In vitro mutagenesis at a single residue introduces B and T cell epitopes into a class I HLA molecule.

We have studied the interaction of HLA class I antigens with alloreactive cytotoxic T lymphocytes and monoclonal antibodies using site-directed mutagenesis and expression of an HLA-Aw68.1 gene. Two mutants containing distinct substitutions at polymorphic residues near the NH2-terminal end of the alpha 2 domain were made. One mutant with substitutions at positions 95 and 97 corresponding to residues found in HLA-A2.1 showed no alterations in binding of HLA-Aw68- or HLA-A2-specific monoclonal antibodies, but was reactive with some HLA-A2-specific CTL clones. A second mutant, in which glycine at position 107 was replaced with tryptophan found at that position in HLA-A2.1, was recognized by HLA-A2-specific CTL clones and HLA-A2, Aw69-specific monoclonal antibodies. Thus, substitution of a single amino acid residue at position 107 of the HLA-Aw68.1 molecule generates an allospecific determinant shared with HLA-A2.1 and recognized by both B and T lymphocytes.

The biosynthesis and secretion of prostate-specific antigen in LNCaP cells.

Prostate-specific antigen (PSA) has been demonstrated to release the active form of insulin-like growth factor I in vitro (P. Cohen et al., J. Clin. Endocrinol. & Metab., 75: 1046-1053, 1992; P. Cohen et al., J. Clin. Endocrinol. & Metab., 79: 1410-1415, 1994; P. Cohen et al., Horm. Metab. Res., 26: 81-84, 1994) and has significant mitogenic activity on osteoblast cells, fibroblasts, and other cultured cells (C. S. Killian et al., Biochem. Biophys. Res. Commun., 192: 940-947, 1993). Recently, PSA has been found not only in prostate tissues but also in breast, colon, ovarian, and other tissues (E. P. Diamandis and H. Yu, J. Clin. Endocrinol. & Metab., 80: 1515-1517, 1995; E. P. Diamandis and H. Yu, Clin. Chem., 41: 204-210, 1995; A. Clements and A. Mukhtar, J. Clin. Endocrinol. & Metab., 78: 1536-1539, 1994). Therefore, PSA has been proposed as a candidate growth factor, cytokine, or growth factor regulator. In this setting, knowing how to manipulate or block the secretion of PSA by the prostate cancer cells could be a useful approach to controlling the progression of human prostate cancers. Using metabolic labeling experiments, we have studied the biosynthesis and secretion of PSA in LNCaP cells. We have also examined the effects of DTT, tunicamycin, 1-deoxymannojirimycin, pilocarpine, and testosterone on PSA biosynthesis and secretion. The results indicate that the secretion of PSA in LNCaP cells is constitutive instead of regulated and that the disruption of intramolecular disulfide bonds affects the transport of PSA from the endoplasmic reticulum to the Golgi apparatus. The biosynthesis of PSA is potentiated by testosterone and inhibited by brefeldin A and DTT. These results will help us understand PSA biosynthesis and secretion in human prostate cancers.

Immunotherapy of NOD mice with bone marrow-derived dendritic cells.

We evaluated two bone marrow-derived dendritic cell (DC) populations from NOD mice, the murine model for type 1 human diabetes. DCs derived from GM-CSF [granulocyte/macrophage colony-stimulating factor] + interleukin (IL)-4 cultures expressed high levels of major histocompatibility complex (MHC) class II, CD40, CD80, and CD86 molecules and were efficient stimulators of naive allogeneic T-cells. In contrast, DCs derived from GM-CSF cultures had low levels of MHC class II costimulation/activation molecules, were able to take up mannosylated bovine serum albumin more efficiently than GM + IL-4 DCs, and were poor T-cell stimulators. The two DC populations migrated to the spleen and pancreas after intravenous injection. To determine the ability of the two DC populations to modulate diabetes development, DCs were pulsed with a mixture of three islet antigen-derived peptides or with medium before injection into prediabetic NOD mice. Despite phenotypic and functional differences in vitro, both populations prevented in vivo diabetes development. Pulsing of the DCs with peptide in vitro did not significantly improve the ability of DCs to prevent disease, which suggests that DCs may process and present antigen to T-cells in vivo. In addition, we detected GAD65 peptide-specific IgG1 antibody responses in DC-treated mice. Overall, these results suggest that a Th2 response was generated in DC-treated mice. This response was optimal when using GM + IL-4 DCs, which suggests that the balance between regulatory Th2 and effector Th1 cells may have been altered in these mice.

Inhibition of horseradish peroxidase activity by specific antibody: determinant specificity of anticatalytic antibodies.

Rabbit antisera specific for horseradish peroxidase inhibit the catalytic activity of the enzyme. All antibodies prepared against the holoenzyme react with the peroxidase apoenzyme. However, only a minority (30-45%) of the total antiperoxidase pool cross react with reduced and alkylated apoenzymes. The antibodies inhibiting peroxidase activity do not bind to S-carboxymethyl of S-carboxamidomethylated apoenzyme derivatives as measured by absorption and competition of inhibition experiments. Glycopeptides derived from horseradish peroxidase also failed to bind anticatalytic antibodies. Antibodies that inhibit enzyme activity have specificity for noncarbohydrate conformation dependent antigenic determinants of horseradish peroxidase. Additional experiments probed the mechanism by which inhibitory antibody decreases the catalytic activity of horseradish peroxidase. Absorption spectra of horseradish peroxidase that has bound Fab fragments sufficient to cause 90% inhibition of the enzyme activity determined that the enzyme retained the ability to bind hydrogen peroxide. Thus, anticatalytic antibodies do not prevent the formation of the first enzyme-substrate intermediate but mediate their inhibitory effects by disrupting a later step in the reaction mechanism.

Beta 2-microglobulin and calnexin can independently promote folding and disulfide bond formation in class I histocompatibility proteins.

Class I histocompatibility proteins fold and assemble with beta 2-microglobulin (beta 2m) into heterodimers before binding short peptides in the endoplasmic reticulum. Here, we show that class I proteins rapidly form disulfide bonds, and that the process is highly reversible in Daudi cells lacking beta 2m. Three distinct class I protein conformations are present in equal amounts in these cells, each associated with the molecular chaperone calnexin. When binding of calnexin is inhibited by the glucosidase inhibitor castanospermine, fully oxidized class I proteins are no longer detected, suggesting that calnexin is required for completion of folding. However, in Daudi cells transfected to express beta 2m, castanospermine decreases only slightly the levels of fully oxidized class I proteins, indicating that folding is much less dependent on calnexin in the presence of beta 2m. Furthermore, calreticulin, a chaperone with functional similarities to calnexin, associates with class I molecules in beta 2m-positive cells. but not in Daudi cells, consistent with completion of folding and disulfide bond formation of class I heavy chains before binding to calreticulin occurs. This study demonstrates that calnexin and beta 2m can function independently to promote folding of class I heavy chains prior to formation of stable class I dimers.

Heterogeneity in anticatalytic activity of antibody specific for horseradish peroxidase.

Rabbit antibodies specific for horseradish peroxidase are heterogeneous in their ability to inhibit enzyme activity. Heterogeneity was demonstrated by fractionation of the total antiperoxidase pool by differential ammonium sulfate precipitation and differential elution of antibody from enzyme affinity columns. Both fractionation methods yielded antibody subpopulations that differed in anticatalytic activity. Some antibody subpopulations decreased enzyme activity almost completely at low molar ratios of antibody to peroxidase. Other subpopulations were not effective inhibitors even at great molar excess. Admixture experiments demonstrated that inefficient antibody pools decreased the anticatalytic effect of highly inhibitory antibody. The degree of inhibition observed with unfractionated antiserum is a reflection of the interaction of various antibody subpopulations with the enzyme. No correlation was found between the immunoglobulin class of antiperoxidase and anticatalytic efficiency in analyses of numerous antisera. The determinant specificity of an antiperoxidase molecule determines its anticatalytic ability. A peptide fragment (mol. wt 22,500) of peroxidase prepared by partial tryptic digestion bount 60 to 70% of the total antiperoxidase in a number of antisera. However, the peptide did not bind inhibitory antibodies.

CD8 independence and specificity of cytotoxic T lymphocytes restricted by HLA-Aw68.1.

The crystal structure of the HLA-Aw68.1 antigen binding site revealed a negatively charged pocket centred on aspartic acid 74 (Garrett et al. 1989). Access to this '74 pocket' is blocked in HLA-Aw68.2 and HLA-Aw69 by two substitutions at positions 97 and 116. This key feature suggests that the Aw68.1-peptide-specific interactions may involve salt bridges between oppositely charged residues. In this paper, the influenza epitope recognized by virus-specific HLA-Aw68.1-restricted cytotoxic T lymphocytes (CTL) has been defined in vitro with a synthetic peptide corresponding to residues 89-101 of the nucleoprotein (NP). Amino acid substitutions of the peptide NP 89-101 showed that the arginine at position 99 is an anchor point of the peptide within the Aw68.1 antigen binding site. Consistent with this we find that neither HLA-Aw68.2 nor HLA-Aw69 positive cells can present peptide NP 89-101 to Aw68.1-restricted CTL. Our results therefore suggest a model in which presentation of NP 89-101 by HLA-Aw68.1 is dependent upon interaction of the positively charged arginine residue at position 99 of the peptide, with the negatively charged aspartic acid in the '74 pocket' of HLA-Aw68.1. We also show that influenza-virus-specific HLA-Aw68.1-restricted CTL are CD8 independent. This result is consistent with the low affinity of HLA-Aw68.1 for CD8 (Salter et al. 1989) and reveals a unique example of CD8-independent priming of CTL by natural infection with a common pathogen in humans.

Mutant HLA-A201 heavy chains with lowered affinity for beta 2m are transported after growth at reduced temperatures.

Forty-five site-directed mutants bearing single amino acid substitutions in the alpha 3 domain of the class I molecule HLA-A201 were previously transfected into CIR cells and screened for surface expression by antibody binding. Eight mutants are expressed at significantly reduced levels relative to HLA-A201, including two with substitutions at residues contacting beta 2m. One of the latter mutations, position 242 gln > lys (242K), is now shown to block assembly with beta 2m and prevent intracellular transport at 37 degrees C. At temperatures ranging from 21 degrees C to 30 degrees C, 242K heavy chains and beta 2m form dimers that are exported to the cell surface. Surface expression at 26 degrees C is not blocked by cycloheximide pretreatment, but is completely inhibited by Brefeldin A, suggesting that at 37 degrees C preformed heavy chains accumulate in the ER. Glycans on the retained heavy chains are sensitive to digestion by Endo H, but become Endo H resistant after cells are grown at 26 degrees C. Preincubation of 242K cells with synthetic peptides shown previously to bind HLA-A201 does not increase reactivity with anti-HLA-A2 antibodies, suggesting that the defective phenotype is not due to instability of cell surface mutant class I dimers, but derives instead from impaired assembly of 242K heavy chains with beta 2m inside the cell. This contrasts with mutant cells such as .174, T2 and RMA-S, which exhibit defects in internal peptide transporters, but assemble and export "empty" dimers to their surfaces that can be stabilized subsequently by exogenous peptides. 242K mutants may therefore be suited uniquely for studying assembly and peptide binding to class I molecules in the ER.

Expression of differentiation antigens by hybrids of human lymphoblastoid cells.

In previous communications, we have described the expression of class I and class II histocompatibility antigens by hybrids of human B and T lymphoblastoid cell lines (B- and T-LCL). In all cases, such hybrids were found to resemble their B-LCL parents, expressing high levels of class I and class II antigens encoded by both parent cell lines. In the current study, we have conducted a more extensive analysis of B-LCLxT-LCL hybrids with a panel of monoclonal antibodies recognizing a variety of B and T lymphocyte differentiation markers. Rather than exhibiting a B-LCL-dominant phenotype, most hybrids were found to express a majority of both T and B lymphocyte antigens expressed by their parent cell lines. Several hybrids of pairs of dissimilar T-LCL were also produced and analyzed. Again, a majority of parental antigens was expressed on the hybrids. However, eight of eight hybrids of the T-LCL CEM and HSB failed to express HNK-1, an antigen strongly expressed by HSB; and two hybrids of the T-LCL CEM and SKW3 expressed CD3, an antigen expressed by neither parent cell line.

Peptide-conformed beta2m-free class I heavy chains are intermediates in generation of soluble HLA by the membrane-bound metalloproteinase.

Molecular mechanisms of soluble HLA-release by a membrane-bound metalloproteinase (MPase) are not defined. We have investigated the possibility that certain beta2-microglobulin (beta2m)-free heavy chains (HC) retain peptide-induced conformations before and after the cleavage by using mutant HLA-A2.242K HC with reduced affinity for beta2m. We show that dissociation of HC/beta2m complexes on the surface of C1R lymphoblastoid cells generates both conformed and non-conformed beta2m-free HC recognized by conformation-dependent antibodies. Conformed HC, having bound the HLA-A2-specific peptide HTLV-1 tax 11-19, can retain their proper conformations after dissociation of beta2m. Further, conformed and non-conformed surface beta2m-free HC are cleaved by the MPase, and some released HC preserve their conformations. Exogenous beta2m binds only to conformed HC, and protects them from cleavage as effectively as the MPase inhibitor BB-2116. We propose that soluble HLA-release requires generation of peptide-conformed beta2m-free HC intermediates on the cell surface, which are then cleaved by the MPase and in solution may reassociate with beta2m. Given the role of soluble HLA in the indirect allorecognition, the activity of this MPase may be important in transplant rejection.

Flow-cytometric determination of peptide-class I complex formation. Identification of p53 peptides that bind to HLA-A2.

A novel class I-peptide-binding assay was developed and used to identify a series of peptides derived from the human p53 tumor-suppressor gene product capable of binding the HLA-A2 class I allele. Brief pH 3.3 acid treatment of human cell lines rapidly denatures pre-existing class I complexes, as detected by loss of binding of conformation-dependent mAbs, leaving only free class I heavy chains associated with the viable cell surface. These heavy chains may be induced to refold and be recognized by antibodies (in 2-4 hours) when acid-treated cells are coincubated with exogenous beta 2-microglobulin and peptides capable of binding the relevant class I allele examined. This assay, with a detection limit of 1-10 nM peptide, was used to screen the capacity of a panel of nine peptides bearing HLA-A2-binding motifs and derived from the human p53 tumor-suppressor protein sequence. Eight of the nine peptides bound to, and reconstituted, HLA-A2 on acid-treated cells. This assay system will enable the rapid identification of peptides binding to any class I allele, which is the initial prerequisite for elucidating potential CD8+ T-cell epitopes.

Intracellular transport of class I HLA molecules is affected by polymorphic residues in the binding groove.

Intracellular transport of class I MHC complexes is dependent on assembly of class I heavy chains with beta 2-microglobulin (beta 2m) and peptides. This suggests that amino acid residues of individual class I molecules which are important for their stability and transport are likely to include those which contribute to binding of a majority of the cleft-associated peptides. To identify such critical residues, substitutions at polymorphic positions within the peptide binding cleft were introduced into a mutant HLA-A*0201 molecule bearing an additional gly > lys substitution at position 242 (242K). The 242K mutation weakens association of the HLA-A*201 heavy chain with beta 2m and was used to enhance potential effects of substitutions in the peptide binding groove on class I stability. Critical in choosing which binding cleft positions to mutate was the observation that HLA-A*6801 was less sensitive to the effects of 242K mutation than HLA-A*0201 and A*6901. This suggested that one or more of the six residues in the alpha 2 domain differing between HLA-A*6901 and A*6801 were likely to affect class I complex stability. Positions 95, 97, 107, 114, 116, and 156 in either 242K or wild-type HLA-A*0201 molecules were therefore each converted to those residues found in HLA-A*6801. One of the second-site substitutions, arg > met at position 97, increased stability and restored surface expression of the 242K molecule. Five other substitutions either had no additional effect or further impaired 242K stability. Substitution of his > arg at position 114 blocked surface expression of both 242K and wild-type HLA-A*0201 molecules. These results demonstrate that polymorphic residues in the binding cleft influence the stability of class I complexes, and suggest that position 97 plays a critical role in stabilizing class I molecules for transport.

Calnexin influences folding of human class I histocompatibility proteins but not their assembly with beta 2-microglobulin.

Class I major histocompatibility complex heavy chains bind to calnexin before associating with beta 2-microglobulin (beta 2m) and peptides. Calnexin has been shown to retain in the endoplasmic reticulum those class I heavy chains which have not assembled properly and, thus, to serve as a quality control mechanism. In addition, calnexin may direct the folding of class I subunits or their subsequent assembly. We asked whether calnexin plays a role in the initial folding of HLA-B*0702 heavy chains by assessing disulfide bond formation in vivo. Our results show that class I heavy chains form intrachain disulfide bonds very soon after translation, and that calnexin is bound to both reduced and oxidized forms during this process. When a cell-permeable reducing agent, dithiothreitol, was added to cells, disulfide bond formation in newly synthesized heavy chains was substantially blocked, as was their association with calnexin. The reducing agent appeared to affect calnexin directly, since binding was similarly abolished to a subset of proteins which do not contain internal disulfide bonds. Addition of the glucosidase inhibitor castanospermine to cells, shown previously to disrupt calnexin binding to ligands, slowed formation of disulfide bonds but did not decrease the amount of assembled heavy chain-beta 2m complexes that formed. Our data suggest that calnexin can promote disulfide bond formation in class I heavy chains but does not directly facilitate subsequent binding to beta 2m.

Impaired assembly and transport of HLA-A and -B antigens in a mutant TxB cell hybrid.

Biosynthesis of HLA class I antigens has been studied in a variant B-LCLxT-LCL hybrid, 174XCEM.T2. This cell line encodes HLA-A2 and -B5, but expresses only small amounts of A2 antigen and undetectable B5 antigen at the cell surface due to a mutation inactivating a trans-acting regulatory gene encoded within the class II region of the human major histocompatibility complex. Northern blot analysis with HLA-A- and HLA-B-specific probes shows that 174XCEM.T2 synthesizes quantities of A and B locus mRNA comparable with its class I antigen-positive parent cell line. Immune precipitation studies indicate that 174XCEM.T2 synthesizes normal HLA heavy chains and beta 2-microglobulin which fail to form dimers. The heavy chains are N-glycosylated normally, but processing of the glycan to the complex form does not occur. In addition, free heavy chains in this cell line are not phosphorylated. Thus, the majority of class I heavy chains in 174XCEM.T2 do not combine with beta 2-microglobulin, and are not processed or transported to the cell surface. As both subunits are synthesized in normal amounts, we propose that an additional molecule absent from 174XCEM.T2 and encoded by an HLA-linked gene is necessary for efficient assembly of class I antigen subunits.

Distinct patterns of folding and interactions with calnexin and calreticulin in human class I MHC proteins with altered N-glycosylation.

Calnexin is a lectin-like chaperone that binds to class I MHC molecules soon after their synthesis, retaining unassembled heavy chains and also assisting their folding. Following association with beta2-microglobulin (beta2m) in the endoplasmic reticulum, a large proportion of human class I molecules release from calnexin, whereas mouse class I molecules do not. We asked whether addition of a second N-glycan to the human class I molecule A*0201 at position 176, a site present in mouse, would affect its binding to calnexin. The 176dg mutant with N-glycans at positions 86 and 176, when transfected into CIR cells, demonstrated increased binding to calnexin, detectable both before and after association with beta2m, and reduced interaction with calreticulin and TAP relative to wild-type protein bearing a single N-glycan at position 86. Cell surface levels of the mutant were decreased only slightly relative to the wild type, suggesting that the protein is not misfolded or grossly altered structurally. A subpopulation of mutant molecules was retained in the endoplasmic reticulum, and surprisingly, these molecules reacted with w6/32, which recognizes an epitope present on transport-competent class I HLA complexes. Transfection into Daudi cells demonstrated that 176dg reacts with w6/32 in the absence of beta2m, suggesting that the Ab epitope can be induced by binding of calnexin. These data may explain previously noted differences between mouse and human class I MHC proteins and demonstrate that the location of N-oligosaccharides within proteins can influence their folding and interactions with chaperones such as calnexin and calreticulin.

Genes regulating HLA class I antigen expression in T-B lymphoblast hybrids.

Regulation of HLA class I and class II antigen expression was studied in hybrids of human T and B lymphoblastoid cell lines (LCL). The T-LCL CEMR.3 expresses no HLA class II antigens. It expresses little total HLA class I antigen and no HLA-B antigens. The B-LCL 721.174 is a radiation-induced variant immunoselected for loss of class II antigen expression. In addition to showing a deletion of all HLA-DR and DQ structural genes, 721.174 expresses no HLA-B antigens and a decreased level of HLA-A antigen compared with the parental cell line. A hybrid of 721.174 and CEMR.3 expresses class II antigens encoded by CEMR.3. Increased expression of HLA class I antigens encoded by both 721.174 and CEMR.3 was also observed. Specifically, the previously undetectable HLA-B5 and HLA-Bw6 antigens encoded by 721.174 and CEMR.3, respectively, were present on the hybrid. Increased expression of the HLA-A2 antigen encoded by 721.174 was also observed. An immunoselected variant of the hybrid lacking both CEMR.3-derived copies of chromosome 6 lost expression of the HLA-B5 antigen encoded by 721.174 and expressed a decreased amount of HLA-A2. From these data, we infer that two complementary trans-acting factors mediate enhanced expression of HLA class I antigens in the hybrid. One of these factors is provided by a gene located on chromosome 6 derived from CEMR.3. The second factor, introduced by 721.174, is the gene previously postulated to induce expression of CEMR.3-encoded class I antigens in hybrids of CEMR.3 with B-LCL.

Role of endogenous peptide in human alloreactive cytotoxic T cell responses.

The T x B hybrid 174 x CEM.T2 (T2) has been shown to be defective in the processing of proteins for presentation by MHC class I molecules. It continues, however, to express significant quantities of HLA-A2.1, suggesting that this class I molecule is expressed either in a largely peptide-free form or in association with a small subset of peptides. In this paper T2 was used in conjunction with limiting dilution analysis to provide a direct estimate of the fraction of alloreactive cytotoxic T lymphocytes (CTLs) that were dependent upon the presence of peptides for their recognition of HLA-A2.1. Alloreactive cytotoxic T cell lines generated by stimulation with HLA-A2.1 expressing peripheral blood lymphocytes recognized T2 poorly. Split-well analysis of 240 clonal limiting dilution cultures demonstrated that this reflected the existence of two subpopulations. An average 85% of HLA-A2.1 specific CTLs recognized HLA-A2.1 on normal cells but not on T2. The remainder recognized HLA-A2.1 on both T2 and normal targets. CTL lines with the latter specificity could be generated by using T2 as a stimulator cell. Using target cells that either expressed a lower density of HLA-A2.1 or that expressed HLA-A2 molecules that had been mutated to affect CD8 binding, no significant differences in avidity between T2-reactive and T2-unreactive CTLs were seen. Thus the failure of the majority of alloreactive CTLs to recognize T2 is not a consequence of the lower level of HLA-A2.1 surface expression on this cell, but is instead due to the absence of appropriate epitopes.(ABSTRACT TRUNCATED AT 250 WORDS)

Phosphatase inhibitors block in vivo binding of peptides to class I major histocompatibility complex molecules.

Class I major histocompatibility complex (MHC) molecules are heterotrimers of heavy chains, beta 2-microglobulin, and 8-10 amino acid-long peptides. Assembly of class I MHC molecules into complexes which are stable and can be transported to the cell surface occurs soon after insertion of individual subunits into the endoplasmic reticulum (ER). To identify subcellular compartments required for class I MHC assembly, we studied class I biosynthesis in human cell lines treated with several inhibitors of intracellular transport. We found that HLA-B701 molecules do not assemble in CIR transfectants in which a block in protein transport from the ER is established by treatment with phosphatase inhibitors. In contrast, stable HLA-B701 complexes form in cells in which the ER becomes mixed with the Golgi after treatment with brefeldin A. Neither treatment impaired binding of HLA-B701 to the ER-resident protein calnexin, and unassembled heavy chains in phosphatase-inhibited cells showed prolonged association with calnexin. In addition, the mouse class I molecule H-2Db, which binds beta 2-microglobulin in human T2 cells in the absence of transporter of antigenic peptides, formed complexes in CIR cell transfectants treated with phosphatase inhibitors. Taken together, these data demonstrate that phosphatase inhibitors do not prevent assembly of class I heavy chain beta 2-microglobulin dimers, but instead interfere with peptide loading. These results are consistent with the possibility that class I MHC molecules are transported from their initial site of insertion into the rough ER before binding peptides, or alternatively that peptide loading mediated by transporter of antigenic peptides is blocked by phosphatase inhibitors.

Calnexin recognizes carbohydrate and protein determinants of class I major histocompatibility complex molecules.

Proper folding of nascent polypeptides is essential for their function and is monitored by intracellular "quality control" elements. The molecular chaperone calnexin participates in this process by retaining in the endoplasmic reticulum a variety of unfolded proteins, including class I major histocompatibility complex molecules. We transfected human B cell lines with genes encoding either wild-type HLA-A2 heavy chains or mutant heavy chains lacking sites for glycosylation or deficient in binding to beta 2-microglobulin (beta 2m). In CIR cells, calnexin did not associate detectably with wild-type heavy chains but bound strongly to mutant heavy chains unable to bind beta 2m. Removal of the glycosylation addition site by further mutagenesis prevented binding of mutant heavy chains to calnexin. In Daudi cells, deficient in synthesis of beta 2m, wild-type HLA-A2 heavy chains, but not a non-glycosylated mutant, bound calnexin. Castanospermine, which blocks trimming of glucose residues from asparagine-linked glycans, inhibited association of calnexin with heavy chains encoded by a second class I gene, HLA-B*0702. Although initiation of calnexin binding appears to depend on the presence of oligosaccharide on the substrate, removal of the glycan from calnexin-associated heavy chains by digestion with endoglycosidase H did not disrupt the interaction. These results suggest that calnexin first recognizes carbohydrate on substrate proteins and then binds more stably to peptide determinants, which disappear upon folding.