Human CD8β, But Not Mouse CD8β, Can Be Expressed in the Absence of CD8α as a ββ Homodimer¹

CD8, a cell-surface molecule expressed on MHC class I-restricted T cells, functions as a coreceptor with the TCR. Two molecular forms of CD8 exist on mature T cells: disulfide-linked homodimers of CD8α and heterodimers of CD8α and CD8β. The CD8αα homodimer is exclusively expressed on subsets of intraepithelial lymphocytes of the intestine (1) and on human NK cells (2).

The crystal structure of CD8αα demonstrates that the structure of the N-terminal 114 aa of each subunit is typical of a light-chain variable domain and that the two Ig domains of the dimer forms a structure very similar to an Ab-variable region (3). The connecting peptide region is a 48-residue linker that precedes the transmembrane sequence and has an extended structure (4). This region contains two cysteine residues that are thought to form disulfide bridges critical for the formation of CD8 dimers and also contains a number of O-linked glycosylation sites. All Ig molecules also contain a canonical disulfide bond between cysteine residues on the B- and F-strands. CD8α has both of these conserved cysteines and also contains an additional conserved cysteine residue on the C-strand. Biochemical and crystal structure analysis provided contradictory results regarding which two cysteines were involved in the formation of the disulfide bridge. All three crystal structures involving CD8α (mouse and human) showed that the B- and F-strand cysteines were the ones making the disulfide bridge (3, 5, 6), whereas two separate biochemical studies have demonstrated that the B- and C-strand cysteines made the disulfide bond in CD8α (7, 8). In this study, we demonstrate that the B- and F-strand cysteines are critical for the proper folding and expression of dimeric CD8 using a mutational analysis.

Functional differences exist between homodimer and heterodimer forms of CD8, although the mechanisms responsible for these differences are not fully understood. It has been shown that both molecules are critical for T cell development (9, 10, 11) and several studies indicate that CD8αβ functions as a better coreceptor with the TCR than CD8αα. One hypothesis to explain the improved coreceptor activity of CD8αβ was that CD8αβ had a stronger affinity for MHC class I. However, we showed that, in the absence of the TCR, CD8αβ does not bind better to MHC class I than CD8αα (12). In addition, we performed a mutational analysis to determine the orientation of CD8αβ relative to MHC class I. The crystal structures of CD8αα and MHC class I demonstrated that the contribution of the two CD8α domains was asymmetric, with one domain contributing more of the contact residues than the other (5). We found that, for human CD8αβ, CD8α corresponded to the domain making 70% of contact and CD8β corresponded to the one making 30% of the contact. Thus, this supported the hypothesis that CD8β played a lesser role in the interaction of CD8αβ with MHC class I (13).

Studies on the function of CD8β have been hindered by the apparent inability of CD8β to be expressed without CD8α (14, 15, 16). In this study, we demonstrate that human, but not mouse, CD8β can be expressed on the cell surface as a homodimer. Creation of chimeric molecules to map the region of the CD8β molecule responsible for this difference demonstrated that only molecules with human CD8β Ig domain could be expressed on the cell surface. This is consistent with studies on TCR Vα-chains in which the Ig domains have also been shown to be critical for those Vα-chains that are able to form homodimers (17, 18). The expression of CD8ββ homodimers has allowed us to directly examine the ability of CD8β homodimers to bind to MHC class I. Results presented here illustrate that, in the absence of CD8α, CD8β cannot mediate binding to MHC class I.

All cell lines were grown in RPMI 1640 medium supplemented with 2 mM l-glutamine, 100 U penicillin/ml, 100 μg streptomycin, and 10% FBS. The cell lines used included the EBV-transformed B cell line UC transfected with luciferase gene (19) and the monkey kidney cell line COS-7 (20).

To determine which domain of human CD8β (hCD8β)³ was responsible for its ability to be expressed in the absence of CD8α, two CD8β chimeric constructs were created. One contained the Ig domain of hCD8β (aa 1–116) and the stalk, transmembrane, and cytoplasmic domain of mouse CD8β (mCD8β) (aa 118–193; chimera A). The other chimeric construct contained the mCD8β Ig domain (aa 1–117) and the remainder from hCD8β (aa 117–201; chimera B). The constructs were generated using overlapping PCR. The first round of PCR involved a standard outside primer (T3 or KS) and an internal primer containing a sequence matching the end of the Ig domain of the template and the beginning of the stalk of the other CD8β. The internal primers used for chimera A were 5′ sense primer (CAG CTG AGT GTG GTT GAT TTC CTT CCT ACA ACT GCC CCA AC) and 3′ antisense primer (GTT GGG GCA GTT GTA GGA AGG AAA TCA ACC ACA CTC AGC). For chimera B, they were 5′ sense primer (GCT GAC TGT GGT TGA TGT CCT TCC CAC CAC TGC CCA G) and 3′ antisense primer (GGC TGG GCA GTG GTG GGA AGG ACA TCA ACC ACA GTC AG). The second round of PCR linked the appropriate Ig domain and remaining regions together using the same external primers in a reaction containing PCR products from the first round as templates. Both constructs were sequenced and subcloned into the expression vector pcDL-SRα-296 (21).

COS-7 fibroblasts were transfected using a modification of a previously described method (22). Briefly, an expression vector containing human and mouse CD8 cDNA (CD8α wild type (WT) or mutants (MTs) and/or CD8β WT or chimeras) were mixed with 8 μl of lipofectamine per 100 μl of the serum-free medium, Optimem (both from Life Technologies, Grand Island, NY) for 30 min at room temperature. This lipofectamine mixture was added to a 35-mm dish of nearly confluent COS-7 cells. Transfection was stopped after 5 or 18 h by replacing the lipofection mix with 2 ml of fresh medium containing 10% FCS. Cells were fed again after 24 h, and after an additional 24 h the dishes were either analyzed for cell-surface expression of CD8 or used in the adhesion assay.

For the cell-surface expression of hCD8, COS-7 transfectants were stained with primary mAbs OKT8 (Coulter, Westbrook, ME), which recognizes an epitope on CD8α; 5F2, which recognizes an epitope on CD8β; and 2ST85H7 (Immunotech, Westbrook, ME), which recognizes a conformational epitope formed by CD8αβ. Expression of mCD8 was examined with PE-conjugated forms of clone 53.6.7 (PharMingen, San Diego, CA) to detect mCD8α and with PE-conjugated forms of clone CT-CD8b (Caltag, Burlingame, CA) to detect mCD8β.

Cells from the blood of hCD8β transgenic mice were stained with the PE-conjugated clones described above for detecting mCD8α and mCD8β. In addition, a FITC-conjugated form of RM 4-5 (PharMingen) was used to detect mCD4, and a biotinylated form of 5F2 (produced on request by Immunotech, Marseille, France) with streptavidin-PE (PharMingen) as the second step was used to detect hCD8β.

The assay used was a modification of the method previously described in detail (22). COS-7 cells expressing CD8 were tested for their ability to bind an MHC class I⁺ B cell line, UC (19). The HLA Ags expressed by the UC cells were A1, A2, B5, B57, Cw4, DR7, DQ2, and DQ3. These cells constitutively expressed the firefly luciferase gene under the control of the CD8α promoter (23), a property which was utilized to measure binding. Transfected COS-7 cells were washed once with PBS, and 10⁷ UC cells were added to each 35-mm dish. The cells were incubated for 1 h at 37°C and the UC cells were aspirated off. After several washes with PBS, the number of bound cells was determined by measuring the amount of luciferase activity in the cell extract.

COS-7 cells transfected with vector alone, CD8β, or both CD8α and CD8β were metabolically labeled with ³⁵S-labeled cysteine and methionine 24–48 h posttransfection, and then they were chased and extracted for immunoprecipitation. Cells were deprived of cysteine and methionine for 1 h in methionine- and cysteine-free RPMI 1640 (ICN, Irvine, CA) containing 3% dialyzed FCS, and then they were pulsed with 0.5 mCi ³⁵S-labeled cysteine and methionine (TRAN-³⁵S-LABEL; ICN) for 1 h at 37°C. The labeled cells were washed with PBS and then lysed in 150 mM NaCl, 10 mM Tris (pH 7.4) containing 1% Triton X-100, 0.5 mM PMSF, 0.2 trypsin inhibitor units/ml aprotinin, 0.1 mM tosyl lysyl chloromethyl ketone, 1 μg/ml pepstatin A, and 5 mM iodoacetamide (pH 8) for 30 min at 4°C. Nuclei were removed by centrifugation for 15 min at 13,000 rpm. Supernatants were then precleared four to six times (30 min, overnight) with protein G zysorbin (Zymed, San Francisco, CA) and then precipitated with 5F2 (CD8β) for 1 h and with protein G-Sepharose (Zymed) for an additional hour. Pellets were washed three times with lysis solution and run on SDS-PAGE under nonreducing and reducing conditions. Gels were then dried and exposed to Kodak (Rochester, NY) Biomax MR film.

Transfection of an expression vector containing human CD8β cDNA alone was sufficient for cell-surface expression of CD8β on COS-7 cells as detected by staining with the hCD8β-specific Ab 5F2 (Fig. 1). These cells did not stain positively with the anti-CD8α Ab OKT8 (Fig. 1) or the anti-CD8αβ Ab 2ST85H7 (data not shown), indicating the absence of CD8α and CD8αβ on the cell surface. The level of expression of CD8β was similar to that of CD8αβ heterodimer, although twice as much CD8β DNA was transfected in the case of CD8β alone compared with the results from using CD8αβ (each plate was transfected with a total of 2 μg of DNA). Unlike hCD8β, expression of mCD8β was only detected when cotransfected with mCD8α, indicating apparent species differences in the expression of CD8β.

The ability of hCD8β to be expressed in vivo in the absence of CD8α was also seen in transgenic mice with a 95-kb human genomic fragment containing the entire hCD8β gene. Using a biotinylated form of 5F2, we were able to detect hCD8β⁺, mCD8α⁻, mCD4⁺ cells in peripheral blood (Fig. 2), lymph node, and thymus (data not shown). These data indicate that the hCD8β gene can be expressed in lymphoid cells in the absence of CD8α. Expression of the endogenous mCD8β was not detected in these CD4-positive cells.

To determine whether hCD8β could bind MHC class I-positive cells (UC-luciferase) in the absence of CD8α, we performed a binding assay. Fig. 3 demonstrates that no significant binding of UC-luciferase cells was detected to COS-7 cells expressing only hCD8β. In contrast, binding was detected to COS-7 cells expressing CD8αβ. Expression of hCD8β and hCD8αβ was as seen in Fig. 1.

Because CD8β is normally expressed as a dimer, we wanted to examine the form of hCD8β expressed on the COS-7 cells in the absence of CD8α. To do this, we metabolically labeled the cells with [³⁵S]l-cysteine and [³⁵S]l-methionine (TRAN-³⁵S-LABEL) and performed an immunoprecipitation with an Ab against CD8β, 5F2. The proteins were run on SDS-PAGE under reducing and nonreducing conditions, and the gel was dried and exposed to x-ray film. Fig. 4 demonstrates that when CD8β is precipitated from COS-7 cells expressing CD8β or CD8αβ under reducing conditions, there are three bands at ∼60, 50, and 25 kDa. The bands at ∼60 and 50 kDa are likely to be homodimers because when the samples are run under reducing conditions these bands disappear and new bands appear (presumably monomers) at ∼30 and 25 kDa. The 50-kDa protein is probably a nonglycosylated form of CD8β because the glycosylated form of CD8αβ is ∼64 kDa (24). Interestingly, CD8β monomers of the nonglycosylated form are also present in the nonreduced lane. However, this is not unique to CD8β because these monomers can also be seen in the samples precipitated from CD8αβ transfectants. Similar results were obtained for CD8αβ transfectants when immunoprecipitation was conducted with an Ab against CD8α (results not shown).

Chimeric CD8β constructs were generated with human and mouse CD8β to determine which region of hCD8β was responsible for its ability to form CD8ββ homodimers. Using flow cytometry we were able to demonstrate that chimera A containing the Ig domain of hCD8β and the stalk, transmembrane, and cytoplasmic domain of mCD8β was sufficient to allow homodimers to form. Chimera B (containing the mCD8β Ig domain with the stalk, transmembrane domain, and cytoplasmic domain of hCD8β) was not expressed unless cotransfected with mouse or human CD8α (Fig. 5). This indicates that the Ig domain of CD8β is the critical factor in determining whether the protein can be expressed as a homodimer in the absence of CD8α.

To determine which of the three conserved cysteines in the Ig domain of CD8α form the disulfide bridge critical for the proper folding of the protein, we conducted site-directed mutagenesis of all three cysteines (Cys²² on the B-strand, Cys³³ on the C-strand, and Cys⁹⁴ on the F-strand). Fig. 6,A shows that only the B- and F-strand MTs had an effect on the ability of CD8α to be expressed efficiently. These mutations dramatically reduced the level of expression of CD8α on the cell surface. Results shown are for staining with OKT8; however, a panel of hCD8α Abs were examined (G10.1, Leu-2a, and 66.2) and all were found to stain in a manner similar to that shown. However, in the presence of hCD8β, the percentage of cells expressing MT CD8α increased by 2-fold (Fig. 6 B). This effect was observed with both the B- and F-strand CD8α MTs.

From our studies with mouse and human CD8, we have discovered that contrary to mCD8β, hCD8β can be expressed on the cell surface of both COS-7 cells and murine lymphocytes in the absence of CD8α. This is consistent with previous studies with mCD8β, which show that cell lines transfected with mCD8β contain mRNA but not cell-surface expression in the absence of CD8α (14, 16). However, our results are discordant with previous studies reporting that hCD8β also required CD8α to be expressed. A probable explanation for this apparent contradiction is that in previous studies the Ab 2ST85H7 was used to detect expression of hCD8β (15, 24). It is now known that this Ab will only detect expression of the CD8αβ heterodimer, whereas 5F2, the Ab we used that was developed later, detects CD8β alone (25). It is unlikely that failure to detect mCD8β in the absence of mCD8α is due to the Abs only recognizing the heterodimer because the anti-mCD8β Ab 53.3.8 was shown to recognize an epitope retained on the mCD8β subunit after dissociation of the heterodimeric mCD8αβ complex (14).

In addition to expression of CD8β on COS-7 cells, in vivo evidence of expression of hCD8β in the absence of CD8α comes from mice expressing a hCD8β transgene. Despite otherwise tissue-specific expression of the transgene, we could detect hCD8β expression on a subpopulation of CD4⁺ lymphocytes, albeit at a lower level than that found on mCD8α cells. Previous experiments to ascertain whether mCD4 cells expressed the hCD8β transgene failed to detect this low-level expression. The reason for the discrepancy appears to be the staining reagents. Previously, we used a streptavidin-Red 670 reagent that stained the cells with a lower intensity than the streptavidin-PE used in this experiment (26).

Because the present transgenic mouse data indicate that hCD8β can be expressed on murine lymphoid cells in the absence of CD8α, what remains to be determined is whether such expression exists on human lymphoid cells. Staining of human blood did not detect any cells that were CD8α⁻CD8β⁺ (results not shown). However, this does not rule out the possibility that CD8ββ homodimers are expressed on the same cells as CD8αα homodimers and CD8αβ heterodimers. It will be important to determine whether this form exists on thymocytes or on mature T cells, either resting or activated.

Using CD8β chimeras, we found that the Ig domain was critical for the ability of hCD8β, to form homodimers because only chimeras with hCD8β Ig domain were expressed without CD8α. Comparison of the Ig domains of human and mouse CD8β indicate some potential differences that could affect the ability of these domains to pair properly. The C′- and G-strand bulges and adjacent loops, regions that are important in dimerization, contain charge differences between the two species. CD8β contains the conserved sequence in the G-strand bulge (F-G-X-G), but in the mouse form the X is threonine and in the human form X is lysine. In addition, comparison of the C-C′ loop between the two species also shows that there are a number of charge differences with the mCD8β containing four charged residues (both positive and negative) that are conserved in the rat but not in the human. Mutational analysis is required to determine whether these charge differences are responsible for the disparity reported here between mouse and human CD8.

The finding that murine Ig domains do not appear to be able to associate properly calls into question the physiological relevance of using chimeric molecules to study the function of mCD8β. Chimeric proteins have been created that contain mCD8α transmembrane and cytoplasmic domains with the extracellular domain of mCD8β (27). These chimeras are reported to be expressed at low levels on T cell hybridomas, and presumably they homodimerize due to the transmembrane domain of CD8α that has been shown to promote dimer formation (28). Results from the study by Wheeler et al. (27) suggest that chimeric CD8β can mediate interaction with MHC class I independent of CD8α. Our results with the hCD8ββ homodimers do not support this observation. Despite the relatively high levels of hCD8β expressed on COS-7 cells, we were unable to detect binding to MHC class I-positive cells. It is likely that the CD8ββ homodimers in the mouse chimera used by Wheeler et al. do not associate properly because the Ig domains of mCD8β do not normally form dimers. This could explain their finding that chimeric CD8ββ homodimers bound better to a form of MHC class I containing a mutation in the α3 domain of MHC class I (27) that was previously shown to be a critical residue for CD8-MHC class I interactions in both binding and CTL assays (29).

Consistent with the importance of the Ig domains of CD8 for dimer formation is our finding that mutant cysteine residues in the B- and F-strands of the CD8α Ig domains resulted in a decrease in CD8 expression, whereas mutation of both cysteines in the stalk had no significant effect (our unpublished observations). Absence of both cysteines in the stalk of CD8α did result in a small decrease in the expression of CD8αβ heterodimers (our unpublished observations), indicating that these residues may play more of a role in the formation of heterodimers. The B- and F-strand mutations most likely affected the ability of the protein to properly fold, hindering their ability to form dimers and hence their ability to be expressed on the cell surface. However, we were able to rescue some expression of MT CD8α as a heterodimer with CD8β WT because CD8β was properly folded and could bring more CD8α to the cell surface.

An interesting point to consider in light of the data we have presented here on the expression and function of CD8β is how the CD8 molecule may have evolved. The two genes for CD8α and β are 36 (in mice) or 56 (in humans) kb apart and probably originated from a common ancestor. Because CD8α can bind to MHC class I in the absence of CD8β and is able to associate with the tyrosine kinase p56^lck, one could assume that this molecule was the first gene. Further support for this hypothesis comes from the expression patterns of the two proteins. CD8αα homodimers, but not CD8αβ heterodimers, are expressed on human and rat NK cells as well as on γδ T cells, all of which are involved in the innate immune response. Once the gene duplicated, CD8β may have evolved for specific immune responses to enhance the ability of CD8 to act as a coreceptor with the TCR because CD8αβ does function as a better coreceptor (30, 31). Why mCD8β has evolved so that it is not expressed in the absence of CD8α, whereas hCD8β has not, remains to be determined.

We thank Dr. Peter Cresswell for helpful discussions.