Expression and characterization of recombinant soluble human CD3 ...

Download PDF
1 downloads 0 Views 279KB Size Report
Comment
2Pacific Northwest Research Institute, 720 Broadway, Seattle, WA 98122, USA. 3Present address: Seattle Genetics, Inc., Bothell, WA 98021, USA. 4Present ...
International Immunlogy, Vol. 14, No. 4, pp. 389±400

ã 2002 The Japanese Society for Immunology

Expression and characterization of recombinant soluble human CD3 molecules: presentation of antigenic epitopes de®ned on the native TCR±CD3 complex Che-Leung Law1,3, Martha Hayden-Ledbetter1,2, Sonya Buckwalter1,2, Lisa McNeill1, Hieu Nguyen1, Phil Habecker1, Barbara A. Thorne1,4, Raj Dua1 and Jeffrey A. Ledbetter1,2 1Xcyte

Therapies, Inc., Seattle, WA 98104, USA Northwest Research Institute, 720 Broadway, Seattle, WA 98122, USA

2Paci®c

3Present 4Present

address: Seattle Genetics, Inc., Bothell, WA 98021, USA address: Targeted Genetics Corp., Seattle, WA 98101, USA

Keywords: author, to, supply, keywords Abstract The TCR±CD3 complex consists of the clonotypic disul®de-linked TCRab or TCRdg heterodimers, and the invariant CD3d, e, g and z chains. We generated plasmid constructs expressing the extracellular domains of the CD3d, e or g subunits fused to human IgG1 Fc. Recombinant fusion proteins consisting of individual CD3d, e or g subunits reacted poorly with anti-CD3 mAb including G19-4, BC3, OKT3 and 64.1. Co-expression of the CD3e±Ig with either the CD3d±Ig (CD3ed±Ig) or the CD3g±Ig (CD3eg±Ig) resulted in fusion proteins with much increased binding to G19-4. A brief acid treatment of the puri®ed CD3ed±Ig fusion protein substantially improved its binding to BC3, OKT3 and 64.1. Surface plasmon resonance analysis revealed that the dissociation constants for CD3ed± Ig and anti-CD3 mAb ranged from 10±8 to 10±9 M. Based on these results, a single-chain (sc) construct encoding the CD3d chain linked to the CD3e chain with a ¯exible linker followed by human IgG1 Fc was expressed. The sc CD3de±scIg reacted with anti-CD3 mAb without requiring acid treatment. Moreover, anti-CD3 mAb bound CD3ed±Ig at a higher af®nity than CD3eg±Ig, suggesting potential structural differences between the CD3ed and CD3eg subunits. In summary, we report the expression of soluble recombinant CD3 proteins that demonstrate structural characteristics of the native CD3 complex expressed on the T cell surface. These CD3 fusion proteins can be used to further analyze the structure of the TCR±CD3 complex, and to identify molecules that can interfere with TCR±CD3-mediated signal transduction by disrupting the interaction between CD3 and TCR subunits. Introduction The TCR±CD3 complex consists of six subunits including the clonotypic disul®de-linked TCRab or TCRdg heterodimers and the invariant CD3 complex. The invariant CD3 complex is made up of four relatively small polypeptides: CD3d (25 kDa), CD3e (20 kDa), CD3g (20 kDa) and CD3z (16 kDa). CD3d, e and g chains possess a signi®cant degree of similarity to each other in their amino acid sequences. They are members of the Ig supergene family and each of them possesses a single extracellular (EC) Ig-like domain. In contrast, CD3z only has a

9-amino-acid EC domain and a longer cytoplasmic domain when compared to CD3 d, e and g. The cytoplasmic domains of the CD3 chains contain one to three copies of a conserved motif termed an immunoreceptor tyrosine-based activation motif (ITAM) that can mediate cellular activation. One consequence of TCR±CD3 complex ligation by peptide±MHC ligands is the recruitment of a variety of signaling factors to the ITAM of the CD3 chains. This initiates the activation of multiple signal transduction pathways, eventually resulting in

Correspondence to: J. A. Ledbetter, Paci®c Northwest Research Institute, 720 Broadway, Seattle, WA 98122, USA; E-mail: [email protected] Transmitting editor: E. A. Clark

Received 30 October 2001, accepted 8 January 2001

390 Soluble recombinant human CD3 gene expression, cellular proliferation and generation of effector T cell functions (1±4). The assembly and surface expression of the TCR complex are tightly regulated processes. It is well established that surface expression of a TCR complex requires the presence of TCRab or TCRgd plus CD3d, CD3e, CD3g and CD3z chains (5,6). Absence of any one chain renders the complex trapped in the endoplasmic reticulum (ER) and subjects them to rapid proteolytic degradation (7±11). The precise stoichiometry of a TCR±CD3 complex awaits complete elucidation. Initially, one copy of the clonotypic TCRab chains was proposed to be in complex with a CD3ed heterodimer, a CD3eg heterodimer and a CD3zz homodimer (12). More recent ®ndings suggest a bigger complex consisting of two copies of the clonotypic TCRab heterodimer in complex with the above six CD3 chains to form a decameric complex (13±15). In this complex, the TCR heterodimers and CD3z homodimers are covalently linked by disul®de bonds, while the CD3ed and CD3eg heterodimers are not covalently linked. Furthermore, the interaction among CD3ed, CD3eg, CD3zz and TCRab or TCRgd chains has been shown to be non-covalent. Assembly of the TCR±CD3 complex begins with pairwise interactions between individual TCRa or TCRb chains with the CD3 chains in the ER. Intermediates consisting of a single TCR chain in association with the CD3 chains are detectable in the ER (16,17). Transfection studies conducted in non-lymphoid cells show that TCRa can associate with CD3d and CD3e, but not CD3z, whereas TCRb can associate with CD3d, e and g, but not CD3z (9,16). Incorporation of the CD3z chain appears to be the rate-limiting step for the formation of a mature TCR±CD3 complex. TCRab, and CD3d, e and g chains must be present in the ER before CD3z can assemble with the partial TCR±CD3 complex to form the ®nal product for surface expression (18). Association between the TCR and CD3 chains depends largely on the charged amino acid residues in their transmembrane (TM) domains. Positively charged amino acid residues are present in the TM domains of the TCRab chains. Negatively charged amino acids are found in the TM domains of the CD3 chains. Formation of salt bridges due to these charged amino acid residues may be the main force driving the association between the TCRab and CD3 chains (19,20). Even though the assembly and signaling properties of the TCR±CD3 complex have been extensively studied, the functions of the EC domains of the CD3d, e or g chains remain to be fully understood. The crystal structure of a TCR±anti-TCR complex provides evidence for the presence of a binding pocket in the TCRb chain large enough to accommodate the EC domain of CD3e (21,22). Deletional analysis has also revealed an extracellular (EC) region proximal to the TM domains of CD3d, e or g with a conserved CXXCXE motif that may mediate hetero-dimerization CD3 chains (23). In order to characterize further the structures and functions of the EC domains of the CD3 subunits, we sought to generate recombinant fusion protein constructs encoding the EC domains of human CD3d, e or g fused with the human IgG1 Fc fragment. We report the expression of soluble human CD3± Ig fusion proteins that contain either one or two of the CD3 chains, and demonstrate expression and processing conditions that allow these fusion proteins to exhibit antigenic epitopes that are de®ned on the native TCR±CD3 complex.

Methods mAb and cells OKT3 was purchased from ATCC (Rockville, MD). BC3 was obtained from Dr Claudio Anasetti (Fred Hutchinson Cancer Research Center, Seattle, WA). 64.1 and 9.3 were obtained from Dr John Hansen (Fred Hutchinson Cancer Research Center). G19-4 has previously been described (24). Normal peripheral blood mononuclear cells (PBMC) were activated by plate-immobilized G19-4 (anti-CD3) and 9.3 (antiCD28) in RPMI 1640 supplemented with 10% FBS for 72 h before total RNA isolation. COS-7 cells were purchased from the ATCC and maintained in DMEM supplemented with 10% FBS. CD3±Ig fusion protein construction and expression For the generation of CD3±Ig fusion proteins, cDNAs encoding EC domains of CD3 were generated using total RNA from antiCD3/anti-CD28-activated peripheral blood mononuclear cells (PBMC) as the template and RT-PCR using the Superscript II RNase H± reverse transcriptase (Life Technologies, Gaithersburg, MD). The PCR primers used were as follows: CD3d EC domain 5¢ primer, 5¢-gcg aag ctt gcc acc atg gaa cat agc acg ttt ctc-3¢; 3¢ primer, 5¢-cgc gga tcc atc cag ctc cac aca gct ctg-3¢; CD3e EC domain 5¢ primer, 5¢-gcg ata aag ctt gcc acc atg cag tcg ggc act cac tgg-3¢; 3¢ primer, 5¢-gcg gga tcc atc cat ctc cat gca gtt ctc aca-3¢; CD3g EC domain 5¢ primer, 5¢-gcg ata aag ctt gcc acc atg gaa cag ggg aag ggc ctg-3¢; 3¢ primer, 5¢- gcg gga tcc att tag ttc aat gca gtt ctg aga c-3¢. Underlined nucleotides indicate HindIII and BamHI sites in the 5¢ and 3¢ primers respectively. All primer sets were designed to include the CXXCXE motifs present in the EC of the CD3d, e and g subunits. HindIII- and BamHI-cut PCR products were ligated in-frame 5¢ to a human IgG1 Fc fragment in a mammalian expression vector controlled by a CMV promoter similar to methods previously described (25). The CD3 de±single chain (sc) Ig fusion protein constructs were made by PCR ampli®cation of the individual chains as cDNA cassettes, restriction digestion at compatible sites and ligation together into an Ig expression vector recipient plasmid. Oligonucleotides used to amplify the CD3d and CD3e EC domains included a leader peptide attached to the 5¢ end of the CD3d domain cassette, with the ®rst two repeats of the (Gly4Ser)3 linker at the 3¢ end of the cassette, including a BamHI site to simplify construction of the complete fusion gene. Similarly, the second domain cassette was ampli®ed using a 5¢ oligonucleotide beginning with a BamHI site and adding the third (Gly4Ser) subunit to the linker, followed by the codons for the ®rst 10 amino acids of the mature peptide and a 3¢ oligonucleotide with a BglII restriction site attached to the 3¢ end. The fragments were then assembled by three-way ligation of the HindIII±BamHI fragment, the BamHI±BglII fragment and HindIII±BglII digested vector already containing the BglII±XbaI human IgG1 fragment. A mammalian expression vector controlled by a CMV promoter was used to harbor the fusion protein expression cassette for expression in COS-7 cells. The PCR primers for construction of the CD3d-linkerCD3e±Ig are as follows: CD3d103, 5¢-ggc aag ctt atg gaa cat agc acg ttt ctc tct ggc ctg-3¢; CD3d104, 5¢-tcc gga tcc gcc acc ccc aga ccc tcc gcc acc atc cag ctc cac aca gct ctg-3¢;

Soluble recombinant human CD3 391 CD3e103, 5¢-ggc gga tcc gga ggt ggt ggc tca gat ggt aat gaa gaa atg ggt ggt att aca-3¢; CD3e104, 5¢-ggc aga tct atc cat ctc cat gca gtt ctc aca cac tct-3¢. Ligation products were transformed into DH5a bacteria and isolated colonies picked to screen for transformants containing the correct insertions. Constructs with the correct digestion pattern were then sequenced using the Big Dye terminator cycle sequencing kit reagents (PE Biosystems, Foster City, CA) on an ABI Prism 310 sequencer (PE Biosystems). The plasmids for positive clones with the correct sequence were then transfected into COS-7 cells. Transient expression of CD3±Ig fusion protein constructs in COS cells was performed as reported (25). To purify Ig fusion proteins, COS cell supernatants were passed through Protein A±Sepharose (Pharmacia Biotech, Piscataway, NJ), and washed successively with 0.14 M potassium phosphate buffer (pH 8.0) and 0.1 M sodium acetate buffer (pH 4.5). Bound fusion proteins were then eluted with 0.1 M citric acid (pH 2) and immediately neutralized with Tris base. Fractions containing Ig fusion proteins were pooled and dialyzed against PBS before further analysis. Acid treatment of CD3±Ig fusion proteins Acid treatment of CD3±Ig was accomplished by incubating 9 parts of fusion protein in PBS with 1 part of 1 M citric acid at pH 2 or 1 M citrate buffers at pH 3, 4, 5 and 6. The resulting citric acid or citrate concentration was 0.1 M. Acidi®ed fusion protein preparations were incubated at 4°C for 60 min. Fusion protein solutions were then neutralized with 1 M Tris base and dialyzed overnight at 4°C against PBS. ELISA Aliquots of 5 mg/ml of the capture anti-CD3 mAb in 0.054 M carbonate/bicarbonate buffer (pH 9.6) were adsorbed onto wells of ELISA plates overnight at 37°C. Wells were then blocked with 5% milk diluent/blocking solution concentrate (Kirkegaard & Perry, Gaithersburg, MD) in PBS at 37°C for 2 h. Before addition of the fusion proteins, wells were rinsed 3 times with wash buffer (0.5% Tween 20 in PBS). Fusion proteins diluted serially into 5% milk diluent/blocking solution concentrate in PBS and added to wells. Human IgG1 or irrelevant, similarly constructed Ig fusion protein was used as negative a control. After incubation at 37°C for 1 h, wells were rinsed 3 times with wash buffer. Horseradish peroxidaseconjugated donkey anti-human IgG (Jackson ImmunoResearch, West Grove, PA) diluted at 1:5000 in 5% milk diluent/blocking solution concentrate in PBS was added to wells and incubated for 1 h at 37°C to detect bound fusion proteins. After 3 rinses with wash buffer, 1 component TMB microwell peroxidase substrate (Kirkegaard & Perry) was added. Color was developed in the dark for 5±15 min, stopped by the addition of an equal volume of 0.1 M HCl, followed by reading on an ELISA reader using a 450 nm ®lter. Anti-CD3 immunoblotting CD3±Ig fusion proteins were resolved on SDS±PAGE under non-reducing conditions and they were then transferred onto either nylon or PVDF membranes. After blocking, membranes were immunoblotted with either horseradish peroxidase-conjugated goat anti-human IgG (Jackson ImmunoResearch) or

anti-CD3 mAb. Binding of anti-CD3 mAb was detected by horseradish peroxidase-conjugated goat anti-mouse IgG Fc (Jackson ImmunoResearch). Blots were then developed using the enhanced chemiluminescence (ECL) kit from Amersham Pharmacia (Piscataway, NJ). Binding af®nity between CD3ed±Ig and anti-CD3 mAb Measurements of af®nity between CD3ed±Ig and anti-CD3 mAb were carried out with a BIAcore 1000 (Biacore, Piscataway, NJ) based on surface plasmon resonance (SPR) technology. Individual anti-CD3 mAb were immobilized onto dextran-based CM5 sensor chips (Biacore) in the presence of N-hydroxysuccinimide and N-ethyl-N¢-[3-di-methyl-amino)propyl] carbodiimide hydrochloride (Biacore). Graded concentrations of CD3ed±Ig, ranging from 20 to 1.25 mg/ml, were injected over the mAb-coated CM5 sensor chip surface to initiate protein association. After maximal binding, dissociation between the bound CD3ed±Ig and anti-CD3 mAb was accomplished by the injection of the carrier buffer free of any Ig fusion protein. The association and dissociation kinetics between CD3ed±Ig and anti-CD3 mAb, and hence the on and off rates respectively, were detected as changes in SPR signals and expressed in sensograms as resonance units versus time. The on and off rates generated from graded doses of CD3ed±Ig interacting with the same anti-CD3 mAbcoated surface were then used to compile the dissociation constants, i.e. af®nity, of interaction. Results Construction and expression of soluble recombinant CD3±Ig fusion proteins In order to characterize the CD3 complex in solution, cDNA encoding the EC domains of CD3d, e and g chains were individually fused at the N-terminus to a human IgG1 hinge± CH2±CH3 Fc fragment and cloned into a mammalian expression vector (Fig. 1A). Two cysteine residues, responsible for the formation of interchain disul®de bonds between the heavy chains of native human IgG1, are present in the hinge region of human IgG1 (Fig. 1A). Accordingly, we reasoned that expression of any individual CD3±Ig fusion plasmid construct in cells would result in the production of one species of disul®de-linked homodimer, hereinafter referred as CD3±Ig homodimers. On the other hand, the expression of two different CD3±Ig fusion plasmid constructs would result in the production of a mixture of two species of disul®de-linked homodimers and one species of disul®de-linked heterodimer (Fig. 1A), hereinafter referred as CD3±Ig heterodimers. Expression of these plasmids in COS cells, either individually or in combinations of two plasmids, resulted in the synthesis of recombinant proteins and their secretion into the culture medium (Fig. 1B). Soluble CD3±Ig proteins could be accumulated to 5±10 mg/l of spent COS cell medium. The sizes of different recombinant proteins, estimated by SDS± PAGE under reducing conditions, were all ~45±50 kDa (Fig. 1B). This range is consistent with the sizes predicted from the amino acid sequences of CD3 EC domains and human IgG Fc. On the other hand, multiple protein bands were detected when these fusion proteins were resolved on SDS±PAGE

392 Soluble recombinant human CD3 protein might contain one more CD3±Ig chain disul®de-linked to the complex than the previous one. Binding of CD3±Ig fusion proteins to anti-CD3 mAb

Fig. 1. CD3±Ig fusion protein. (A) This schematic depicts cDNA constructs derived from the fusion of various CDe EC domains to the human IgG1Fc (hinge±CH2±CH3) fragment and their expression to generate homo- and heterodimers of CD3±Ig. (B) COS cells were transiently transfected by single or two CD3±Ig fusion protein constructs. Spent supernatants were incubated with Protein A± Sepharose. Bound fusion proteins were eluted by either reducing or non-reducing sample buffer as indicated and resolved on SDS± PAGE. The gels were then stained by Coomasie brilliant blue.

under non-reducing conditions (Fig. 2B). The lowest molecular weight bands were ~90±100 kDa, corresponding to the predicted size of a homo- or heterodimer with two CD3±Ig chains. Each higher mol. wt protein appeared to be ~45±50 kDa larger in size successively. Hence, each higher molecular

Some anti-CD3 mAb have been reported to bind individual CD3 chains, whereas others recognize only CD3e in complex with either CD3d or CD3g (26). We applied four anti-CD3 mAb (G19-4, OKT3, BC3 and 64.1) to probe the structures of the CD3±Ig homo- and heterodimers. Plate-immobilized anti-CD3 mAb were used to capture CD3±Ig fusion proteins, which was then detected by antibodies directed speci®cally against the Fc fragment of human IgG. Figure 2 shows that supernatants from COS cells transfected with individual CD3±Ig constructs did not react with anti-CD3 mAb, with the exception of G19-4, which consistently showed low levels of binding to CD3ee±Igcontaining supernatants. Co-transfection of COS cells with CD3e±Ig and CD3d±Ig plasmids or CD3e±Ig and CD3g±Ig plasmids generated supernatants that bound much better to G19-4. OKT3, BC3 and 64.1 did not show any consistent binding to CD3ed±Ig or CD3eg±Ig supernatants. Mixing of CD3e±Ig supernatants with CD3d±Ig or CD3g±Ig supernatants did not increase the binding to G19-4, suggesting that CD3e± Ig and CD3d±Ig fusion proteins formed G19-4-binding complexes in COS cells before being secreted into the culture medium. We then examined the binding of Protein A±Sepharosepuri®ed CD3±Ig preparations to the same panel of anti-CD3 mAb (Fig. 3). All anti-CD3 mAb tested showed improved reactivity and saturable binding to puri®ed CD3±Ig fusion proteins (Fig. 3A and B, upper panel). The concentration of CD3ed±Ig that half-saturated G19-4 was reproducibly 5- to 10fold less than that required to saturate OKT3, BC3 and 64.1 (Fig. 3B, upper panel). The CD3eg±Ig heterodimer interacted with the panel of anti-CD3 mAb at a much wider range of af®nity. Saturable binding was observed between CD3eg±Ig and G19-4 and CD3eg±Ig and BC3, but the concentration of CD3eg±Ig to half-saturate G19-4 was ~20-fold less than that needed to half-saturate BC3 (Fig. 3A and B, lower panel). In contrast to G19-4 and BC3, CD3eg±Ig showed minimal binding to both OKT3 and 64.1. In addition, the CD3ed±Ig and CD3eg± Ig heterodimers bound much better to anti-CD3 mAb than the CD3dd±Ig, CD3ee±Ig and CD3gg±Ig homodimers. Although detectable CD3ee±Ig/G19-4, CD3gg±Ig/G19-4, CD3gg±Ig/BC3 CD3ee±Ig/64.1 and CD3gg±Ig/64.1 binding was observed, such binding was not saturable even at fusion protein concentrations of 100 mg/ml (Fig. 3A). Finally, none of the four mAb used in this study recognized the CD3dd±Ig homodimer (Fig. 3A). Acid-induced conformational changes of CD3ed±Ig The ability of OKT3, BC3 and 64.1 to bind puri®ed CD3ed±Ig (Fig. 3) was in stark contrast to the lack of binding of the same mAb to unpuri®ed CD3ed±Ig fusion proteins present in the COS cell supernatants shown in Fig. 2. This prompted us to examine the possibility that changes in the conformation of CD3ed±Ig occurred during puri®cation, which generated the epitope(s) recognized by OKT3, BC3 and 64.1. After binding of CD3±Ig fusion proteins to Protein A±Sepharose, we washed the matrix with pH 8 and 4.5 buffers successively, and eluted the bound fusion protein at pH 2. Eluted fusion proteins were

Soluble recombinant human CD3 393 Table 1. Af®nity between CD3ed±Ig and anti-CD3 mAb determined by surface plasmon resonance

G19-4 OKT3 BC3 64.1 Fig. 2. Co-expression of CD3e±Ig with CD3d±Ig or CD3g±Ig in COS cells produced heterodimeric CD3±Ig fusion proteins that bound anti-CD3 mAb. COS cells were transiently transfected with constructs encoding CD3d±Ig (d), CD3e±Ig (e), CD3g±Ig (g), CD3e±Ig + CD3d±Ig (ed) or CD3e±Ig + CD3g±Ig (eg). COS cell supernatants were tested for binding of CD3±Ig fusion proteins to G19-4, OKT3, BC3 and 64.1 by ELISA. Supernatants containing CD3e±Ig were either mixed with those containing CD3d±Ig (e + d) or CD3g±Ig (e + g) and tested for binding to the same panel of anti-CD3 mAb.

then neutralized by Tris buffer. When CD3ed±Ig heterodimers were neutralized by Tris buffer immediately after elution, they showed extremely low binding activity to OKT3, BC3 and 64.1 (Fig. 4A and data not shown). However, the same fusion protein still bound to G19-4 in a saturable manner (Fig. 4A, untreated line). A brief, as short as 15 min, additional exposure of the CD3ed±Ig to pH 2 enabled it to bind OKT3 and BC3. Interestingly, this acid treatment did not change the binding activity of CD3ed±Ig to G19-4. This suggests that the G19-4 epitope on CD3ed±Ig may be different from those of the OKT3 and BC3 epitope(s). When the effect of different pH was examined, we found that pH 2 was most effective (Fig. 4B). Frozen preparations of acid-treated CD3ed±Ig retained its ability to bind OKT3, BC3 and 64.1 over a long period of time. Non-reducing SDS±PAGE revealed the presence of multiple species of disul®de-linked fusion proteins in all of the puri®ed CD3±Ig preparations (Fig. 1B). We examined which of the species could bind anti-CD3 mAb. Western blot analysis on CD3±Ig fusion proteins resolved under non-reducing conditions revealed that G19-4 could recognize CD3ee±Ig, CD3gg± Ig, CD3ed±Ig and CD3eg±Ig, but not CD3dd±Ig (Fig. 5A). This is consistent with the failure of CD3dd±Ig to bind G19-4 on ELISA shown in Fig. 3. G19-4 preferentially bound the 90 kDa dimeric form of the CD3gg±Ig, CD3ed±Ig and CD3eg±Ig, with the exception of CD3ee±Ig to which G19-4 appeared to bind better to the 150 kDa trimer. On the other hand, when OKT3 was used to probe similar blots, it speci®cally bound only to the 90 kDa heterodimer band in the CD3ed±Ig lane and it did not recognize any other forms of CD3±Ig fusion proteins (Fig. 4B). As OKT3 did not bind any of the three homodimers, it is most likely that the 90 kDa protein recognized by OKT3 on Western blot was the heterodimeric form consisting of one chain of CD3e±Ig and one chain of CD3d±Ig. Hence the fusion protein bound by OKT3 on ELISA shown in Fig. 2±4 might also be predominantly the same heterodimer.

kd ( s±1)a ka (M±1 s±1)b Kd (M)c kd ( s±1) ka (M±1 s±1) Kd (M) kd ( s±1) ka (M±1 s±1) Kd (M) kd ( s±1) ka (M±1 s±1) Kd (M)

Experiment 1

Experiment 2

7.8 3 10±4 6.1 3 105 1.3 3 10±9 172 3 10±4 16.2 3 105 10.6 3 10±9 108 3 10±4 17.1 3 105 6.3 3 10±9 97.4 3 10±4 12.9 3 105 7.6 3 10±9

15.6 3 10±4 5.5 3 105 2.8 3 10±9 181 3 10±4 14.6 3 105 12.4 3 10±9 124 3 10±4 15.3 3 105 8.1 3 10±9 91.4 3 10±4 3.2 3 105 28.6 3 10±9

aRate

constant for dissociation, off rate. constant for association, on rate. cDissociation constant (af®nity) which equals to k /k . d a bRate

High-af®nity interaction between CD3ed±Ig and anti-CD3 mAb The af®nity of interaction between CD3ed±Ig and anti-CD3 was determined by SPR in which graded concentrations of CD3ed± Ig in liquid phase were allowed to bind anti-CD3 mAb immobilized on chips. Interactions between CD3ed±Ig and anti-CD3 showed on rates ranging from 3.2 3 105 to 17.1 3 105 M±1 s±1 and off rates ranging from 7.8 3 10±4 to 181 3 10±4 s±1, indicative of high-af®nity interaction (Table I). Whereas the on rates of CD3ed±Ig bound onto OKT3, BC3 and 64.1 were similar to each other, the off rate of CD3ed±Ig bound onto G19-4 was 6- to 20-fold lower. Thus, in two separate experiments G19-4 was found to bind CD3ed±Ig at the highest af®nity. The average dissociation constant between CD3ed±Ig and G19-4 was 3.5- to 9-fold lower than those between CD3ed±Ig and the other three anti-CD3 mAb. These results are consistent with the ELISA data shown in Fig. 2. Construction and expression of CD3de±scIg fusion protein Co-expression of plasmid constructs encoding CD3e±Ig and CD3d±Ig in COS cells clearly generated disul®de-bonded heterodimeric proteins bearing apparently native epitopes for anti-CD3 mAb. However, the molecular composition of CD3ed±Ig was likely to be a complex mixture of homo- and heterodimers and oligomers (Figs 1 and 5). Acid treatment of the puri®ed CD3ed±Ig protein was also required to generate the BC3, OKT3 and 64.1 epitope(s) (Fig. 4). We sought to resolve some of these problems by re-engineering the CD3±Ig fusion protein expression constructs. Two major changes were incorporated as depicted in Fig. 6. First, a cDNA fragment encoding the leader plus the EC domain of CD3d was placed N-terminal to a cDNA fragment encoding the EC domain of CD3e. A Gly/Ser linker was used to separate the two fragments, which may also add ¯exibility to the expressed protein and enhance association between the CD3e and CD3d subunits. Second, we substituted serine residues for the two cysteine residues in the hinge region of the human IgG1 Fc fragment. We reasoned that since the CD3d and CD3e subunits were present on the same polypeptide they should be able to associate with each other to form the epitope(s)

394 Soluble recombinant human CD3

Fig. 3. Binding of puri®ed CD3±Ig fusion proteins to anti-CD3 mAb. (A) Various CD3±Ig homodimeric or heterodimeric fusion proteins were puri®ed from spent medium collected from COS cells transiently expressing different CD3±Ig constructs, singly or in combinations of two. The binding of puri®ed fusion proteins to antiCD3 mAb was determined by ELISA. Anti-CD3 mAb was used to capture CD3±Ig fusion proteins as described in Methods. (B) CD3ed±Ig and CD3eg±Ig binding curves from (A) were re-plotted to allow direct comparison of their binding activities to anti-CD3 mAb.

recognized by anti-CD3 mAb. In the absence of the cysteine residues in the Fc fragment, monomeric forms of the CD3de± scIg fusion protein might be produced. The re-engineered CD3de±scIg plasmid construct was transiently expressed and puri®ed fusion proteins were analyzed by Western blotting. An anti-human IgG reagent detected three major species of proteins of ~60, 120 and >200 kDa, the predominant forms being the 120 kDa species (Fig. 7A). The 60 kDa protein corresponded to the predicted size of a monomeric form of CD3de±scIg. Accordingly, the 120

and >200 kDa proteins might be disul®de-linked dimeric and tetrameric forms of CD3de±scIg respectively. Four unpaired cysteine residues were present in the EC domains of CD3d and CD3e in this construct (Fig. 6); they were probably involved in the disul®de linking of CD3de±scIg monomers to give dimers and tetramers. When G19-4 was used to probe Western blotted CD3de±scIg, it bound predominantly to the 60 kDa monomers and reacted very poorly with the 120 kDa dimer (Fig. 7B). On the other hand, OKT3 bound exclusively to the 60 kDa protein with virtually no binding to the 120 kDa dimer.

Soluble recombinant human CD3 395

Fig. 4. Acid-induced conformational changes in CD3ed±Ig. (A) Puri®ed CD3ed±Ig was treated with pH 2 citric acid buffer at 4°C for the indicated lengths of time and its binding activity to G19-4, OKT3 and BC3 was determined by ELISA. The quantity of Ig fusion protein in each batch of acid-treated CD3ed±Ig was monitored by an antihuman IgG ELISA. (B) Puri®ed CD3ed±Ig was treated by citric acid/ citrate buffers at different pH values as indicated for 45 min at 4°C, and its binding activity to OKT3 and BC3 was determined by ELISA.

These results suggest that when the CD3d and CD3e subunits were expressed as a single translational unit separated by a ¯exible linker they were able to associate with each other to regenerate epitopes recognizable by anti-CD3 mAb. It is also noteworthy that, unlike the CD3ed±Ig used in Fig. 5, the CD3de± scIg fusion protein was directly eluted from Protein A± Sepharose using SDS sample buffer without any acid treatment before SDS±PAGE and Western blotting. Binding of CD3de±scIg fusion protein to anti-CD3 mAb In order to con®rm that an OKT3 epitope was already present on the CD3de±scIg fusion protein as secreted from

COS cells, we compared the binding of puri®ed CD3de± scIg and puri®ed acid-treated CD3ed±Ig to G19-4 and OKT3 by ELISA. Puri®ed CD3de±scIg was eluted from Protein A±Sepharose and immediately neutralized by Tris buffer and dialyzed against PBS. No additional acid treatment was performed. Figure 8 shows that both CD3ed±Ig and CD3de±scIg bound to G19-4. The concentration of CD3de±scIg needed to half-saturate G19-4 was ~3-fold higher than that needed by CD3ed±Ig. Interestingly, unlike untreated CD3ed±Ig (Fig. 4), CD3de±scIg showed saturable binding to OKT3. The concentration needed to half-saturate OKT3 was ~6-fold less than that needed by

396 Soluble recombinant human CD3 CD3ed±Ig. Supernatants from COS-7 or CHO cells expressing CD3de±scIg also bound OKT3 by ELISA without the need of any acid treatment (data not shown). Hence, acid treatment of CD3de±scIg was not required to generate the OKT3 epitope. Discussion

Fig. 5. Immunoblotting of CD3±Ig fusion proteins by G19-4 and OKT3. Puri®ed CD3±Ig fusion proteins were resolved by nonreducing SDS±PAGE, transferred onto membranes, and then probed with G19-4 (A) and OKT3 (B).

Fig. 6. CD3de±scIg fusion protein construct. The coding region for the leader and EC domain of CD3d was joined to the 5¢ end of the coding sequence for the EC domain of CD3e via a ¯exible linker. This fragment was in turn fused 5¢ to a cDNA encoding the human IgG1 Fc fragment. The two cysteine residues in the hinge region of the human IgG Fc fragment were replaced by serine residues by site-directed mutagenesis.

The role of the CD3 subunits in the TCR complex has been the subject of intense research. Various gene deletion experiments unequivocally demonstrated the importance of the CD3d (27,28), e (29), g (28,30) and z (31±33) in T cell differentiation and functions. Assembly and traf®cking of the TCR±CD3 complex (5±10,12,16±18), interactions between charged amino acid residues in the TM domains of the CD3 and TCR ab subunits (19,20) and signaling transduction functions of CD3 cytoplasmic domains (2±4) are the other wellcharacterized aspects of the CD3 complex. Recently, the structure of a murine CD3eg complex in solution has become available (34). The construction and expression of soluble human CD3 complexes will certainly facilitate elucidation of the human CD3 structures and functional interaction with the clonotypic subunits of the human TCR. In this study, we report the expression and secretion of soluble Ig fusion proteins consisting of the EC domains of human CD3. Previous studies both in T cells and non-lymphoid cells have shown that partial TCR complexes are invariably trapped in ER (17,26,35). TCR ab and CD3d chains are further directed to pre-Golgi proteolytic degradation pathways, presumably due to the exposure of peptide sequences in their TM domains signaling for proteolysis. We have successfully expressed the soluble CD3±Ig fusion proteins by simply fusing the CD3d, e and g EC domains to the N-terminus of human IgG1. We have also produced stable CHO cell transfectants secreting comparable levels of CD3ed±Ig (data not shown). Expression of murine CD3e and CD3eg sc proteins in a bacterial expression system has been reported. These proteins were expressed as inclusion bodies in bacteria. Detergent solubilization and refolding could successfully regenerate a soluble murine CD3eg protein that bound antiCD3 mAb (36).

Fig. 7. Immunoblotting of CD3de±scIg fusion proteins by G19-4 and OKT3. Puri®ed CD3de±scIg fusion proteins were resolved by nonreducing SDS±PAGE, transferred onto membranes, and then probed with anti-human IgG (A), G19-4 (B) and OKT3 (C).

Soluble recombinant human CD3 397

Fig. 8. Binding of puri®ed CD3de±scIg fusion protein to G19-4 and OKT3. CD3de±scIg fusion protein was puri®ed from spent medium collected from COS cells transiently expressing the CD3de±scIg construct. The binding of puri®ed CD3de±scIg to G19-4 and OKT3 was compared to that of CD3ed±Ig by ELISA.

One characteristic of the CD3±Ig and CD3de±scIg fusion proteins is their tendency to oligomerize via disul®de linkages (Figs 1, 5 and 7A). Although the presence of two cysteine residues in the hinge region of human IgG1 can in theory explain the oligomeric structures, we consider this unlikely. First, most recombinant Ig fusion proteins reported by various groups, including us, exist predominantly as dimers. Second, as shown in Figs 6 and 7, oligomeric structures were still evident even when both cysteine residues in the IgG hinge of the CD3de±scIg were removed. The unpaired, membraneproximal cysteine residues in the CXXCXE motifs of CD3 may be one reason for oligomerization of the CD3±Ig fusion proteins. Although most CD3ed and CD3ed heterodimers do not appear to be covalently linked in their native state in the TCR complex, disul®de-bonded CD3e homodimers have been reported on human and mouse T cells before (37,38). Disul®de-linked oligomeric CD3e, d and g have also been detected both in murine T cells (39), and in an experimental system designed for in vitro translation of the TCR±CD3 complex (40). The cysteine residues in CXXCXE motifs have been proposed to be responsible for disul®de bonding of CD3e and CD3g (40). A sc recombinant murine CD3eg heterodimer without any CXXCXE motifs expressed by bacteria bound to anti-mCD3 mAb (36). Experiments to substitute the corresponding cysteine residues in the CD3ed±scIg construct (Fig. 6) with serine residues should be able to test the possibility that expression of human CD3e and CD3d in a co-linear translational unit may permit proper pairing even in the absence of the CXXCXE motifs. Most anti-human CD3 mAb recognize the CD3e subunit (26,41,42). A subset of them requires the presence of either CD3d or CD3g before they can recognize CD3e, suggesting

that the conformation of CD3e is modi®ed when it is in association with CD3d or CD3g (26,40). Binding results obtained from this study largely agree with previously published studies. Thus, G19-4, OKT3, BC3 and 64.1 demonstrated saturable high-af®nity interaction with CD3ed±Ig (Figs 3 and 4) (Table I), while G19-4 and OKT3 also bound at high af®nity to CD3de±scIg (Fig. 8). When G19-4 and OKT3 were used to Western blot CD3ed±Ig (Fig. 5) and CD3de±scIg (Fig. 7), both mAb recognized almost exclusively the dimeric form of CD3ed±Ig and the monomeric form of CD3de±scIg. These results are consistent with a model in which Ig fusion protein molecules consisting of only one CD3e chain and one CD3d chain are most ef®cient in generating an epitope for anti-CD3 mAb binding. The ability of CD3ed±Ig and CD3de±scIg to be bound by anti-CD3 mAb suggests that both CD3±Ig fusion proteins are structurally similar to the native TCR±CD3 complex. An acid treatment of CD3ed±Ig was needed before it could demonstrate high-af®nity binding to OKT3 and BC3, whereas the interaction between CD3ed±Ig and G19-4 was independent of this acid treatment (Figs 2, 3 and 4). This may re¯ect that the epitope recognized by OKT3 and BC3 is distinct from that recognized by G19-4. Thus, in the presence of CD3d chain acid treatment might induce proper folding of the OKT3/BC3 epitope for antibody recognition. On the other hand, as long as CD3e is in complex with CD3d chain a high-af®nity epitope for G19-4 binding was already formed. The ability of G19-4 to recognize CD3e alone while OKT3 only bound CD3ed±Ig or CD3de±scIg (Figs 2 and 5) is consistent with the idea that there are separate epitopes on the CD3e chain for G19-4 and OKT3 or BC3. Molecular refolding induced by acid treatment also seemed to be a speci®c property of CD3ed±Ig; acid treatment of CD3eg±Ig, CD3dd±Ig, CD3ee±Ig and CD3gg±Ig did not signi®cantly alter their binding af®nities to any of the four antiCD3 mAb (data not shown). A number of molecular chaperones have been identi®ed or proposed to regulate the assembly of TCR and CD3 chains into native TCR±CD3 complexes in the ER. These include CD3w (6), calnexin (39,40,43,44) and calreticulin (45). Through their binding to the TCR and CD3 chains, these chaperones may facilitate CD3 folding and subunit association. The acid treatment might substitute some functions performed by the chaperones to generate a conformational epitope on CD3ed±Ig. The CD3de± scIg bound G19-4 and OKT3 at high af®nity without any acid treatment (Figs 7 and 8), suggesting that a co-linear CD3d and CD3e polypeptide unit may be more autonomous in forming the OKT3/BC3 epitope. The binding af®nity between various anti-CD3 mAb and CD3±Ig fusion proteins was assessed by ELISA and SPR (Figs 3 and 8) (Table 1). Among the four anti-CD3 mAb studied, G19-4 consistently gave the highest af®nity in binding to CD3ed±Ig, CD3eg±Ig and CD3de±scIg (Figs 3 and 8) (Table 1). Anti-CD3 mAb also showed detectable binding to CD3 homodimers at concentrations >10 mg/ml. The concentrations of CD3ed±Ig or CD3eg±Ig that half-saturated ant-CD3 mAb were at least 2 orders of magnitude lower than those of the CD3 homodimers (Fig. 3A). Hence, it is unlikely that the presence of CD3dd±Ig, CD3ee±Ig or CD3gg±Ig in the CD3±Ig heterodimer preparations could substantially modify the apparent af®nity between anti-CD3 mAb and CD3ed±Ig or

398 Soluble recombinant human CD3 CD3eg±Ig using ELISA or SPR. The af®nity between OKT3 and TCR±CD3 expressed on T cells was determined to be 8.3 3 10±10 M (46,47). The af®nity for OKT3 binding to CD3ed±Ig was found to be ~2 3 10±9 M in this study, ~ 14-fold less than 8.3 3 10±10 M (Table I). Several possibilities may account for this apparent lower af®nity of OKT3. First, the af®nity of OKT3 previously reported was determined by the binding of OKT3 to CD3 molecules expressed on the T cell surface (46,47). Hence, it was a measure of overall avidity. We have shown that OKT3 binds exclusively to the 90 kDa form of CD3ed±Ig containing one chain each of CD3e and CD3d (Fig. 5). Accordingly, the af®nity measured in this study most likely re¯ects a monovalent interaction between a single CD3ed dimer and one F(ab) arm of an OKT3 molecule. Second, we cannot exclude the possibility that CD3ed±Ig did not fully resemble its native form in the TCR±CD3 complexes expressed on T cells. Only G19-4 and BC3, but not OKT3 or 64.1, demonstrated saturable binding to CD3eg±Ig (Fig. 3). However, they bound at af®nities that were 1±2 orders lower than their binding to CD3ed±Ig (Fig. 3). Acid treatment did not improve the binding of CD3eg±Ig to the anti-CD3 mAb tested in this study (data not shown). These results suggest that structural and possibly functional differences may exist between the CD3ed and CD3eg dimers in the TCR complex. Structural determination of the CD3ed±Ig and CD3de±scIg should provide valuable information on this question, and on how CD3ed and CD3eg may function differently in TCR-mediated signal transduction. Members of the Ig supergene family are well known for their functions as adhesion molecules. Therefore it would not be surprising if there were ligands for the EC domains of CD3 of Ig-like domains. During TCR±CD3 synthesis in the ER, formation of CD3ed and CD3eg pairs takes place ®rst and involves the speci®c interaction between their EC domains (48). CD3d and CD3g compete for binding to the same site on CD3e (17,49). Besides interacting with CD3d and CD3g chains, the EC domain of CD3e may also interact with the TCRab chains via a pocket formed by the FG loop of the TCR Cb domain (21,22). This interaction is mirrored in the B cell receptor complex in which the EC domains of Iga/Igb interact with membrane IgM and IgD molecules (50,51). Mouse CD3e may also interact with CD4 via their EC domains during T cell activation (52). A possible role played by the CD3e chain in T cell±B cell interaction is also implicated by the binding involving the EC domains of CD3e and Igb (53). We investigated whether CD3±Ig fusion proteins can displace the native CD3 subunits from the TCR±CD3 complex by binding to the TCRab chain, thereby compromising signal transduction via the TCR. So far we have not detected any high-af®nity binding of CD3±Ig fusion proteins to intact normal PBMC and T cell lines (data not shown). Likewise, we also have not detected any signi®cant effects of CD3±Ig fusion proteins on the proliferation of normal PBMC induced by either superantigens or anti-CD2 + anti-CD28mAb (data not shown). A couple of explanations can be offered for the failure of the CD3±Ig fusion proteins to displace CD3 subunits from the TCR±CD3 complex. First, the overall binding avidity among components of the TCR±CD3 complex may be too high to be disrupted by soluble CD3±Ig fusion proteins. Second, the Fc tails of the CD3±Ig proteins may impose strong steric hindrance to

prevent them from interacting with components of the TCR± CD3 complexes expressed on T cells. Additional studies to examine the ability of CD3±Ig to bind targets in detergentsolubilized cell lysates or puri®ed TCRab/dg chains should provide more insight into this question. The availability of soluble recombinant human CD3ed fusion proteins resembling the native proteins in the TCR±CD3 complex provides us with the starting materials for threedimensional structural determination of the CD3ed subunits. This could be the ®rst step to elucidate how the EC domains of CD3 subunits interact with each other and the TCRab/dg chains. Such information would facilitate the screening and identi®cation of high potency molecules that can disrupt the native TCR±CD3 complexes expressed on T cells. Such agents could be small molecules or biologics like new antiCD3 mAb directed against speci®c regions of the CD3 chains. Not only can these molecules be applied to further our understanding in the signaling function of the TCR±CD3 complex during T cell development, they can also be tested as immunosuppressive agents in the clinic. Acknowledgements This study was supported by Xcyte Therapies, Inc. and NIH grant CA90143

Abbreviations EC ITAM ER PBMC sc SPR TM

extracellular immunoreceptor tyrosine-based activation motif endoplasmic reticulum peripheral blood mononuclear cell single chain surface plasmon resonance transmembrane

References 1 Clevers, H., Alarcon, B., Wileman, T. and Terhorst, C. 1988. The T cell receptor/CD3 complex: a dynamic protein ensemble. Annu. Rev. Immunol. 6:629. 2 Weiss, A. and Littman, D. R. 1994. Signal transduction by lymphocyte antigen receptors. Cell 76:263. 3 Wange, R. L. and Samelson, L. E. 1996. Complex complexes: signaling at the TCR. Immunity 5:197. 4 Cantrell, D. 1996. T cell antigen receptor signal transduction pathways. Annu. Rev. Immunol. 14:259. 5 Minami, Y., Weissman, A. M., Samelson, L. E. and Klausner, R. D. 1987. Building a multichain receptor: synthesis, degradation, and assembly of the T-cell antigen receptor. Proc. Natl Acad. Sci. USA 84:2688. 6 Alarcon, B., Berkhout, B., Breitmeyer, J. and Terhorst, C. 1988. Assembly of the human T cell receptor±CD3 complex takes place in the endoplasmic reticulum and involves intermediary complexes between the CD3-gamma.delta.epsilon core and single T cell receptor alpha or beta chains. J. Biol. Chem. 263:2953. 7 Chen, C., Bonifacino, J. S., Yuan, L. C. and Klausner, R. D. 1988. Selective degradation of T cell antigen receptor chains retained in a pre-Golgi compartment. J. Cell Biol. 107:2149. 8 Bonifacino, J. S., Cosson, P. and Klausner, R. D. 1990. Colocalized transmembrane determinants for ER degradation and subunit assembly explain the intracellular fate of TCR chains. Cell 63:503. 9 Wileman, T., Carson, G. R., Concino, M., Ahmed, A. and Terhorst, C. 1990. The gamma and epsilon subunits of the CD3 complex

Soluble recombinant human CD3 399 10

11

12

13 14

15

16 17

18

19 20 21

22

23

24

25

26

inhibit pre-Golgi degradation of newly synthesized T cell antigen receptors. J. Cell Biol. 110:973. Wileman, T., Pettey, C. and Terhorst, C. 1990. Recognition for degradation in the endoplasmic reticulum and lysosomes prevents the transport of single TCRb and CD3d subunits of the T-cell antigen receptor to the surface of cells. Int. Immunol. 2:743. Wileman, T., Kane, L. P., Young, J., Carson, G. R. and Terhorst, C. 1993. Associations between subunit ectodomains promote T cell antigen receptor assembly and protect against degradation in the ER. J. Cell Biol. 122:67. Manolios, N., Letourneur, F., Bonifacino, J. S. and Klausner, R. D. 1991. Pairwise, cooperative and inhibitory interactions describe the assembly and probable structure of the T-cell antigen receptor. EMBO J. 10:1643. Exley, M., Wileman, T., Mueller, B. and Terhorst, C. 1995. Evidence for multivalent structure of T-cell antigen receptor complex. Mol. Immunol. 32:829. San Jose, E., Sahuquillo, A. G., Bragado, R. and Alarcon, B. 1998. Assembly of the TCR±CD3 complex: CD3 epsilon/delta and CD3 epsilon/gamma dimers associate indistinctly with both TCR alpha and TCR beta chains. Evidence for a double TCR heterodimer model. Eur. J. Immunol. 28:12. Fernandez-Miguel, G., Alarcon, B., Iglesias, A., Bluethmann, H., Alvarez-Mon, M., Sanz, E. and de la Hera, A. 1999. Multivalent structure of an alphabetaT cell receptor. Proc. Natl Acad. Sci. USA 96:1547. Manolios, N., Kemp, O. and Li, Z. G. 1994. The T cell antigen receptor alpha and beta chains interact via distinct regions with CD3 chains. Eur. J. Immunol. 24:84. Alarcon, B., Ley, S. C., Sanchez-Madrid, F., Blumberg, R. S., Ju, S. T., Fresno, M. and Terhorst, C. 1991. The CD3-gamma and CD3-delta subunits of the T cell antigen receptor can be expressed within distinct functional TCR±CD3 complexes. EMBO J. 10:903. Weissman, A. M., Frank, S. J., Orloff, D. G., Mercep, M., Ashwell, J. D. and Klausner, R. D. 1989. Role of the zeta chain in the expression of the T cell antigen receptor: genetic reconstitution studies. EMBO J. 8:3651. Hall, C., Berkhout, B., Alarcon, B., Sancho, J., Wileman, T. and Terhorst, C. 1991. Requirements for cell surface expression of the human TCR±CD3 complex in non-T cells. Int. Immunol. 3:359. Cosson, P., Lankford, S. P., Bonifacino, J. S. and Klausner, R. D. 1991. Membrane protein association by potential intramembrane charge pairs. Nature 351:414. Wang, J., Lim, K., Smolyar, A., Teng, M., Liu, J., Tse, A. G., Hussey, R. E., Chishti, Y., Thomson, C. T., Sweet, R. M., Nathenson, S. G., Chang, H. C., Sacchettini, J. C. and Reinherz, E. L. 1998. Atomic structure of an alphabeta T cell receptor (TCR) heterodimer in complex with an anti-TCR fab fragment derived from a mitogenic antibody. EMBO J. 17:10. Ghendler, Y., Smolyar, A., Chang, H. C. and Reinherz, E. L. 1998. One of the CD3epsilon subunits within a T cell receptor complex lies in close proximity to the Cbeta FG loop. J. Exp. Med. 187:1529. Borroto, A., Mallabiabarrena, A., Albar, J. P., Martinez, A. C. and Alarcon, B. 1998. Characterization of the region involved in CD3 pairwise interactions within the T cell receptor complex. J. Biol. Chem. 273:12807. Ledbetter, J. A., June, C. H., Martin, P. J., Spooner, C. E., Hansen, J. A. and Meier, K. E. 1986. Valency of CD3 binding and internalization of the CD3 cell-surface complex control T cell responses to second signals: distinction between effects on protein kinase C, cytoplasmic free calcium, and proliferation. J. Immunol. 136:3945. Law, C. L., Aruffo, A., Chandran, K. A., Doty, R. T. and Clark, E. A. 1995. Ig domains 1 and 2 of murine CD22 constitute the ligandbinding domain and bind multiple sialylated ligands expressed on B and T cells. J. Immunol. 155:3368. Salmeron, A., Sanchez-Madrid, F., Ursa, M. A., Fresno, M. and Alarcon, B. 1991. A conformational epitope expressed upon association of CD3-epsilon with either CD3-delta or CD3-gamma is the main target for recognition by anti-CD3 monoclonal antibodies. J. Immunol. 147:3047.

27 Dave, V. P., Cao, Z., Browne, C., Alarcon, B., Fernandez-Miguel, G., Lafaille, J., de la Hera, A., Tonegawa, S. and Kappes, D. J. 1997. CD3 delta de®ciency arrests development of the alpha beta but not the gamma delta T cell lineage. EMBO J. 16:1360. 28 Wang, B., Wang, N., Salio, M., Sharpe, A., Allen, D., She, J. and Terhorst, C. 1998. Essential and partially overlapping role of CD3gamma and CD3delta for development of alphabeta and gammadelta T lymphocytes. J. Exp. Med. 188:1375. 29 Malissen, M., Gillet, A., Ardouin, L., Bouvier, G., Trucy, J., Ferrier, P., Vivier, E. and Malissen, B. 1995. Altered T cell development in mice with a targeted mutation of the CD3-epsilon gene. EMBO J. 14:4641. 30 Haks, M. C., Krimpenfort, P., Borst, J. and Kruisbeek, A. M. 1998. The CD3gamma chain is essential for development of both the TCRalphabeta and TCRgammadelta lineages. EMBO J. 17:1871. 31 Liu, C. P., Ueda, R., She, J., Sancho, J., Wang, B., Weddell, G., Loring, J., Kurahara, C., Dudley, E. C., Hayday, A., et al. 1993. Abnormal T cell development in CD3-zeta±/± mutant mice and identi®cation of a novel T cell population in the intestine. EMBO J. 12:4863. 32 Ohno, H., Aoe, T., Taki, S., Kitamura, D., Ishida, Y., Rajewsky, K. and Saito, T. 1993. Developmental and functional impairment of T cells in mice lacking CD3 zeta chains. EMBO J. 12:4357. 33 Koyasu, S., Hussey, R. E., Clayton, L. K., Lerner, A., Pedersen, R., Delany-Heiken, P., Chau, F. and Reinherz, E. L. 1994. Targeted disruption within the CD3 zeta/eta/phi/Oct-1 locus in mouse. EMBO J. 13:784. 34 Sun, Z. J., Kim, K. S., Wagner, G. and Reinherz, E. L. 2001. Mechanisms contributing to T cell receptor signaling and assembly revealed by the solution structure of an ectodomain fragment of the CD3 epsilon gamma heterodimer. Cell 105:913. 35 Berkhout, B., Alarcon, B. and Terhorst, C. 1988. Transfection of genes encoding the T cell receptor-associated CD3 complex into COS cells results in assembly of the macromolecular structure. J. Biol. Chem. 263:8528. 36 Kim, K. S., Sun, Z. Y., Wagner, G. and Reinherz, E. L. 2000. Heterodimeric CD3epsilongamma extracellular domain fragments: production, puri®cation and structural analysis. J. Mol. Biol. 302:899. 37 Sancho, J., Chatila, T., Wong, R. C., Hall, C., Blumberg, R., Alarcon, B., Geha, R. S. and Terhorst, C. 1989. T-cell antigen receptor (TCR)-alpha/beta heterodimer formation is a prerequisite for association of CD3-zeta 2 into functionally competent TCR.CD3 complexes. J. Biol. Chem. 264:20760. 38 Jin, Y. J., Koyasu, S., Moingeon, P., Steinbrich, R., Tarr, G. E. and Reinherz, E. L. 1990. A fraction of CD3 epsilon subunits exists as disul®de-linked dimers in both human and murine T lymphocytes. J. Biol. Chem. 265:15850. 39 Kearse, K. P. 1998. Calnexin associates with monomeric and oligomeric (disul®de-linked) CD3delta proteins in murine T lymphocytes. J. Biol. Chem. 273:14152. 40 Huppa, J. B. and Ploegh, H. L. 1997. In vitro translation and assembly of a complete T cell receptor±CD3 complex. J. Exp. Med. 186:393. 41 Moingeon, P., Alcover, A., Clayton, L. K., Chang, H. C., Transy, C. and Reinherz, E. L. 1988. Expression of a functional CD3±Ti antigen/MHC receptor in the absence of surface CD2. Analysis with clonal Jurkat cell mutants. J. Exp. Med. 168:2077. 42 Tunnacliffe, A., Olsson, C. and de la Hera, A. 1989. The majority of human CD3 epitopes are conferred by the epsilon chain. Int. Immunol. 1:546. 43 David, V., Hochstenbach, F., Rajagopalan, S. and Brenner, M. B. 1993. Interaction with newly synthesized and retained proteins in the endoplasmic reticulum suggests a chaperone function for human integral membrane protein IP90 (calnexin). J. Biol. Chem. 268:9585. 44 van Leeuwen, J. E. and Kearse, K. P. 1996. Calnexin associates exclusively with individual CD3 delta and T cell antigen receptor (TCR) alpha proteins containing incompletely trimmed glycans that are not assembled into multisubunit TCR complexes. J. Biol. Chem. 271:9660. 45 van Leeuwen, J. E. and Kearse, K. P. 1996. The related molecular chaperones calnexin and calreticulin differentially associate with

400 Soluble recombinant human CD3 46 47

48 49 50

nascent T cell antigen receptor proteins within the endoplasmic reticulum. J. Biol. Chem. 271:25345. Jolliffe, L. K. 1993. Humanized antibodies: enhancing therapeutic utility through antibody engineering. Int. Rev. Immunol. 10:241. Adair, J. R., Athwal, D. S., Bodmer, M. W., Bright, S. M., Collins, A. M., Pulito, V. L., Rao, P. E., Reedman, R., Rothermel, A. L., Xu, D., et al. 1994. Humanization of the murine anti-human CD3 monoclonal antibody OKT3. Hum. Antibodies Hybridomas 5:41. Dietrich, J., Neisig, A., Hou, X., Wegener, A. M., Gajhede, M. and Geisler, C. 1996. Role of CD3 gamma in T cell receptor assembly. J. Cell Biol. 132:299. Geisler, C. 1992. Failure to synthesize the CD3-gamma chain. Consequences for T cell antigen receptor assembly, processing, and expression. J. Immunol. 148:2437. Pogue, S. L. and Goodnow, C. C. 1994. Ig heavy chain

extracellular spacer confers unique glycosylation of the Mb-1 component of the B cell antigen receptor complex. J. Immunol. 152:3925. 51 Li, Q., Santini, R. and Rosenspire, A. R. 1998. Glycosylated extracellular domains of membrane immunoglobulin M contribute to its association with mb-1/B29 gene products and the B cell receptor complex. Immunol. Invest. 27:57. 52 Portoles, P., Rojo, J., Golby, A., Bonneville, M., Gromkowski, S., Greenbaum, L., Janeway, C. A., Jr, Murphy, D. B. and Bottomly, K. 1989. Monoclonal antibodies to murine CD3 epsilon de®ne distinct epitopes, one of which may interact with CD4 during T cell activation. J. Immunol. 142:4169. 53 Muller, B., Cooper, L. and Terhorst, C. 1995. Interplay between the human TCR±CD3 epsilon and the B-cell antigen receptor associated Ig-beta (B29). Immunol. Lett. 44:97.