Novel CD8+ T Cell Antagonists Based on β2-Microglobulin*

The CD8 coreceptor of cytotoxic T lymphocytes binds to a conserved region of major histocompatibility complex class I molecules during recognition of peptide-major histocompatibility complex (MHC) class I antigens on the surface of target cells. This event is central to the activation of cytotoxic T lymphocyte (CTL) effector functions. The contribution of the MHC complex class I light chain, β2-microglobulin, to CD8αα binding is relatively small and is mediated mainly through the lysine residue at position 58. Despite this, using molecular modeling, we predict that its mutation should have a dramatic effect on CD8αα binding. The predictions are confirmed using surface plasmon resonance binding studies and human CTL activation assays. Surprisingly, the charge-reversing mutation, Lys58 → Glu, enhances β2m-MHC class I heavy chain interactions. This mutation also significantly reduces CD8αα binding and is a potent antagonist of CTL activation. These results suggest a novel approach to CTL-specific therapeutic immunosuppression.

The CD8 coreceptor of cytotoxic T lymphocytes binds to a conserved region of major histocompatibility complex class I molecules during recognition of peptidemajor histocompatibility complex (MHC) class I antigens on the surface of target cells. This event is central to the activation of cytotoxic T lymphocyte (CTL) effector functions. The contribution of the MHC complex class I light chain, ␤ 2 -microglobulin, to CD8␣␣ binding is relatively small and is mediated mainly through the lysine residue at position 58. Despite this, using molecular modeling, we predict that its mutation should have a dramatic effect on CD8␣␣ binding. The predictions are confirmed using surface plasmon resonance binding studies and human CTL activation assays. Surprisingly, the charge-reversing mutation, Lys 58 3 Glu, enhances ␤ 2 m-MHC class I heavy chain interactions. This mutation also significantly reduces CD8␣␣ binding and is a potent antagonist of CTL activation. These results suggest a novel approach to CTL-specific therapeutic immunosuppression.
The peptide-MHC 1 class I complex (pMHC) on a target cell (antigen presenting cell) is recognized by a specific T cell receptor on the surface of CD8 ϩ cytotoxic T lymphocytes (CTL). The pMHC consists of a heavy chain, which is attached to the cell membrane and contains the peptide binding site, and a light chain, ␤ 2 -microglobulin (␤ 2 m). The CD8 molecule is a cell-surface glycoprotein present on CTL, which acts as a "coreceptor"; it is not peptide-specific, but binds to a conserved site on the pMHC molecule, which comprises several regions on the heavy chain and the small DE loop of ␤ 2 m consisting of residues 58 -60 (Lys 58 -Asp 59 -Trp 60 ) (1).
After CTL engage pMHC, the earliest intracellular events induce specific phosphorylation of tyrosine residues in the immunoreceptor tyrosine activation motifs within the cytoplasmic tails of the TCR-associated CD3 complex. The cytoplasmic tail of the CD8 ␣-chain is associated with the protein tyrosine kinase p56 lck . Active p56 lck initiates TCR signal transduction by phosphorylating the immunoreceptor tyrosine activation motifs within the CD3 complex. Inhibition of CD8 binding to pMHC therefore inhibits T cell activation (2). Exogenous soluble ␤ 2 m can exchange with cell surface-associated ␤ 2 m complexed to pMHC (3). Therefore, by mutating the CD8 contact site on ␤ 2 m, and exchanging the mutant ␤ 2 m into the native MHC, it should be possible to inhibit CTL activation.

Molecular Dynamics and Free Energy
Perturbations-Initial coordinates were taken from the crystal structure of the complex between human MHC class I HLA-A2 and the T cell coreceptor CD8␣␣ solved at 2.65-Å resolution and deposited in the Protein Data Bank (4) under the name 1akj (1). Molecular dynamics and free energy perturbations were performed using CHARMM (version 27) (5) and the standard all-atom parameter set (6). Hydrogens were added using the HBUILD module in CHARMM. Water molecules were added to the complex by superimposing a 16-Å sphere of TIP3P water molecules centered at the ␤ 2 m Lys 58 N atom.
The solvent atoms were minimized by 500 steps of steepest descents followed by 1000 steps of conjugate gradient. At the next step, the entire system was relaxed with 500 steps of steepest descents that were switched to conjugate gradient until convergence criteria of r.m.s. gradient of the potential energy lower than 0.3 Kcal/mol⅐Å has been achieved. A 14-Å nonbonded cutoff was employed. The dielectric constant was unity. The system was simulated using a stochastic boundary molecular dynamics (7). The reference point for partitioning the system was the ␤ 2 m Lys 58 N atom. The system was divided into a 12-Å reaction region, a 4-Å buffer region, and a reservoir. The frictional coefficients for water oxygen and heavy atoms in the protein were 62 and 200 ps Ϫ1 , respectively (8). The relaxed system was equilibrated at 300 K for 150 ps with a time step of 1 fs followed by 1 ns performed for data collection with coordinates and energies saved to a disc every 1 ps. The ␤ 2 m Lys 58 was mutated using the Biopolymer and Homology modules in the MSI software package. For each mutation the procedure described above has been repeated. Relative binding Helmholtz free energies were calculated by the perturbation method (9) as follows: ⌬G 3 ϭ HLA-A2/␤ 2 m (native) complex 3 HLA-A2/ ␤ 2 m (Lys3 Glu) complex and ⌬G 4 ϭ HLA-A2/␤ 2 m (native) /CD8␣␣ complex 3 HLA-A2/␤ 2 m (Lys3 Glu) /CD8␣␣ complex.
Since free energy is a state function, it is path-independent, and the free energy difference: ⌬G 4 -⌬G 3 is equal to the difference ⌬G 2 -⌬G 1 (see "Results"). Each perturbation was performed in two steps using a total number of 26 windows. At the first 16 windows the lysine H␦1, H⑀1, H⑀2, N, H1, H2, and H3 atoms (including their charges) were deleted. C␦ atom type was modified to sp 2 carbonyl carbon. H␦2 and C⑀ atom types were modified to carboxylate oxygens. At the last 10 windows, the charges of the remaining side chain were adjusted to asp side chain. Trajectories were produced by MD simulations at the same conditions to these described above with 150 ps of equilibration and 100 ps of data collection at every window.
Soluble TCR Preparation-The TCR used for the SPR experiments derives from the JM22 T cell clone (10,11). It is specific for an HLA-A2-restricted peptide (GILGFVFTL) from the influenza matrix protein (58 -66) and uses gene segments TCRAV10S2J9S11C1 and TCRBV17S1J2S7C2. The two fragments of soluble JM22-TCR␣␤ (residues 1-204 for the ␣-chain and 1-245 for the ␤-chain) were expressed in E. coli, refolded and purified as described previously (12). The TCR concentration was determined from the extinction coefficient (105,500 M Ϫ1 x cm Ϫ1 , determined by amino acid analysis), assuming 100% activity.
Soluble HLA-A2/␤ 2 m Complex Preparation-Soluble influenza peptide-HLA-A2/␤ 2 m complexes were prepared by refolding HLA-A2 heavy chain carrying the biotin tag with ␤ 2 m wildtype or mutant (both expressed in E. coli) and the synthetic peptide corresponding to influenza matrix protein 58 -66 GILGFVFTL (Genosys, Woodlands, TX) as described in Garboczi et al. (13). The refolded complexes were purified by both anion exchange and gel filtration before being used in SPR experiments. HLA-A2 heavy chain was enzymatically biotinylated as described (14) using N-hydroxysuccinimidobiotin (Sigma) and BirA enzyme. Tetrameric pMHC I complexes were produced by conjugation with phycoerythrin-labeled extravidin as described previously (14). DNA constructs encoding the ␤ 2 m mutants (Lys 58 to Arg, Asp, Glu, Ser, Val, Tyr, Cys, SES, Trp, and GRG) were produced using the Quick-Change Site-Directed Mutagenesis Kit (Stratagene) and checked by DNA sequencing of the entire coding fragment. The mutant proteins were expressed and the inclusion bodies purified as for the wild type.
Surface Plasmon Resonance Experiments-SPR binding studies were performed at 25°C using a BIAcore TM 3000 (BIAcore AB, St. Albans, UK) in HBS (BIAcore AB). HBS contains 10 mM HEPES, pH 7.4, 150 mM NaCl, 3.4 mM EDTA, and 0.005% Surfactant P20. Peptide-HLA class I complex was bound to the BIAcore chip by producing recombinant soluble pMHC fused to a biotinylation tag, which was specifically biotinylated in vitro (14) and flowed over a streptavidin-coated chip surface. Streptavidin (Sigma) was covalently coupled to Research Grade CM5 sensor chip (BIAcore) via primary amines using the Amine Coupling kit (BIAcore). For coupling, the streptavidin was dissolved in 10 mM sodium acetate, pH 5.5, and injected at 0.2 mg ml Ϫ1 . Immobilization level was 6000 response units on average. Biotinylated HLA-A2/␤ 2 m (wild type and mutants) complexes were immobilized at 12,000 -12,500 Response units by injection of 5-35 l at 40-100 g ml Ϫ1 , at a flow rate of 5 l min Ϫ1 . The injections of the different CD8␣␣ (sCD8) and JM22-TCR␣␤ (sTCR) solutions were performed at a flow rate of 5 l min Ϫ1 . K d values were obtained either by Scatchard plots or by nonlinear fitting of the Langmuir binding isotherm (where B is sCD8 (or sTCR) concentration and AB max is maximum sCD8 (or sTCR) binding) to the data using the Levenberg-Marquardt algorithm as implemented in the Windows 98 application Origin (version 6.1; Microcal Software, Northampton, MA).
Cell Culture and CTL Activation Assays-pBMC were isolated from fresh blood by Ficoll-Hypaque density gradient centrifugation. CD8 ϩ CTL clones were generated and maintained as described previously (15). Target cells in cytotoxicity assays were HLA-matched immortalized T cell hybridomas or Epstein-Barr virus transformed B-lymphoblastoid cell lines (B-LCL) incubated with peptide as shown, or infected with recombinant vaccinia virus (rVV), and labeled with 51 Cr (Amersham Biosciences, Amersham, UK). Peptides were synthesized using standard Fmoc (N-(9-fluorenyl)methoxycarbonyl) chemistry and were Ͼ90% pure as determined by high performance liquid chromatography (Research Genetics, Huntsville, AL). Vaccinia infections were effected at 3-5 plaque-forming units/cell for 1 h and followed by a 6-h incubation period to allow expression prior to 51 Cr labeling. The rVV expressing HIV-1 Nef was constructed from a full-length proviral clone isolated from donor SC1 according to standard protocols; wild type WR rVV was used to infect control target cells (16). Lysis assays were performed in low percentage fetal calf or human serum using standard 51 Cr release methodology (17). Tetrameric pMHC I complex activation assays were performed by measurement of extracellular RANTES release after 30min exposure as described previously (18). All data points for both lysis and RANTES release assays represent the mean of triplicate readings; y axis error bars show the S.D. in each case.
Flow Cytometric Analysis-Cells were stained with phycoerythrinlabeled tetrameric pMHC I complexes containing wildtype or Lys 58 3 Glu forms of ␤ 2 m as described previously (19). Stained cells were analyzed using a Becton Dickinson Calibur flow cytometer with CellQuest software.  does not interact with the CD8␣-1. In the Lys 58 3 Arg mutant (side chain in purple) the H-bond to CD8␣-1 Asp 75 is preserved. b, in the Lys 58 3 Ser and Lys 58 3 Cys mutants the interaction with the CD8␣␣ is mediated through a water molecule: the side chain donates its hydroxyl or thiol hydrogen to a water molecule, which donates its hydrogen to CD8␣␣ Val 24 carbonyl oxygen. c, the native structure ␤ 2 m Lys 58 and CD8␣-1 Asp 75 (blue) compared with steric hindrance via a bulky side chain mutants: Lys 58 3 Tyr (side chain in green) and Lys 58 3 Trp (side chain in purple). The bulky side chain fills the cavity between the HLA-A2/␤ 2 m CD8␣␣, and the interaction with the CD8␣␣ is poor. d, comparison between the native ␤ 2 m DE loop consisting of residues 58 -60 (Lys 58 -Asp 59 -Trp 60 ), shown in blue, to insertions: Lys 58 3 SES (green) and Lys 58 3 GRG (purple). ␤ 2 m residues Leu 40 -Ala 79 are shown in azure. e, comparison between the native structure to a Lys 58 3 Val mutation (blue). The hydrophobic valine side chain repels CD8␣-1 Asp 75 carboxylate. f, mutations to a negatively charged side chain: Lys 58 3 Asp (yellow), Lys 58 3 Glu (purple) compared with the native lysine (gray). In both Lys 58 3 Asp and Lys 58 3 Glu mutations the positive charge on HLA-A2 Arg 6 heavy chain attracts the mutant's carboxylate. Only Lys 58 3 Glu carboxylate is in close proximity to CD8␣-1 Asp 75 O␦2. mutations to short polar side chains (Lys 58 3 Ser, Lys 58 3 C), mutations introducing steric hindrances via bulky side chains (Lys 58 3 Tyr, Lys 58 3 Trp) or insertions (Lys 58 3 SES, Lys 58 3 GRG), a mutation to a medium length hydrophobic side chain (Lys 58 3 Val), mutations to negatively charged side chains (Lys 58 3 Asp, Lys 58 3 Glu).
We studied the hydrogen bond network formed between the wild type Lys 58 , or mutations of this residue, to CD8␣-1 as shown in Table I. The perturbation to the conformation of the contact residues (CD8␣-1:Arg 4 , Asp 75 , Val 24 -Val 26 , ␤ 2 m:Trp 60 , Lys 58 ) (1) caused by various Lys 58 mutations is presented as the root mean square (r.m.s.) deviation of these residues to that observed in the crystal structure.
Neutralizing the Lys 58 positive charge in silico leads to the loss of two H-bonds to CD8␣-1 Asp 75 ⌷␦2 and Val 24 O atoms, as illustrated in Fig 1a and Table I, and formation of alternative hydrogen bonds with water molecules in the cavity of the HLA-A2/CD8␣␣/␤ 2 m complex. A mutation that preserves the positive charge (Lys 58 3 Arg) has a less dramatic effect: the H-bond to CD8␣-1 Asp 75 ⌷␦2 is preserved, and the r.m.s. value is similar to that of the wild type structure and smaller than that of the neutral Lys 58 side chain. Therefore, eliminating the positive charge on the side chain is likely to reduce the binding affinity to the CD8␣␣.
In short polar side chain mutants, Lys 58 3 Ser and Lys 58 3 Cys, the interaction with CD8␣␣ is mediated through a water molecule as shown in Fig. 1b and Table I. The side chain donates its hydroxyl or thiol hydrogen to a water molecule, which donates its hydrogen to CD8␣␣ Val 24 O. Unlike Lys 58 3 Ser, where the H-bond network is stable, the H-bond between the thiol moiety and the water molecule in the Lys 58 3 Cys mutant fluctuates during the simulation. This has an impact on the r.m.s. values shown in Table I, which increase from 0.70 for Lys 58 3 Ser to 0.93 for Lys 58 3 Cys. These results suggest that removing the H-donor group from the side chain at position 58 should reduce the binding affinity to CD8␣␣.
Mutations introducing steric hindrance via a bulky side chain, Lys 58 3 Tyr and Lys 58 3 Trp, are illustrated in Fig. 1c. The bulky side chain fills the cavity between HLA-A2/␤ 2 m and CD8␣␣, and the interaction with CD8␣␣ is impaired by the lack of H-bonding, resulting in high r.m.s. values (Table I). Lys 58 3 GRG and Lys58 3 SES insertions (Fig. 1d) perturb the tertiary structure and lead to increased r.m.s. values of 1.21 and 1.78, respectively. The higher r.m.s. value of the Lys 58 3 SES insertion is due to reversal of the positive charge and the fact that serine has a side chain that contributes to the overall steric hindrance.
The Lys 58 3 Val mutation yielded a higher r.m.s. value than all other single substitution mutations. In bulkier mutations, such as Lys 58 3 Trp, the side chain can orient itself toward the HLA-A2/␤ 2 m-CD8␣␣ cavity, whereas the valine side chain is not long enough to have this effect. As a result, the hydrophobic side chain repels the polar CD8␣-1 Asp 75 (Fig. 1e). Fig. 1f shows mutations to a negatively charged side chain (Lys 58 3 Asp, Lys 58 3 Glu). Repulsion between the carboxylates leads to an average distance of 6.27 Å between the CD8␣-1 Asp 75 O␦2 and Lys 58 3 Asp O␦1 oxygens. Strikingly, the distance between CD8␣-1 Asp 75 O␦2 and Glu 58 O⑀1 oxygens in the Lys 58 3 Glu is only 4.09 Å. Both Lys 58 3 Asp and Lys 58 3 Glu are attracted to the positive charge on HLA-A2 Arg 6 heavy chain: in 52.9% of the frames taken from the MD simulation, the distance between the mutated Glu 58 O⑀2 and HLA-A2 Arg 6 N1 atoms was lower than 5 Å, fluctuating to a minimum of 3.18 Å. This interaction stabilizes the HLA-A2/ ␤ 2 m complex and orients the Lys 58 3 Glu side chain toward CD8␣-1 Asp 75 . HLA-A2 Arg 6 heavy chain attracts the Lys 58 3 Asp carboxylate as well; however since Lys 58 3 Asp side chain is shorter, it is distant from Asp 75 O␦2.
The theoretical results show that, although the contribution of the ␤ 2 m Lys 58 to CD8␣␣ binding is relatively small (1), its mutation should have a marked effect upon CD8␣␣ binding.
To test these predictions, we used SPR to measure the binding of soluble CD8␣␣ (sCD8) and soluble JM22 T cell receptor (sTCR) to HLA-A2-influenza matrix peptide complex, containing wild type or mutant ␤ 2 m (Fig. 2). The affinity of sTCR for pMHC was similar for all of the complexes studied (Table II), indicating that the active material on the chip surface was correctly folded. However, the absolute measurements of response varied markedly between the different complexes, indicating varying proportions of active material on the chip surface. This implies that some mutant ␤ 2 m complexes are more stable than others. In particular, HLA complexes containing Lys 58 to Tyr, Trp, SES, and GRG mutations show reduced stability (data not shown). These observations correlate with refolding efficiency, with yields of Lys 58 3 Arg and Lys 58 3 Glu complexes being several times higher than those for Lys 58 3 SES and Lys 58 3 GRG complexes (data not shown). This difference agrees with the MD simulations, which predict  that the mutant Lys 58 3 Arg complex structure is closest to that of the wild type (Table I), and that the complex containing the ␤ 2 m Lys 58 3 Glu mutant is stabilized by the interaction with HLA-A2 heavy chain R 6 . There is a strong correlation between the measured sCD8 binding (Table II) (Table I). The wild type structure forms two hydrogen bonds and shows the highest affinity for sCD8. Complexes containing ␤ 2 m mutants Lys 58 3 Arg, Ser, or Cys are predicted to form only a single H-bond (mediated through a water molecule for Lys 58 3 Ser and Lys 58 3 Cys) and show decreased binding affinity for sCD8 (Fig. 2, Table II). In complexes containing other ␤ 2 m mutants (Lys 58 mutations to Val, Asp, Tyr, Trp, GRG, SES, and Glu), where no H-bond was predicted, the sCD8 binding was negligible (K d Ͼ 1 mM). Similarly, mutations of Asp 59 and Trp 60 also greatly reduced sCD8 binding (data not shown). The BIAcore binding data showed that sCD8 binding is undetectable in HLA complexes containing ␤ 2 m in which Lys 58 is mutated to Glu (Fig. 2, Table II), yet this complex behaves identically to complexes containing wildtype ␤ 2 m in terms of sTCR binding, refolding yield, and stability.  (2). Tetramer containing wild type ␤ 2 m folded around the cytomegalovirus-derived HLA-A2-restricted peptide NLVPMVATV was included as a control for nonspecific binding. b, tetramer-induced activation of 003 CTL. Tetramers folded around SLYNTVATL containing either wild type or Lys 58 3 Glu ␤ 2 m were incubated in 96-well plates with 10 4 CTL per well at the concentrations shown. Release of RANTES into the culture supernatant was measured after 30 min by enzyme-linked immunosorbent assay. The tetramer containing Lys 58 3 Glu ␤ 2 m induces less activation compared with the tetramer containing wildtype ␤ 2 m at low concentrations. The control tetramer containing wild type ␤ 2 m folded around the NLVPMVATV peptide failed to activate CTL at 10 Ϫ7 g/ml. c, antagonism of CTL activation by soluble Lys 58 3 Glu ␤ 2 m. Target T2 hybridomas were plated at 5 ϫ 10 3 per well in 96-well round-bottomed plates with soluble proteins at 60 g/ml as indicated and cognate peptide at 10 Ϫ6 M. After incubation for 20 min at room temperature, CTL were added at an effector:target (E:T) ratio of 2:1. Effector CTL (clone 4D5) were specific for the HLA-A2-restricted MAGE-3 tumor antigen epitope FLWGPRALV (17). The assay was harvested after 6 h. Similar results were obtained with the HLA-A2-restricted SLYNTVATL-specific CTL clone 5D8 (data not shown). d, inhibition of direct ex vivo fresh PBMC response to endogenously processed antigen. Target cells were autologous SC21 B-LCL infected with either Nef or control WR rVV as described and incubated to allow expression in the presence of 50 g/ml protein as indicated for 6 h. Proteins were present at this concentration throughout the 51 Cr labeling process and for the entire duration of the assay. The assay was harvested after 12 h. Vaccinial expression of HLA class I-restricted antigens obviates the problems associated with potential effects of ␤ 2 m or mutants thereof on peptide exchange at the cell surface when external loading protocols are followed. The inhibitory effect of the mutant Lys 58 3 Glu protein is controlled by the addition of equimolar concentrations of wild type ␤ 2 m. Similar results were seen in direct ex vivo assays with RPMTYKGAL-pulsed SC21 B-LCL (data not shown). The Lys 58 3 Glu mutant also inhibited a primary CTL line (data not shown) specific for the HLA B35-restricted parvovirus B19-derived epitope QPTRVDQKM (NS1, residues 391-399) (22).
The strong correlation between the predictions of the molecular dynamics simulations, free energy calculations, and biophysical measurements of sCD8 binding led us to conclude that the ␤ 2 m mutant containing a glutamate residue at position 58 (Lys 58 3 Glu) was the most promising candidate for further investigation.
The impact of mutating ␤ 2 m on the interaction between soluble antigen and TCR expressed on the cell surface was investigated using tetrameric pMHC I complexes. Fluorochrome-labeled tetramers containing either wild type or Lys 58 3 Glu forms of ␤ 2 m-stained CD8 ϩ CTL at equivalent levels across a range of concentrations (Fig. 3a). However, tetrameric complexes containing the Lys 58 3 Glu form of ␤ 2 m were substantially impaired in their ability to activate CTL. This effect was titratable and most apparent at lower concentrations (Fig. 3b). These data indicate that, at similar levels of interaction with cell surface TCR, the impaired ability of complexes containing the Lys 58 3 Glu form of ␤ 2 m to engage the CD8 coreceptor translates into a biologically significant effect. This is consistent with previous work demonstrating that the major role of the CD8 coreceptor is the provision of signaling moieties that allow an enhanced sensitivity to antigen (2).
The effects of soluble exogenous wild type and mutant ␤ 2 m proteins on CTL interactions with target cells expressing pMHC I antigenic complexes were assessed using standard 51 Cr release cytolytic assays. At concentrations of pMHC on the target cell surface insufficient to saturate the maximal lytic capacity of the CTL, exogenous wild type ␤ 2 m, and those mutants that efficiently refold with soluble MHC class I heavy chain to form stable complexes, consistently enhanced CTLmediated lysis and interferon-␥ production (data not shown). This effect likely relates to the ability of these proteins to stabilize pMHC class I on the target cell surface in a TCRcognate form and is consistent with previous work (3,20). At higher concentrations of pMHC class I on the target cell surface, exogenous mutant forms of ␤ 2 m that impair CD8 binding to pMHC class I were found to inhibit CTL-mediated lysis (Fig.  3c). In all experiments with CD8-dependent CTL, the Lys 58 3 Glu mutant was the most potent inhibitor of CTL activation. These results indicate that mutant forms of ␤ 2 m can, through inhibition of CD8 binding to complexed pMHC I, antagonize CTL activation.
The potential use of such reagents in a therapeutic setting requires an effect on whole blood antigen-specific responses mounted by polyclonal unmanipulated CTL. We examined this using peripheral blood mononuclear cells (PBMC) isolated from donor SC21, an HIV-1-infected individual with a potent CTL response to the viral Nef protein that maps predominantly to the HLA B7-restricted epitope RPMTYKGAL (residues 75-83).
The Lys 58 3 Glu mutant was found to inhibit this fresh PBMC lytic response to endogenously presented pMHC class I antigen in comparison to equimolar levels of wild type ␤ 2 m (Fig. 3d).
In conclusion, we have developed stable ␤ 2 m mutants that inhibit CD8 coreceptor binding to pMHC I and exert an inhibitory effect on CTL activation. These data suggest that such mutant forms of ␤ 2 m could be used to selectively modulate the CD8 ϩ cellular immune response, a principle that could be applied therapeutically (21).