rabbit or mouse serum and formalin fixed Staphylococcus aureus and then immunoprecipitated with the indicated antibodies as described elsewhere (38). Digestions with endoglycosidase H (EndoH)_f (New England Biolabs, Beverly, MA) were performed on immune complexes bound to S. aureus according to the vendor's instruction.

The translational performance of the mRNAs encoding the TCR subunits was optimized to ensure the likelihood of subsequent assembly of subunits translated simultaneously. We substituted the 5′ untranslated sequence as well as the signal peptide encoding sequence of all cDNAs with the corresponding sequence of the murine MHC class I molecule K^b, the in vitro transcript of which had proven to be translated with high efficiency (40). A comparison of the translation products of both sets of mRNAs bearing either the endogenous or the K^b signal sequence shows that in all cases this substitution led to an enhanced translation efficiency (Fig. 1). This holds true as well for the translation of CD3-δ, the nonmodified in vitro transcript that failed to be translated (data not shown). Since sequence substitution as described above did not affect the mobility of the translation products in SDS-PAGE, we performed all subsequent experiments with modified mRNAs.

In living cells, TCR-α, -β, CD3-γ and δ, but not CD3-ε and the ζ subunit are co-translationally N-glycosylated in the lumen of the ER. To confirm proper glycosylation in vitro, we translated the mRNAs of all TCR subunits individually in the presence of microsomes, followed by recovery of the translation products by immunoprecipitation using the appropriate antibodies. Immunoprecipitates were digested with EndoH or mock digested to remove N-linked glycans (Fig. 2). The number of glycans attached could be estimated for in vitro translated TCR-α (four glycans), -β, CD3-γ and -δ (two glycans), and CD3-ε and -ζ (no glycans), all in agreement with in vivo data (2–5, 25).

The formation of disulfide bridges occurs in the lumen of the ER both co- and posttranslationally, and requires an oxidizing environment. Disulfide bonds constrain the tertiary structure and help maintain the conformation of a polypeptide chain (41–43). In this study, we used conformation-specific mAbs to monitor disulfide bond–dependent folding. TCR-β and CD4 translated in the presence of the reducing agent DTT do not acquire their native conformation, demonstrated by their failure to bind to JOVI.1, JOVI.3 (anti–TCRβ mAbs) and OKT4 (anti-CD4 mAb), respectively. However, the epitopes recognized by all three antibodies were recovered upon inclusion of oxidized glutathion (GSSG) (Fig. 3 A) in a dose-dependent manner.

Similarly, disulfide bond–mediated homodimerization of the ζ subunit is dependent on proper redox conditions. In the presence of 4 mM DTT, ζ homodimers could not be detected with TIA-2 (anti-ζ mAb, reacting preferentially with dimeric ζ) but were recovered upon addition of GSSG (Fig. 3 B). We conclude that the proper formation of intra- and intermolecular disulfide bonds in vitro requires an oxidizing environment and is essential for correct folding.

Because the complexity of polypeptides resulting from in vitro translation is limited and their specific radioactivity is high, specificity of immunoprecipitation must be rigorously established. All TCR subunits were therefore translated separately (Fig. 4,A, left), and the microsomes lysed in digitonin lysis buffer. Lysates of the different translations were pooled and subjected to immunoprecipitation using the antibodies as indicated (Fig. 4,A), or lysates of the individual translations were immunoprecipitated with a mixture of antibodies lacking only the antibody specific for the translated polypeptide (Fig. 4 A, middle).

All of the antibodies used precipitated exclusively the appropriate chain from a polypeptide mix (Fig. 4,A, middle) but failed to recover inappropriate polypeptides (Fig. 4 A, right). This result also demonstrates that assembly of the TCR does not occur artefactually after lysis, but requires the presence of all subunits in a shared membrane compartment (see below).

To examine subunit-specific assembly of the TCR in vitro, we translated simultaneously all chains of the TCR as well as three unrelated polypeptides; the human CD4 coreceptor and the α and β chain of the human MHC class II molecule HLA-DR1 (DR). The peptide of the influenza hemagglutinin (HA306-318) was added to the reaction as a ligand of DR to promote the formation of peptide-bound SDS-stable MHC class II molecules, which migrate as SDS-stable heterodimer in SDS-PAGE (44). This control serves to show the ability of the in vitro system to support the proper assembly of this heterodimer, in accord with our earlier studies (40). After translation, microsomes were solubilized in digitonin lysis buffer, subjected to immunoprecipitation using OKT4 (anti-CD4 mAb), Tü36 (anti-DR mAb), JOVI.1 (anti–TCR-β mAb), or OKT3 (anti-CD3 mAb). The resulting immunoprecipitates were analyzed by SDS-PAGE under reducing conditions (Fig. 4 B, left). CD4 and DR were recovered only by their appropriate antibodies. A sizable proportion of DR displayed the SDS-stable phenotype, indicative of proper peptide loading and specific assembly. The immunoprecipitates obtained with JOVI.1 (anti–TCR-β) and OKT3 (anti-CD3) contained subunits of the TCR exclusively and were devoid of co-translated CD4 or DR.

To assess the formation of disulfide-linked subunits, complexes recovered with JOVI.1 and OKT3 were analyzed by SDS-PAGE under nonreducing conditions (Fig. 4,B, middle). We observed the presence of α/β heterodimers and ζ homodimers, as well as the absence of monomeric α, β, and ζ subunits in complexes immunoprecipitated with OKT3. To resolve TCR-α from TCR-β, immunoprecipitates from JOVI.1 and OKT3 were digested with EndoH before separation by reducing SDS-PAGE (Fig. 4 B, right). The methionine content (TCR-α, 4 methionines; TCR-β, 6 methionines) and relative intensity of both subunits recovered with OKT3 strongly suggest a 1:1 stoichiometry for both chains, in accordance with in vivo data.

We conclude that the formation of TCR complexes as monitored by coimmunoprecipitation of the relevant subunits occurs in a specific manner. The recovery of TCR-ζ with both JOVI.1 and OKT3 indicate the formation of complete TCR complexes, since this subunit joins the TCR during the last stage of assembly (10). Moreover, specific assembly of multimolecular complexes (TCR versus DR) can occur simultaneously in an in vitro translation system, as complexes containing inappropriate subunits were not observed.

The CD3-ε chain contains two cysteine residues presumably involved in stabilization of the Ig fold, and two more cysteines in the lumenal domain (2), one of which is located within a few residues of the lipid bilayer (Fig. 5). The occurrence of disulfide-linked ε homodimers has been reported in vivo (45, 46), but the monomer is the prevalent form found in living cells. We investigated the biochemical properties of CD3-ε translated in vitro under both reducing and oxidizing conditions. Immunoprecipitations were performed with the anti–CD3-ε antibody in the presence of 50 mM iodoacetamide to prevent the formation of disulfide bridges after lysis. As visualized by SDS-PAGE under nonreducing conditions, translation of CD3-ε performed under oxidizing conditions resulted in oligomeric structures that collapsed to the monomer when analyzed by SDS-PAGE under reducing conditions (Fig. 6,A). Moreover, the difference in mobility for the oxidized and reduced monomeric form of CD3-ε is clearly visible (Fig. 6 A, left). We noted that it is only the oxidized form of CD3-ε that coimmunoprecipitated with fully assembled TCR complexes. We do not know whether the more rapidly migrating, oxidized form of CD3-ε contains any remaining free sulfhydryl groups, but we consider this unlikely. Inclusion of DTT in the translation reaction abrogated oligomerization but led to a monomeric polypeptide with a slightly decreased mobility, as a consequence of alkylation by iodoacetamide upon lysis. We confirmed the homooligomeric nature of CD3-ε by two-dimensional gel electrophoresis (first dimension: isoelectric focusing under nonreducing conditions; second dimension: reducing SDS-PAGE; data not shown). We conclude that CD3-ε translated in vitro in the absence of related subunits tends to form mostly disulfide-linked oligomers, a process that appears to be redox dependent.

To investigate whether the quantity of CD3-ε expressed in microsomes exerts any influence on the formation of disulfide-linked ε oligomers, we translated CD3-ε together with increasing amounts of an unrelated chain, the DR-β chain. Total microsome-associated material as well as immunoprecipitates of CD3-ε were analyzed by SDS-PAGE performed under both reducing and nonreducing conditions (Fig. 6 B). As the ratio between monomeric and disulfide-linked oligomeric CD3-ε does not depend on the amount of translated CD3-ε (and DR-β), the formation of homooligomers must be an inherent property of the CD3-ε subunit.

We addressed the kinetics of this reaction (Fig. 6 C) by terminating the translation of CD3-ε after 15 min by the addition of cycloheximide. Aliquots of microsomes were lysed in the presence of iodoacetamide at indicated times after translation and subjected to immunoprecipitation with anti–CD3-ε antibodies. Analysis of immune complexes by SDS-PAGE performed under nonreducing conditions reveals that dimeric CD3-ε was already prevalent at the end of the translation period, but higher order forms gradually increased with time.

The propensity of CD3-ε to form disulfide-linked oligomers was examined by translating CD3-ε either alone or together with the other subunits of the TCR. Total microsome-associated material as well as immunoprecipitates obtained from lysed microsomes were analyzed by SDS-PAGE performed under both reducing and nonreducing conditions (Fig. 7 A). Lysis in 0.5% NP-40 facilitates the visualization of CD3-ε as it leads to the disruption of all noncovalent interactions between the subunits of the TCR, but not those between CD3-ε, -γ, and -δ (6).

As shown above, CD3-ε translated on its own is found primarily in oligomeric structures. However, its behavior changes dramatically in the presence of TCR subunits, where CD3-ε is now predominantly monomeric. The role of each individual subunit in “chaperoning” CD3-ε was studied by translating ε alone and pairwise with other subunits of the TCR or with DR-β. Analysis of anti–CD3-ε immunoprecipitates by nonreducing SDS-PAGE (Fig. 7 B) shows that in the presence of TCR-α, -β, -ζ and DR-β the formation of ε oligomers is not affected. However, co-translation of CD3-γ and CD3-δ drastically shifts the equilibrium towards the formation of ε monomers. The strong interaction between CD3-ε and CD3-γ or -δ (which survives solubilization in 0.5% NP-40) may account for this effect. As binding of both γ and δ to CD3-ε appears to compete with the formation of ε homooligomers, the high degree of homology between CD3-γ, -δ, and, to some degree, CD3-ε may be particularly noteworthy and suggests the presence of a common binding motif for complex formation. Since Ig-like domains are present not only in the CD3 subunits, but also in the clonotypic α/β heterodimer and in DR-β, the mere presence of Ig-like domains is clearly not sufficient to confer the ability of blocking homooligomerization of CD3-ε.

Calnexin, a molecular chaperone, binds preferentially to monoglucosylated trimming intermediates in the ER (21). Interestingly it also interacts with free CD3-ε and CD3-ε complexed with CD3-γ or -δ, although CD3-ε lacks N-linked glycans (24, 47). How calnexin binds to CD3-ε is not known. To study calnexin–CD3-ε interactions and their importance for proper folding of CD3-ε, we designed the following experiment: CD3-ε alone, ε and γ, or γ alone were translated in the absence and presence of the glucosidase inhibitor N-7-oxadecyl-dNM (7-O-dec), which blocks glucose trimming of N-linked glycans (48) and consequently prevents a glycan-dependent interaction between calnexin and glycoproteins. The change in mobility for the glycosylated γ subunit upon translation in the presence of 7-O-dec is due to glucose retention and is clearly visible on SDS-PAGE. After translation (1-h duration), microsomes were lysed in digitonin lysis buffer. Immunoprecipitates obtained with antibodies directed against calnexin, CD3-ε, and -γ were analyzed by SDS-PAGE under nonreducing conditions (Fig. 8).

Translated alone, oligomeric ε subunits could be recovered with the anticalnexin antibody regardless of the absence or presence of 7-O-dec in the translation reaction. However, the γ subunit strongly interacted with calnexin in a glycan-dependent manner, as this interaction appeared to be 7-O-dec sensitive. Monomeric CD3-ε was co-immunoprecipitated with calnexin in the presence of CD3-γ. Complete sensitivity of this interaction to 7-O-dec implies a role of CD3-ε as a bridging intermediate for CD3-γ in this binding.

Notably, 7-O-dec did not exert any effect on the oligomerization of CD3-ε translated alone or together with CD3-γ. Therefore, for both glycan-dependent and -independent modes of binding, the interaction of calnexin with the CD3-ε does itself not influence the oligomerization of CD3-ε. Hence, we conclude that the folding of ε is guided by the formation of a complex with CD3-γ and most likely also with CD3-δ, two subunits competing for binding to CD3-ε at an early stage of TCR assembly.

Our understanding of the requirements for the assembly of the TCR stems largely from studies in which the individual subunits were transfected into suitable recipient cell lines, e.g., mutant T cell lines deficient in one of the six subunits or by transfections into nonlymphoid cells. The expression of all subunits of the TCR–CD3 complex is necessary and sufficient to accomplish its assembly in the ER and its subsequent expression at the cell surface (49, 50), without obvious involvement of lymphoid-specific factors.

The analysis of heterooligomerization of the TCR-CD3 complex in living cells is complicated by the presence of widely different pool sizes of the participating subunits, and by the egress of incompletely assembled receptors out of the ER, a complication avoided in the present study. We sought to reproduce the early stages of the TCR assembly in a complete in vitro system, with particular emphasis on the role of the CD3-ε subunit. We optimized the translation efficiency of all mRNAs involved to allow the simultaneous translation of multiple mRNAs, without having to consider the confounding effects of the 5′ untranslated sequence and signal peptides on the efficiency of translation and membrane insertion. In this fashion, we successfully translated nine different mRNAs simultaneously, and, provided proper redox conditions were imposed on the ER-derived microsomes, observed specific complex formation amongst the appropriate subunits produced in vitro. Specifically, the co-translation of class II HLA-DR–α and –β subunits in the presence of an appropriate peptide ligand, along with CD4 and all subunits of the TCR–CD3 complex, resulted in the formation of properly assembled TCR–CD3 complexes, correctly folded CD4 molecules, as well as peptide-loaded class II molecules, without cross-contamination of any of these complexes by inappropriate subunits. To our knowledge, no oligomeric membrane proteins of greater complexity have been generated by in vitro translation. The ability to obtain such complexes testifies to the robustness of these microsome-supplemented in vitro systems. The actual concentrations of the TCR subunits obtained by in vitro translation are difficult to estimate, but likely to be quite low, compared to the concentrations of the proteins in the lumen of the microsomes used. The remarkably efficient assembly and folding of the polypeptides analyzed thus suggests the possibility that these processes might be limited to small numbers of “insertion sites”, where conditions for folding and assembly are favorable. Through association of the translocon with accessory proteins (N-oligosaccharyl transferase, peptidyl disulfide isomerase, calnexin), some specialization through accomplishing this goal within the ER could occur.

The sequence of subunit interactions that leads to a complete TCR–CD3 complex has been addressed for living cells (7, 9). Given the complexity of this receptor, it is perhaps not surprising that assembly occurs in a well ordered and stepwise fashion; the formation of CD3-γ/ε and δ/ε pairs takes place early and depends on the specific interaction between their ectodomains (15). Both CD3-γ and CD3-δ compete for binding to CD3-ε (51). The crucial sites in the extracellular portion of CD3-γ for its interaction with CD3-ε are also conserved in the primary structure of CD3δ, which suggests that both CD3-γ and -δ share a common binding motif in CD3-ε (15). Next, pairs of CD3-δ/ε and γ/ε recruit TCR-α and -β, respectively. Upon formation of an intrachain disulfide bridge between TCR-α and -β, trimeric complexes merge to a hexameric εγβ–αδε complex, into which the ζ homodimer is integrated as the last remaining subunit. It is noteworthy that the stoichiometry of the CD3–TCR complex is still considered tentative (8).

When CD3-ε is translated alone, it tends to form disulfide-linked homooligomers; moreover, a substantial amount of monomeric CD3-ε is found in a reduced, i.e., incompletely, folded state. The quantity of CD3-ε translated and inserted into microsomes does not influence the ratio between monomeric (reduced and oxidized) and oligomeric structures. In all likelihood, homooligomerization of CD3-ε is an inherent property of the ε subunit; it is fast and occurs under physiological redox conditions. However, co-translation of TCR subunits under conditions that support complex formation leads to a dramatic change in the balance of newly synthesized CD3-ε subunits towards oxidized and properly folded monomers, the only form of CD3-ε recovered with the conformation-dependent mAb OKT3. The mere co-translation of CD3-γ or -δ was sufficient to keep CD3-ε primarily in an oxidized and monomeric state, indicating that CD3-γ and -δ guide the folding of CD3-ε at the initial stage of TCR assembly.

Although the disulfide bonding status for individual cysteine residues in CD3-ε has not been established, we consider it likely that the Ig fold may facilitate formation of the disulfides between the cysteines at position 27 and position 76, which stabilize this structure. The extent to which the remaining extracellular cysteine residues (at position 97 and 100) are involved in disulfide bonding in the mature TCR–CD3 complex is not known. Our data suggest that when no DTT is added, some CD3-ε may not acquire a proper disulfide bonding status. We therefore cannot indicate which cysteine residues are involved in the formation of these ε oligomers.

The presence of disulfide-linked ε homodimers in a completely assembled and functional TCR complex has been reported for several human and murine T cell lines (45, 46). However, the presence of these ε homodimers was established either using conformation-independent anti–CD3-ε reagents (46) or by examining immunoprecipitates obtained with the conformation-specific antibody OKT3 after cell surface labeling under oxidative conditions (45). Although we observe oligomerizaton of CD3-ε as a major pathway in the course of its biosynthesis in the absence of CD3-γ and -δ, we could not detect these structures as part of a complete TCR–CD3 complex by immunoprecipitation with the mAb OKT3. This antibody, the epitope of which has not been mapped accurately, only recovered CD3 subunits in immunoprecipitations when these were translated simultaneously with TCR-α and -β under redox conditions that allow proper folding (data not shown). Intracellular immunostaining of CD3-ε transiently expressed in COS cells could only be achieved with OKT3 when either CD3-γ or CD3-δ were coexpressed (52).

Furthermore, using OKT3 we could only co-immunoprecipitate TCR-α and -β as disulfide-linked heterodimers, despite the presence of free TCR-α, -β, and CD3 pairs in the lysate. Hence, OKT3 most likely recognizes a conformation-dependent epitope on CD3-ε that becomes available when CD3-γ or -δ is present, and remains stable after solubilization in detergent when the hexameric εγβ–αδε structure has been formed. A conformational change of murine CD3-ε, dependent on the presence of murine CD3-γ or -δ, has been reported (53). Taken together, our data strongly support the notion that CD3-ε requires the presence of CD3-γ or -δ to fold correctly and to allow the assembly of the TCR–CD3 complex to proceed.

CD3-ω, a nonglycosylated polypeptide of 28 kD, is expressed in human T cells and transiently interacts with pairs of CD3-δ/ε at an early stage of TCR assembly (54). Upon maturation and egress of the nascent TCR from the ER, it is released from the complex (54). CD3-ω is commonly regarded as a T cell–specific molecular chaperone and hence, is likely to be absent from the canine microsomes used in this study. The function of CD3-ω is unknown since the expression of CD3-ω is dispensable for the assembly and cell surface expression of the TCR–CD3 complex.

We have also investigated the role of the molecular chaperone calnexin in the folding of CD3-ε. Calnexin binds in a prolonged fashion to misfolded proteins or single subunits that fail to join a complex, suggesting that it may participate in the ER quality control machinery that prevents egress of incompletely folded or unassembled polypeptides from the ER (24, 55). A characteristic of calnexin is its preference for monoglucosylated trimming intermediates of N-linked glycoproteins (21). Calnexin can bind exclusively in a lectin-like fashion, i.e., independent of the folding state of the substrate (20), an observation that ties the cycle of deglucosylation and reglucosylation to glycoprotein folding (56). Although the interaction between heavy chains of MHC class I molecules and calnexin is only initiated by partially trimmed glycans, removal of glycans of class I–calnexin complexes by treatment with EndoH does not lead to the dissociation of this complex (21). Calnexin-dependent ER retention has been observed for the nonglycosylated CD3-ε subunit transfected into COS cells, based on colocalization of CD3-ε with mutant forms of calnexin, engineered with respect to their retention in the ER (24). The mode of binding between calnexin and CD3-ε therefore appears to deviate from that between calnexin and other molecules.

When we translated CD3-ε in the absence of CD3-γ and -δ, we could only detect oligomeric CD3-ε subunits in a complex with calnexin. Theoretically, this interaction could involve an unlabeled ER-resident protein that resides in the microsomes used in the translation experiments and would escape detection. Calnexin can interact in a glycan-independent manner with aggregates of the glycosylated VSV–G protein (22). Possibly some structural attribute of such VSV–G protein aggregates might be mimicked by the CD3-ε oligomer. It will be important to define more accurately and in molecular terms, how calnexin can engage in glycan-independent interactions, and whether the outcome of these interactions could, under certain circumstances, lead to the resolution of the aggregates.

In conclusion, we have demonstrated that the oligomerization of CD3-ε is a pathway essentially not used when CD3-γ and/or CD3-δ subunits are available. In this sense, the CD3-ε could serve as a subunit that monitors proper oligomerization of the CD3 complex.

We would like to thank Dr. Armin Rehm for continuous discussions and all present lab members for helpful suggestions. We thank Dr. O. Acuto, Dr. P. Anderson, Dr. J. Coligan, Dr. M. Owen, Dr. C. Terhorst, and Dr. J. Trowsdale for providing essential reagents.