Conformational and biochemical differences in the TCR.CD3 complex of CD8(+) versus CD4(+) mature lymphocytes revealed in the absence of CD3gamma.

Mature CD4(+) and CD8(+) T lymphocytes are believed to build and express essentially identical surface alphabeta T-cell receptor-CD3 (TCR.CD3) complexes. However, TCR.CD3 expression has been shown to be more impaired in CD8(+) cells than in CD4(+) cells when CD3gamma is absent in humans or mice. We have addressed this paradox by performing a detailed phenotypical and biochemical analysis of the TCR.CD3 complex in human CD3gamma-deficient CD8(+) and CD4(+) T cells. The results indicated that the membrane TCR.CD3 complex of CD8(+) T lymphocytes was conformationally different from that of CD4(+) lymphocytes in the absence of CD3gamma. In addition, CD8(+), but not CD4(+), CD3gamma-deficient T lymphocytes were shown to contain abnormally glycosylated TCRbeta proteins, together with a smaller, abnormal TCR chain (probably incompletely processed TCRalpha). These results suggest the existence of hitherto unrecognized biochemical differences between mature CD4(+) and CD8(+) T lymphocytes in the intracellular control of alphabetaTCR. CD3 assembly, maturation, or transport that are revealed when CD3gamma is absent. Such lineage-specific differences may be important in receptor-coreceptor interactions during antigen recognition.

Mature CD4 ؉ and CD8 ؉ T lymphocytes are believed to build and express essentially identical surface ␣␤ T-cell receptor-CD3 (TCR⅐CD3) complexes. However, TCR⅐CD3 expression has been shown to be more impaired in CD8 ؉ cells than in CD4 ؉ cells when CD3␥ is absent in humans or mice. We have addressed this paradox by performing a detailed phenotypical and biochemical analysis of the TCR⅐CD3 complex in human CD3␥-deficient CD8 ؉ and CD4 ؉ T cells. The results indicated that the membrane TCR⅐CD3 complex of CD8 ؉ T lymphocytes was conformationally different from that of CD4 ؉ lymphocytes in the absence of CD3␥. In addition, CD8 ؉ , but not CD4 ؉ , CD3␥-deficient T lymphocytes were shown to contain abnormally glycosylated TCR␤ proteins, together with a smaller, abnormal TCR chain (probably incompletely processed TCR␣). These results suggest the existence of hitherto unrecognized biochemical differences between mature CD4 ؉ and CD8 ؉ T lymphocytes in the intracellular control of ␣␤TCR⅐CD3 assembly, maturation, or transport that are revealed when CD3␥ is absent. Such lineage-specific differences may be important in receptor-coreceptor interactions during antigen recognition.
Mature ␣␤ T lymphocytes recognize pathogen-derived peptides on antigen-presenting cells by means of the multimeric membrane protein ensemble termed the T-cell receptor (TCR) 1 ⅐CD3 complex. This TCR⅐CD3 complex includes two clonally distributed variable chains that directly interact with antigens (TCR␣ and TCR␤) and four invariant polypeptides that regulate assembly and signal transduction (CD3␥, CD3␦, CD3⑀, and ) (1). The assembly of complete TCR⅐CD3⅐ complexes takes place in a highly ordered manner within the endoplasmic reticulum: first CD3 chains, then TCR chains, and finally chains. Further conformational maturation, including carbohydrate processing, occurs in the Golgi apparatus before exportation of mature complexes to the T cell surface. The biochemical machinery involved in the assembly, processing, and exportation of TCR⅐CD3 complexes is assumed to be shared by all ␣␤ T-lineage cells. Thus, CD4 ϩ and CD8 ϩ are believed to build biochemically and conformationally identical antigen receptors, although differences in the numbers that reach or remain at the cell surface have been noted (2). Therefore, the lack of any CD3 chain would be expected to affect to a similar extent the assembly and exportation of TCR⅐CD3 complexes by mature CD4 ϩ and CD8 ϩ T cells. However, this was not the case in several murine and human CD3 deficiencies (reviewed in Ref. 3). In particular, it has been consistently shown that in the absence of CD3␥ or CD3␦, TCR⅐CD3 expression (or conformation) is more impaired in mature peripheral CD8 ϩ cells than in their CD4 ϩ counterparts, both in human and in murine deficiencies (3)(4)(5)(6)(7). Three other observations suggested the existence of CD8 ϩ cell-specific defects in human CD3␥ deficiency: first, the proband died after a viral infection (a cytolytic T-celldependent function) despite normal antibody responses (helper T-cell-dependent) (8); second, the peripheral blood CD8 ϩ T cell subset was more strongly reduced (5-fold) than the CD4 ϩ T cell subset (only 2-fold) (4); and third, the scanty peripheral CD3␥deficient CD8 ϩ cells, but not CD4 ϩ cells, failed to grow in vitro under optimal stimuli (PHA and allogeneic feeder cells) (4). This paradox may be the reflection either of subset-specific defective maturation of CD3 low to CD3 high thymocytes in these mutants (3) or of hitherto unrecognized biochemical differences in the assembly or maturation of TCR⅐CD3 complexes between peripheral CD8 ϩ and CD4 ϩ T lymphocytes, revealed only when certain CD3 chains are absent. Alternatively, as suggested previously (4), the peripheral CD8 ϩ T cells of such CD3 deficiencies may belong to a minor population (in CD3-sufficient individuals) that was relatively expanded when either CD3 chain was absent (9).
The recent availability of Herpesvirus saimiri C-488 (HVS)immortalized mature peripheral CD3␥-deficient CD8 ϩ lymphocytes (5) has allowed us to address this paradox in the case of human CD3␥ deficiency, by studying phenotypically and biochemically the TCR⅐CD3 complex of human CD3␥-deficient CD8 ϩ T cells in comparison with CD4 ϩ T cells lacking CD3␥ and with appropriate ␥-sufficient controls. The results support the notion that, indeed, there are differences in the way mature CD8 ϩ and CD4 ϩ T lymphocytes assemble or process TCR⅐CD3 complexes, which are revealed when CD3␥ is absent.
Flow Cytometry Analyses-The expression of different surface markers was studied by flow cytometry according to a standard procedure (12). The antibodies used in these studies are listed in Table I. All commercial antibodies were fluorescein isothiocyanate-or phycoerythrin-conjugated and, for the rest, a second fluorescein isothiocyanate-or phycoerythrin-conjugated antibody (anti-mouse IgG or IgM and anti-rabbit IgG, from Caltag) was used.
Intracellular stainings were done following the protocol described in Ref. 15. In all cases, cell viability was determined by the expression of CD45, and intracellular markers were analyzed within CD45 ϩ cells. Anti-CD1a (OKT6, 1:1000) and anti-CD74 (VICY1, 1:100) ascites were used as negative and positive controls, respectively, for intracellular stainings with ascitic fluids. For intracellular stainings with rabbit antisera, preimmune serum was used as a negative control. Epstein-Barr virus-transformed lymphoblastoid B cells were used as specificity controls.
For comparative stainings the mean fluorescence intensity (MFI) was used, which is defined as the average fluorescence value of the corresponding mAb referred to the logarithmic scale of fluorescence intensity along the x axis of the histograms.
Calcium Flux Measurement-Intracellular Ca 2ϩ release was induced in PBLs loaded with the fluorescent dye Fluo3AM (Sigma) according to a standard procedure (12).
Radiolabeling and Immunoprecipitation-For metabolic labeling, at least 10 7 cells were washed once with DPBS (0.14 M NaCl, 8  S]Cys) was added to the labeling tubes. Cells were incubated in these conditions for 1 h more at 37°C/5% CO 2 . Then cells were washed twice in DPBS and subsequently lysed on ice (45 min incubation) in 1% digitonine-containing lysis buffer (1% digitonine (RBI, Natick, MA), 50 mM Tris-HCl, 150 mM NaCl, 1 mM MgCl 2 , 0.1 mM EDTA, 1 mM phenylmethylsulfonyl fluoride (Sigma) and 8 mM iodoacetamide (Sigma), pH 7.6). Cell debris was removed by centrifuging the tubes at 2000 ϫ g for 10 min. Then supernatants were collected and centrifuged for 30 min at 14,000 ϫ g to eliminate the smallest cell debris. Again, supernatants, hereafter referred to as lysates, were collected and transferred to new tubes.
The lysates were precleared twice by incubation with Sepharose beads (Amersham Pharmacia Biotech) containing 1% digitonine, followed by a 2-min centrifugation at 12,000 ϫ g in an Eppendorf centrifuge at 4°C. The precleared supernatants were subsequently incubated for 1.5 h at 4°C with 0.5 ml of OKT3 or APA1/2 supernatants (anti-CD3⑀␦/⑀␥ or anti-CD3␦, respectively) coupled to protein G-Sepharose beads (Amersham Pharmacia Biotech). The beads were afterward washed five times in lysis buffer. For deglycosylation with endo-␤-Nacetylglucosaminidase H (Endo H) (Roche Molecular Biochemicals), immunoprecipitates were resuspended, after the last wash, in 45 l of denaturing buffer (5% SDS, 10% 2-mercaptoethanol) and boiled for 10 min. After a 2-min/12,000 ϫ g centrifugation, supernatants were pipetted to new Eppendorf tubes, dividing each immunoprecipitate in two tubes, one to be deglycosylated (Endo H ϩ ) and the other to be used as non-deglycosylation control (Endo H Ϫ ). 4.5 l of 10ϫ Endo H buffer (0.5 M sodium citrate, pH 5.5) and 1 l of Endo H (1000 IU/ml) were added to the Endo H ϩ samples, which were in turn incubated overnight at 37°C. The Endo H Ϫ tubes were meanwhile kept at 4°C. Samples were then resuspended in Laemmli sample buffer and boiled for 5 min before a short spin at 12,000 ϫ g. SDS-polyacrylamide gel electrophoresis was performed on 10% or 12% polyacrylamide gels (see legends to Figs. 5 and 6), and the samples were analyzed by autoradiography or in a PhosphorImager SI (Molecular Dynamics, Sunnyvale, CA). The films were electronically scanned (Bio-Rad Geldoc 2000 analyzer) to determine the relative intensity and molecular weight of each protein.
For surface labeling, at least 10 7 cells were washed twice with phosphate-buffered saline and resuspended in 150 l of phosphate-buffered saline. Then cells were 125 I-labeled using the lactoperoxidase method by adding 2 mCi of Na 125 I (Amersham Pharmacia Biotech), 30 l of a 140 IU/ml lactoperoxidase solution (Sigma), and 10-l aliquots of a 0.06% H 2 O 2 solution four times at 5-min intervals. To stop the labeling reaction, a solution of 20 mM KI and 1 mM tyrosine phosphate-buffered saline was added to the tubes (Sigma). The samples were then lysed in a lysis buffer containing 1% Brij 96 (Sigma), 150 mM NaCl, 20 mM Tris-HCl, pH 7.8, 10 mM iodoacetamide, 1 mM phenylmethylsulfonyl fluoride, 1 g/ml each leupeptin and aprotinin. The lysates were precleared twice with Sepharose beads as described above. Immunoprecipitation was performed with 2-4 g of APA 1/1 (anti-CD3⑀) coupled to protein A-Sepharose beads (Amersham Pharmacia Biotech). This antibody has been shown to coprecipitate CD3␥ and CD3␦ chains (12), which can be then resolved by deglycosylation. Samples were then divided in two, and after five washings with lysis buffer, one part was digested overnight with N-glycosidase F (N-Gly ϩ , Roche Molecular Biochemicals), whereas the other was left undigested (N-Gly Ϫ ). Briefly, N-Gly ϩ samples were denatured by boiling for 2 min in 7 l of a solution containing 0.5% SDS and 0.8% 2-mercaptoethanol. After cooling the samples on ice, 15 l of N-Gly buffer (250 mM Na 2 PO 4 , pH 8.8, 1% Nonidet P-40, 10 mM EDTA) were added to all tubes, and then 1 l/sample of 200 unit/ml N-Gly was added to the N-Gly ϩ samples. N-Gly ϩ tubes were incubated overnight at 37°C, whereas N-Gly Ϫ tubes were maintained at 4°C. Finally, all samples, N-Gly ϩ and N-Gly Ϫ , were resuspended in Laemmli sample buffer, boiled for 5 min before a short spin at 12,000 ϫ g, and resolved by SDS-polyacrylamide gel electrophoresis on 12% polyacrylamide gels. Northern Blot Analysis-Total RNA was isolated with Ultraspec (Biotecx, Houston, TX) following the manufacturer's instructions, and then at least 10 g of RNA were run on 1% agarose-formaldehyde gels, transferred to nylon membranes, and hybridized as described elsewhere (Ref. 16 and references therein) with 32 P-labeled cDNA probes corresponding to the TCR C␣, C␤ (17), C␥ (18), C␦ (19), and pT␣. The same blot was subsequently stripped and hybridized with a ␤-actin probe.
V␤ Usage Analysis by Reverse Transcription-PCR-Total RNA (2 g) was reverse-transcribed into cDNA using an oligo-dT primer according to the manufacturer's protocol (Roche Molecular Biochemicals). cDNA from HVS-immortalized cells or from PBLs was amplified by PCR using 22 different 5Ј V␤-specific primers (V␤1-V␤20; see Fig. 7) together with a common 3Ј C␤-specific primer (20). Amplified products were electrophoresed, transferred to nylon membranes, and analyzed by Southern blot hybridization with a C␤ probe (Jur␤2, 17).
Statistical Analysis-Student's t test was used for all comparisons. Only p values below 0.05 were considered significant. Data are presented as mean Ϯ S.D.

RESULTS
Peripheral Blood CD8 ϩ Lymphocytes from a Human CD3␥ Deficiency Were Functional T Cells with a Normal V␤ Repertoire-The barely detectable surface expression of certain TCR⅐CD3 epitopes in CD8 ϩ lymphocytes from human CD3␥ deficiency ( Fig. 1A and Ref. 4), together with the fact that they did not grow in optimal T-cell culture conditions (4), could be interpreted as an indication that they were not T lymphocytes at all but rather NK-lineage cells. Our previous phenotypic analysis of such CD8 ϩ lymphocytes did not support that hypothesis for two reasons: 1) most CD8 ϩ cells in CD3␥ deficiency were CD8 bright (that is, CD8␣␤ ϩ ), whereas CD8 ϩ NK cells are normally CD8 dull (CD8␣␣ ϩ ); and 2) NK markers (CD16, CD56, and CD57) were not overrepresented within the CD8 ϩ cell subset in CD3␥ deficiency (Ref. 4). Nevertheless, an additional phenotypical and functional assay was performed to further ascertain the T-lineage descent of CD3␥-deficient CD8 ϩ cells and to rule out their putative NK-lineage descent. First, V␤ usage within CD8 bright cells was assayed by cytofluorometry and found to be normal (Fig. 1B). Second, an early TCR⅐CD3dependent functional response, namely calcium flux, was tested and found also to be normal (Fig. 2). This ruled out the existence of an early signal transduction defect in CD8 ϩ ␥ Ϫ cells as a cause of the observed selective growth defect. Taken together, these results suggested that peripheral blood CD8 ϩ cells from human CD3␥ deficiency were polyclonal, functionally competent T lymphocytes.
The Membrane TCR⅐CD3 Complex of Immortalized CD3␥-deficient CD8 ϩ T Lymphocytes Was Conformationally Different from That of CD3␥-deficient CD4 ϩ Lymphocytes-The limited number of peripheral blood CD8 ϩ lymphocytes in CD3␥ deficiency and our inability to obtain CD8 ϩ T cell lines (4) precluded further studies on the CD8 ϩ T cell subset. This prompted us to try HVS immortalization. An HVS-immortalized ␥-deficient T cell line termed DSF8 (CD8␣␤ ϩ , data not shown) was obtained and characterized phenotypically with a large panel of TCR⅐CD3-specific mAbs (Fig. 3). For comparative purposes, the results for each mAb are depicted as the MFI ratio of ␥ ϩ cells to ␥ Ϫ cells, either CD8 ϩ (Fig. 3A, black columns) or, for comparison, CD4 ϩ (Fig. 3A, shaded columns) (5). A ␥ ϩ :␥ Ϫ ratio close to 1 (indicated by the horizontal dotted line) reflects an equivalent expression of the epitope detected by that particular mAb in ␥ Ϫ and ␥ ϩ cells (see, for example, CD2 and CD4 or CD8). ␥ ϩ :␥ Ϫ ratios above 1 reflect an impaired expression of the detected epitope by ␥ Ϫ cells relative to ␥ ϩ controls (impaired 2-, 3-, and 4-fold for ratios of 2, 3, and 4, respectively, and so forth). By this criterion, three conclusions emerged. First, all anti-CD3 anti-TCR␣␤ mAbs used stained both CD4 ϩ ␥ Ϫ and CD8 ϩ ␥ Ϫ cells poorly relative to ␥ ϩ controls (␥ ϩ :␥ Ϫ MFI ratios Ն2; Fig. 3A). Second, there were, however, clear relative staining differences among different mAbs, which were shared by CD4 ϩ ␥ Ϫ and CD8 ϩ ␥ Ϫ cells. Thus, the ␥ ϩ :␥ Ϫ MFI was around 2 when SPV.T3b was used, but it reached 4 when RW2-8C8, X35, UCHT-1, or OKT3 were analyzed (see Fig. 3A and representative stainings of the two patterns in Fig. 3B). These results suggest that in addition to displaying lower overall levels of TCR⅐CD3 complexes, CD3␥ Ϫ cells were more deficient in the relative expression of certain TCR⅐CD3 epitopes. The relative expression of other T cell surface molecules was not affected in ␥ Ϫ cells (e.g. CD2, Fig. 3, A and B). A mAb hierarchy of CD3␥-dependence may be established based on these binding results, as shown in Fig. 3A. In general, TCR␣␤specific mAbs bound ␥ Ϫ T cells more poorly than CD3-specific mAbs, relative to ␥ ϩ controls. It has been shown that TCR⅐CD3 recognition by certain mAbs may be influenced by glycosylation (e.g. WT31, Ref. 21), and we have reported previously that the lack of CD3␥ affects TCR⅐CD3 assembly and glycosylation in CD4 ϩ T cells (12,22). Therefore, the impaired binding observed in Fig. 3 may in fact be a reflection of the higher or lower glycosylation of the particular epitope recognized by each mAb. Third, as observed in PBLs (Fig. 1A), TCR⅐CD3 binding was significantly more impaired in CD8 ϩ ␥ Ϫ cells than in CD4 ϩ ␥ Ϫ cells, relative to ␥ ϩ controls, with three mAbs: Leu4, 2Ad2, and BMA031 (see representative reactivity patterns in Fig. 3B).
These findings showed that, in the absence of CD3␥, immortalized CD8 ϩ cells expressed essentially the same number of membrane TCR⅐CD3 complexes as CD4 ϩ cells (as shown with SPV.T3b), which were, however, different from those of CD4 ϩ cells (as shown with Leu4, 2Ad2, and BMA031). These results are concordant with PBLs and thus validate the immortalized ␥ Ϫ cells as a model system to study such differences. The general decrease in TCR⅐CD3 staining observed in ␥ Ϫ cells (Fig.  3) could be due to the lack of intracellular subunits (other than CD3␥) to build TCR⅐CD3 complexes (due, for instance, to degradation), or to poor transport of the ␥ Ϫ TCR⅐CD3 complex to the surface. To distinguish between these two possibilities, intracellular stainings of permeabilized ␥ Ϫ and ␥ ϩ T cells, both CD4 ϩ and CD8 ϩ , were performed and compared as above. The results indicated that ␥ Ϫ cells had slightly more intracellular CD3␦, CD3⑀, and TCR␤ relative to ␥ ϩ controls (Fig. 4). Therefore, these data are consistent with the hypothesis that the decreased cell surface TCR⅐CD3 expression seen in ␥ Ϫ T cells (Fig. 3) is likely due to poor transport of the TCR to the cell surface.
Taken together, the phenotypic data suggested that HVSimmortalized CD8 ϩ cells lacking CD3␥ built surface TCR⅐CD3 complexes, which were conformationally different from those of CD4 ϩ ␥ Ϫ cells, despite the apparent intracellular availability of several relevant TCR and CD3 components other than CD3␥. This was found within the background of a poor transport of ␥ Ϫ TCR⅐CD3 complexes to the surface of both cell types. Such CD8 ϩ cell-specific features could be due to: 1) a change in TCR⅐CD3 subunit composition; 2) a change in TCR⅐CD3 subunit glycosylation; or 3) a change in accessibility of certain TCR⅐CD3 epitopes (due to steric hindrance). Further studies were thus undertaken to explore the first two possibilities.
Immortalized CD3␥-deficient CD8 ϩ , but Not CD4 ϩ , Lymphocytes Lacked Normal TCR␣ Chains-Unimmortalized CD3␥deficient CD4 ϩ cells had previously been analyzed by immunoprecipitation (12,22) and immunohistochemistry (23) and had been found to contain all TCR␣␤ and CD3 chains except CD3␥. Thus, the impaired expression of their mutant TCR⅐CD3 complex could be explained as a simple consequence of the lack of CD3␥ on several CD3␥-dependent TCR⅐CD3 epitopes. The unexpected finding that some of the same epitopes were conformationally different in CD8 ϩ cells prompted us to perform biosynthetic studies on these cells.
First, a [ 35 S]methionine/cysteine pulse-chase immunoprecipitation was done using OKT3 (anti-CD3, Fig. 5A) and nonreducing conditions. The results confirmed the lack of CD3␥ in CD8 ϩ ␥ Ϫ cells and an impaired association of 2 to the mutant complex, as described previously in unimmortalized CD4 ϩ ␥ Ϫ cells (12,22). Interestingly, whereas in ␥ ϩ cells, the immature ␣␤ complex (␣␤i) progressively gave rise to the mature form (␣␤m), in ␥ Ϫ cells, the complex had an intermediate size throughout the experiment. Furthermore, the CD3-associated TCR␣ single chains, either totally (␣tg) or partially (␣pg) glycosylated, were not detected in ␥ Ϫ cells. Single TCR␤ chains (␤tg), in contrast, were present and were apparently more stable in ␥ Ϫ cells. However, an additional labeled protein running slightly faster than ␤tg was observed only in ␥ Ϫ cells. This protein could be a partially glycosylated form of the TCR␤ protein, which would be consistent with the hypothesis that the transport and processing of the TCR␤ chain is slowed in the absence of CD3␥. Alternatively, the additional labeled protein could represent a new, unidentified protein (see below). Therefore, the TCR heterodimer in CD8 ϩ ␥ Ϫ cells was smaller in size than the normal mature form found in ␥ ϩ cells and lacked the maturation-associated (glycosylation) size changes found in normal TCR␣␤ heterodimers. In addition, a lack of CD3-associated TCR␣ chains and CD3 was prominent, and an additional protein, probably partially glycosylated TCR␤, was detected in CD8 ϩ ␥ Ϫ cells.
Second, a more detailed analysis of [ 35 S]methionine/cysteinelabeled CD3-associated TCR chains was performed under reducing conditions (Fig. 5B), with or without Endo H treatment to remove pre-Golgi N-linked oligosaccharides. The results using two different anti-CD3 mAbs (OKT3 and APA1/2) clearly confirmed that the TCR composition of CD8 ϩ ␥ Ϫ cells was different from that of ␥ ϩ controls. Indeed, it involved not only the loss of normal TCR␣ (␣dg) as observed in Fig. 5A, but also the addition of another, unidentified 32-kDa protein (Fig. 5B, *). By contrast, TCR␤ (␤dg) was present and normal in size in CD8 ϩ ␥ Ϫ cells. There were also some notable differences be- tween ␥ Ϫ Endo H Ϫ lanes and ␥ ϩ Endo H Ϫ lanes: 1) as shown in Fig. 5A, totally glycosylated TCR␣ (␣tg) chains were absent or very reduced in ␥ Ϫ cells; 2) as in Fig. 5A, totally glycosylated TCR␤ (␤tg) was present in ␥ Ϫ cells, and an additional labeled protein running slightly faster than ␤tg (probably partially glycosylated TCR␤) was observed only in ␥ Ϫ cells; and 3) a very strongly labeled 38-kDa protein was prominent in ␥ Ϫ cells, coincident in apparent molecular mass with partially glycosylated TCR␣ (␣pg). This protein could represent the glycosylated form of the unidentified 32-kDa protein (Fig. 5B, *). However, it may also contain some abnormally glycosylated forms of TCR␤, because the labeling intensity of deglycosylated TCR␤ (␤dg) apparently exceeds the combined labeling intensity of normal glycosylated TCR␤ (␤tg) and the additional labeled protein mentioned above. For comparative purposes, immortalized CD4 ϩ cells, both ␥ ϩ and ␥ Ϫ , and CD8 ϩ ␥ Ϫ cells were [ 35 S]methionine/cysteine-labeled and analyzed under reducing conditions using OKT3, with or without Endo H (Fig. 5C). The results indicated that immortalized CD4 ϩ ␥ Ϫ cells contained normal-sized TCR␣ (␣tg; ␣pg is not detected in this particular gel) and TCR␤ (␤tg) proteins and an additional minor labeled protein running slightly faster than ␤tg (again, probably partially glycosylated TCR␤). CD8 ϩ ␥ Ϫ cells, in contrast, lacked normal TCR␣ (␣tg) but showed normal TCR␤ (␤tg; partially glycosylated TCR␤ is not apparent in this particular gel), and the strongly labeled 38-kDa protein seen in Fig. 5B, which perhaps is the glycosylated form of the unknown 32-kDa protein (Fig. 5C, *), and some glycosylated form of TCR␤.
Finally, direct immunoprecipitations of TCR chains were carried out in [ 35 S]methionine/cysteine-labeled cells and analyzed under reducing conditions (Fig. 5D). The results using an anti-TCR␣ mAb (␣F1, Ref. 13) confirmed the absence of normal TCR␣ in CD8 ϩ ␥ Ϫ cells but not in CD8 ϩ ␥ ϩ controls or in CD4 ϩ cells (␥ ϩ or ␥ Ϫ ). The TCR␣ chain of the CD8 ϩ ␥ ϩ sample had slightly smaller (ϳ1 kDa) relative molecular mass than in CD4 ϩ cells, perhaps due to genetic variation (the specific transcripts were also slightly smaller in this particular cell line, see below and Fig. 8). Similar results (i.e. lack of normal TCR␣ in CD8 ϩ ␥ Ϫ cells, sample ␥ 1 Ϫ ) were obtained with an unrelated antiserum against TCR␣ (H36, Ref. 14). The lack of TCR␣ was confirmed with H36 in an independently derived HVS-immortalized CD8 ϩ ␥ Ϫ cell line termed D8EDTA (sample ␥ 2 Ϫ in Fig.  5D). By using an anti-TCR␤ mAb (␤F1, Ref. 13), totally glycosylated TCR␤ (␤tg) was detected in CD8 ϩ ␥ Ϫ cells as well in their CD8 ϩ ␥ ϩ controls, and also in immortalized CD4 ϩ ␥ Ϫ cells and their ␥ ϩ controls (Fig. 5D). Interestingly, in addition to ␤tg, two highly labeled 40-and 42-kDa proteins were observed in CD8 ϩ ␥ Ϫ cells. The proteins were also present in CD8 ϩ ␥ ϩ controls, although at clearly lower relative levels, despite the fact that the number of cells used for the analysis was similar (5 ϫ 10 6 in all lanes). Such excess of ␤F1-reactive proteins in CD8 ϩ ␥ Ϫ cells may explain the significant increase of intracellular ␤F1 binding observed in CD8 ϩ ␥ Ϫ cells, as compared with CD4 ϩ ␥ Ϫ cells (relative to ␥ ϩ cells, Fig. 4A). The two additional proteins are probably glycosylated forms of TCR␤, because they resolved into a single protein with the apparent molecular mass of deglycosylated TCR␤ (␤dg) after Endo H digestion. Indeed, at least one of these proteins may be present in the very strongly labeled 38-kDa CD3-associated protein observed in CD8 ϩ ␥ Ϫ cells (Fig. 5, B and C). Alternatively, they may represent unidentified proteins that pair with TCR␤ in some T cell types. Similar additional proteins have also been observed in Jurkat cells using ␤F1 (24).
Collectively, the results of the biosynthetic studies indicated that immortalized CD8 ϩ cells lacking CD3␥ contained an abnormally small TCR heterodimer composed of a normal TCR␤ chain, some abnormally glycosylated TCR␤ chains and an additional unidentified chain that was smaller than normal TCR␣ and could not be recognized by two unrelated TCR␣-specific antibodies. To ascertain whether the intracellular findings were representative of the surface mutant TCR⅐CD3 complex, CD8 ϩ ␥ Ϫ (and, as a control, CD8 ϩ ␥ ϩ ) cells were radioiodinated, lysed, and immunoprecipitated with APA1/1 (anti-CD3⑀). Half of the precipitates were digested with N-Gly to remove all N-linked oligosaccharides before electrophoresis under reducing conditions. The results are shown in Fig. 6A. Clear differences were observed between ␥ Ϫ and ␥ ϩ cells, both in TCR and in CD3 chains. In contrast to biosynthetic results, the amount of CD3-associated TCR proteins was clearly lower (around 3-fold by electronic densitometry) in ␥ Ϫ cells than in ␥ ϩ cells, so that only major TCR proteins were visible in ␥ Ϫ samples. With this limitation, the surface-labeled TCR of ␥ Ϫ cells (Fig. 6A $) apparently lacked both normal (i.e. fully glycosylated) TCR␣ and TCR␤. The lack of normal TCR␣ in CD8 ϩ ␥ Ϫ cells confirmed previous biosynthetic results (see above, Fig. 5). The lack of normal TCR␤, in contrast, was not found in those biosynthetic experiments, in which TCR␤ was present as a minor but detectable protein (Fig. 5, B and C). Thus, it cannot be ruled out that the surface TCR of CD8 ϩ ␥ Ϫ cells contained a small amount of fully glycosylated TCR␤ that cannot be detected by radioiodination due to the impaired association of surface TCR chains to CD3 in ␥ Ϫ cells. Two labeled TCR chains, with apparent molecular masses of 40 and 43 kDa, respectively, are discernable in ␥ Ϫ cells before deglycosylation. Such proteins could in part represent abnormally glycosylated TCR␤, as they are reminiscent in size and intensity of the two highly labeled proteins that coprecipitated with normal TCR␤ when ␤F1 was used (Fig. 5D) and, in the case of the smaller one, of the strongly labeled 38-kDa protein coprecipitated with CD3 in Fig. 5, B and C. The fact that, upon deglycosylation, a major portion of those proteins had the same apparent molecular mass as normal deglycosylated TCR␣ ϩ TCR␤ ((␣ ϩ ␤) dg) would lend support to this notion. In addition, a minor portion of the 40-or 43-kDa protein resolved upon deglycosylation into an abnormally small protein of about 30 kDa (Fig. 6A, *), which may be identical to the deglycosylated 32-kDa protein precipitated with anti-CD3 in Fig. 5, B and C. This unidentified protein may represent the protein that is associated with TCR␤ in place of normal TCR␣ in CD8 ϩ ␥ Ϫ cells. These results confirmed that the surface TCR composition of CD8 ϩ ␥ Ϫ cells was essentially similar to the aberrant intracellular complex characterized previously. However, in the extracellular complex, an impaired association of TCR chains to CD3 components was observed, and neither normal totally glycosylated TCR␤ nor TCR␣ was detectable. Rather, abnormally glycosylated TCR␤ and an unidentified 30-kDa chain were observed. CD8 ϩ ␥ Ϫ showed also some clear differences in CD3 chains composition as compared with CD8 ϩ ␥ ϩ controls, which were not evident in biosynthetic experiments. Indeed, the ␥ Ϫ sample had, in addition to normal CD3⑀, one or more highly labeled proteins with an apparent molecular mass in the range of 25-30 kDa, within the size range of normal CD3␥ and CD3␦ (Fig. 6A, ?). Upon deglycosylation, the protein or proteins had the same apparent molecular mass as normal CD3␦ (␦dg). Therefore, the 25-30-kDa proteins may be abnormally glycosylated CD3␦. However, due to the increased amount of labeled protein in the deglycosylated ␥ Ϫ sample, there could also be more than one protein with a size similar to that of CD3␦ present. To further substantiate that the highly labeled 25-30-kDa protein(s) of CD8 ϩ ␥ Ϫ cells was abnormally glycosylated CD3␦, it was eluted from the gel, deglycosylated with Endo H, and analyzed by electrophoresis under reducing conditions. Normal CD3␥ and CD3␦ were analyzed in parallel, as controls. The results are shown in Fig.  6B. The normal CD3␥ sample showed mature glycosylation, whereas normal CD3␦ showed about half mature and half immature glycosylation. The 25-30-kDa protein(s) from ␥ Ϫ cells did not run at the same size as CD3␥ or CD3␦ from ␥ ϩ cells, and, in contrast to normal CD3␦, showed no change in size upon Endo H treatment, consistent with mature glycosylation of surface-labeled proteins. Taken together, these results suggested that the surface TCR⅐CD3 complex of ␥ Ϫ cells contained normal CD3⑀ and an additional protein, which most likely represented abnormally glycosylated CD3␦. Experiments are in progress to prove the identity of such protein, by amino acid analysis and partial protease digestion of the deglycosylated form of CD3␦ in ␥ Ϫ versus ␥ ϩ cells.
The surface labeling experiments collectively indicated that immortalized CD8 ϩ ␥ Ϫ cells expressed an aberrant TCR⅐CD3 complex with fewer CD3-associated TCR proteins, lacking nor- mal TCR␣ and normally glycosylated TCR␤ among them, and with what appeared to be abnormally glycosylated forms of TCR␤ and CD3␦, together with an unidentified 30-kDa protein (in deglycosylated form), which may represent the protein paired with TCR␤ on the ␥ Ϫ cell surface. Some of these biochemical results were reminiscent of TCR␥␦ ϩ cells (25), but the possibility that CD8 ϩ ␥ Ϫ cells were TCR␥␦ ϩ , rather than aberrant TCR␣␤ ϩ , was ruled out by staining with ␥␦-specific monoclonals TCR␦1 (26) and 11F2 (27) (data not shown) and by Northern blot analysis (see below).
TCR Gene Expression Was Normal in Immortalized CD3␥deficient CD8 ϩ Lymphocytes-The lack of normal TCR␣ in CD8 ϩ ␥ Ϫ cells suggested the possibility that they expressed alternative mature (␥␦ ϩ ) or immature (␤pT␣ ϩ ) TCR ensembles. To explore such possibilities, total RNA was isolated from CD8 ϩ ␥ Ϫ cells, and from appropriate controls, and probed for several TCR specific transcripts by Northern analyses (Fig. 8).
The results demonstrated that CD8 ϩ ␥ Ϫ cells contained normal levels and sizes of mature TCR␣ and TCR␤ mRNA, whereas they lacked expression of TCR␦, TCR␥, and pT␣ transcripts. In addition, they did not display germline transcription at the TCR locus and, therefore, T early ␣ transcripts were absent (data not shown). These results support the conclusion that CD8 ϩ ␥ Ϫ cells are not NK cells, nor ␥␦ T cells. Interestingly, they express the TCR␣ gene at the transcriptional level, although it is apparently not expressed at the protein level. DISCUSSION In the present study, we have investigated TCR⅐CD3 assembly and expression in immortalized natural mutant CD8 ϩ and CD4 ϩ T lymphocytes lacking CD3␥. We were prompted to un-dertake this study by the unexpected observation that peripheral blood TCR⅐CD3 expression was more impaired in CD8 ϩ than in CD4 ϩ cells when CD3␥ is absent (Refs. 4 and 5 and Fig.  1A). Despite their phenotypical defect, CD8 ϩ ␥ Ϫ cells were functional (Fig. 2). Therefore, we tried to understand the mechanisms behind the feeble expression of surface TCR⅐CD3 on those cells. The unexpected finding was that, in addition to lacking CD3␥, immortalized CD8 ϩ ␥ Ϫ cells, but not CD4 ϩ ␥ Ϫ cells, lacked or expressed few mature TCR␣ chains (see below).
The consequence of this biochemical finding was that membrane TCR⅐CD3 complexes from immortalized CD3␥-deficient CD8 ϩ T cells were conformationally different from those of CD3␥-deficient CD4 ϩ T cells or normal CD8 ϩ and CD4 ϩ T cells (Fig. 3). These conformational differences may explain the selective TCR⅐CD3 expression impairment detected previously in peripheral blood (4, 7) 2 or immortalized (Ref. 5 and present data) murine or human CD8 ϩ ␥ Ϫ T cells by cytofluorometry. Greater sensitivity of CD8 ϩ cells than CD4 ϩ cells to the absence of CD3␥ (in terms of TCR⅐CD3 surface expression) might be related to differences in the intracellular control of ␣␤ TCR assembly, maturation, and/or transport between the two lin-2 M. van Tol, personal communication.
eages. The lower levels of surface ␣␤ TCR found in normal mature CD8 ϩ T lymphocytes (2) may be the reflection of such biochemical differences (thus qualitative rather than quantitative) and perhaps relevant in the context of receptor-coreceptor interactions (see below). Indeed, the comparative staining of peripheral blood CD4 ϩ and CD8 ϩ cells using a broad panel of TCR⅐CD3-specific mAbs revealed that, in contrast to expectations, certain mAbs stained CD8 ϩ cells better than CD4 ϩ cells. 3 A similar situation, i.e. increased sensitivity of CD8 ϩ versus CD4 ϩ cells in TCR⅐CD3 expression, has also been observed in mice lacking CD3␦ (3,6). The biochemical basis, however, remained unexplored, perhaps due to the scant peripheral blood lymphoid compartment.
A further study of the influence of CD3␥ deficiency on expression of other TCR⅐CD3 components was performed. Pulsechase experiments clearly showed the following points: 1) absence of CD3␥ chains, 2) absence of CD3 chain coprecipitation, and 3) absence of TCR␣␤ heterodimer maturation in CD8 ϩ ␥ Ϫ T cells (Fig. 5A). In addition, TCR␤ chains seemed more stable, a finding that, together with absence of CD3 coprecipitation and of TCR␣␤ heterodimer maturation, strongly indicated that TCR⅐CD3 complexes (or at least most of them) stayed in the endoplasmic reticulum (29). The same results were obtained when several different anti-TCR⅐CD3 mAbs were used in the other immunoprecipitations.
Nonglycosylated TCR␣ chains have a molecular mass of about 30 kDa, and partially or fully glycosylated TCR␣ chains have molecular masses of about 40 -45 kDa (up to six N-linked sugars). Nonglycosylated TCR␤ chains have a molecular mass of 35 kDa, and partially or fully glycosylated TCR␤ chains have a molecular mass of about 37-40 kDa. Deglycosylation experiments with Endo H demonstrated that deglycosylated TCR␤ chains from the CD3␥-deficient CD4 ϩ or CD8 ϩ cells migrate to the same position. However, deglycosylated TCR␣ chains from CD3␥-deficient CD8 ϩ T cells have a molecular mass of 2-3 kDa lower than normal (32-34 versus 30 -32 kDa; see Fig. 5, B and C). As TCR C␣ cDNA sequences in CD4 ϩ and CD8 ϩ CD3␥deficient T cells are the same as in normal T cells, 4 the most logical explanation seems to be that TCR␣ chains from CD3␥deficient CD8 ϩ T cells are more easily deglycosylated compared with CD3␥-deficient CD4 ϩ T cells (or normal T cells). This may be due to the fact that TCR␣ chains in CD3␥-deficient CD8 ϩ T cells are more immature than those of CD3␥-deficient CD4 ϩ T cells or normal T cells. It is likely that immature TCR␣ loses recognition by ␣F1 and other C␣-specific reagents. Indeed, the ␣F1 monoclonal recognizes an exposed epitope spanning residues 141-159 of the C␣ domain (13), including an Asn residue, which is one of the four potential N-glycosylation sites of the human C␣ domain (30). Loss of recognition by ␣F1 has been reported previously in a Jurkat variant lacking the transmembrane and cytoplasmic domains of TCR␣ (31). Parallel studies on CD3␥-deficient Jurkat T cells have shown that 1) an inverse relationship exists between the amount of TCR␣ chains coprecipitated and the amounts of CD3␥ chains produced in a T cell line, and 2) the production of only partially glycosylated CD3␥ chains influences TCR␣ chain glycosylation. 5 Abnormal assembly and glycosylation of the TCR⅐CD3 complex have been reported previously in CD4 ϩ ␥ Ϫ cells (12,22). However, the differences in TCR␣ observed in CD8 ϩ versus CD4 ϩ ␥-deficient cells have not been reported previously. Nascent TCR␣ chains have been shown to be uniquely unstable in immature murine T cells (32), perhaps because of altered proc-essing of oligosaccharide side-chains in the endoplasmic reticulum (33). These differences are believed to cause the reduced TCR⅐CD3 expression observed in immature versus mature T lymphocytes (34) or in transfected eukaryotic cells (35). Similarly, our results suggest that human mature CD8 ϩ and CD4 ϩ T lymphocytes may differ in the intracellular control of TCR⅐CD3 assembly, maturation, or transport. Such differences would be revealed when CD3␥ is absent, particularly in TCR␣, which is the most unstable chain.
The most difficult data to interpret are those from external labeling of TCR⅐CD3 complexes from CD3␥-deficient CD8 ϩ T cells: 1) few TCR molecules are labeled/coprecipitated (in proportion to CD3 chains precipitated), and 2) deglycosylated TCR␣ chains have an apparent molecular mass lower than deglycosylated TCR␣ chains from normal T cells (Fig. 6). In addition, CD3␥-deficient CD8 ϩ T cells seem to express CD3␦ (and probably TCR␤) chains, which are more glycosylated than the normal counterparts. Thus, are the TCR⅐CD3 complexes on the surface of CD3␥-deficient CD8 ϩ T cells normal, with expression of low levels of mature TCR␣ chains? Or alternatively, are the TCR⅐CD3 complexes on the surface of CD3␥-deficient CD8 ϩ T cells constructed with a TCR␣ chain different from normal TCR␣ chains?
Despite the observed biochemical differences in the TCR⅐CD3 complex of CD8 ϩ ␥ Ϫ versus CD4 ϩ ␥ Ϫ cells, their functional behavior was undistinguishable when several CD3-induced activation events were analyzed (5). Thus, it may well be that the peripheral T lymphocytes that we can analyze when CD3␥ is absent are those carrying TCR⅐CD3 complexes that are conformationally compatible with the corresponding coreceptor. A direct interaction of the TCR with CD4 and CD8 has been proposed before (36 -39).
TCR⅐CD3-induced calcium flux was normal in CD4 ϩ and CD8 ϩ CD3-deficient PBLs (Fig. 2) and HVS-immortalized (5) T cells. However, it was partially impaired in an unimmortalized CD4 ϩ T cell line (DIL2) derived from the same donor (12). This discrepancy may have been due to clonal variation of DIL2 or, alternatively, to differences in the selective pressures imposed by the culture conditions. DIL2 was grown with PHA, which is a strong TCR⅐CD3 stimulant, whereas HVS cells grow by TCR⅐CD3-independent interactions (11) and may therefore be more representative of PBL behavior.