Trimeric Interactions of the Invariant Chain and Its Association with Major Histocompatibility Complex Class II (cid:97)(cid:98) Dimers*

The invariant chain (I chain) associates with major histocompatibility complex class II (cid:97)(cid:98) heterodimers upon synthesis, preventing them from binding peptides and unfolded proteins in the endoplasmic reticulum and directing class II transport to post-Golgi endosomal compartments. To assess which regions of the I chain are involved in binding class II molecules, we have stud- ied proteolytic fragments of the I chain generated both by natural proteolytic degradation of (cid:97)(cid:98) dimer-invari-ant chain complexes ( (cid:97)(cid:98) (cid:122) I) within human B cells and by in vitro digestion of purified (cid:97)(cid:98) (cid:122) I complexes with proteinase K. The 18-kDa luminal I chain fragment gener- ated by proteinase K, called K3, remains associated with (cid:97)(cid:98) dimers and retains the complex ( (cid:97)(cid:98) (cid:122) K3) in a high molecular mass nonameric configuration. The N termi- nus of the K3 fragment was identified as glycine 110. This indicates that the K3 fragment lies outside of the class II-associated invariant chain peptide region (ami-no acids 81–104) of the I chain, shown to be important for initial (cid:97)(cid:98) (cid:122) I assembly. An N-terminal 12-kDa I chain fragment called p12, generated intracellularly, was also an- alyzed and was found to remain associated with (cid:97)(cid:98) dimers in a high molecular mass form analogous to the nonameric (cid:97)(cid:98) (cid:122) I complex. These results demonstrate that at least two class II contact points exist along the length of the I chain and that different regions of the I chain can stabilize the (cid:97)(cid:98) (cid:122) I nonamer. Additional evidence suggests that the O -linked glycan(s) characteristic of the I chain is added to the short C-terminal region absent from the K3 fragment. AR film. In some experiments, radioactive protein bands were quantitated by a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA) and expressed in arbitrary PhosphorImager units.

Molecules encoded by the major histocompatibility complex (MHC) 1 genes serve to display peptides processed from endogenous or exogenous protein antigens to T cells. This process leads to T cell activation and amplification of both cellular and humoral immune responses to foreign antigen. MHC class I molecules bind peptides in the endoplasmic reticulum (ER) that are delivered from the cytoplasm by the transporters associated with antigen processing (reviewed in Ref. 1). In contrast, MHC class II molecules associate with the invariant chain (I chain) upon synthesis in the ER (2)(3)(4). This association prevents the class II ␣␤ heterodimer from binding peptides and unfolded proteins during transport of ␣␤ dimer-invariant chain complexes (␣␤⅐I) from the ER through the Golgi apparatus (5)(6)(7)(8)(9)(10). Inhibition of binding appears to result from the interaction of the class II-associated I chain peptide (CLIP) region of the I chain with the peptide-binding groove of the MHC class II molecule (11)(12)(13).
The I chain associates with class II molecules in the ER to form a nine-chain (nonameric) complex containing three I chain molecules and three ␣␤ dimers (14). After assembly with the assistance of the chaperone calnexin (15,16), the nonameric complex exits the ER, is transported through the Golgi apparatus, and localizes to a compartment in the endocytic pathway. A percentage of ␣␤⅐I complexes may transport to the cell surface and rapidly internalize before localizing in this compartment (17). There is an endosomal localization signal encoded within the cytoplasmic tail of the I chain (18). The exact nature of the class II molecule-containing endocytic compartments is not yet clear, but they are biochemically distinct from classical endosomes or lysosomes (19 -21). Class II ␣␤ dimers, the I chain, and endocytosed antigen colocalize in these compartments (22)(23)(24)(25). It is possible that the I chain encodes additional localization signals elsewhere within the molecule. The four different forms of the human I chain (p33, p35, p41, and p43) may regulate the localization of class II molecules in distinct endocytic compartments (26 -28).
Once ␣␤⅐I complexes reach late endocytic compartments, the I chain is degraded through a series of proteolytic steps in which cathepsins B and D have been implicated (29 -32). After removal of the I chain, ␣␤ dimers are capable of binding processed peptides and of being transported to the cell surface. A nested set of I chain peptides, from amino acid residues 81 to 104 in the human system, can be found associated with both human and mouse class II molecules (33)(34)(35)(36). This is the CLIP region of the I chain, which seems to be important for initial assembly of ␣␤⅐I complexes (37,38) and the removal of which is mediated by HLA-DM molecules (39 -41).
We initiated these experiments to determine how the I chain interacts with ␣␤ dimers and what regions are important in forming the nonameric ␣␤⅐I structure. We have previously demonstrated that ␣␤⅐I complexes remain nonameric until and even during endosomal proteolytic degradation of the I chain (42). To further characterize sites within the I chain that are important for contact with the class II ␣␤ dimer, we have analyzed additional I chain fragments formed in vivo and in * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  vitro. The results suggest that at least two sites along the length of the I chain molecule contact ␣␤ dimers and that fragments containing either of them can maintain the nonameric complex.

MATERIALS AND METHODS
Cells-The human B-lymphoblastoid cell line Pala (HLA-DR3) was grown in Iscove's modified Dulbecco's medium supplemented with 5% bovine calf serum (Life Technologies, Inc.). For generation of ␣␤⅐I complexes, Pala cells (5 ϫ 10 6 ) were radiolabeled with 0.25 mCi of [ 35 S]methionine (Tran 35 S-label, ICN Radiochemicals, Irvine, CA) in 1 ml of methionine-free Dulbecco's modified Eagle's medium (Life Technologies, Inc.) supplemented with 3% dialyzed fetal calf serum and 20 g/ml gentamicin (Life Technologies, Inc.) for 6 h at 37°C in the presence of 10 M monensin (Boehringer Mannheim). In some experiments, cells were labeled in the presence of 1 mM leupeptin (Sigma). For generation of ␣␤⅐p12 complexes, Pala cells were first starved in methionine-free medium, pulse-labeled as described above for 30 min in methionine-free medium, and then chased in medium containing excess unlabeled methionine for 3 h.
Antibodies-The hybridoma cell line L243 (anti-DR mAb) (43) was obtained from American Type Culture Collection (Rockville, MD). DA6.147 (anti-DR ␣ chain mAb) (44) was a generous gift of Dr. V. van Heyningen (Medical Research Council Clinical and Population Genetics Unit, Edinburgh, Scotland). BU45 (anti-invariant chain C-terminal mAb) was obtained from Binding Site, Inc. (San Diego, CA). The antipeptide mAb PIN.1.1 was developed in this laboratory by immunization with a peptide representing amino acids 12-28 of the I chain. The rabbit antiserum R.I.EQLP was generated against the same peptide. Rabbit anti-invariant chain serum to the CLIP region (amino acids 81-104) was developed by immunization with this peptide coupled to keyhole limpet hemocyanin. It was affinity-purified using the immunizing peptide coupled to -aminobutyl-agarose (Sigma). For immunoaffinity columns, purified Ig from mAb-containing ascites was coupled to cyanogen bromide-activated Bio-Gel A-15m resin (Bio-Rad) (45) at 2-3 mg/ml of resin.
Generation of ␣␤⅐K3 Complexes-Pala cells labeled as described above in monensin to generate a large pool of ␣␤⅐I complexes were lysed in TS buffer (0.01 M Tris, pH 7.4, 0.15 M NaCl) containing 2% C 12 E 9 (polyoxyethylene 9 lauryl ether, Sigma) with protease inhibitors (0.2 mM N ␣ -p-tosyl-L-lysine chloromethyl ketone, 0.5 mM phenylmethylsulfonyl fluoride, and 5 mM iodoacetamide). The clarified lysate was then passed over tandem immunoaffinity columns of normal mouse immunoglobulin, L243, and DA6.147, respectively. ␣␤⅐I complexes were eluted at high pH from the DA6.147 column as described (6) and then neutralized. Purified ␣␤⅐I was digested with 10 g/ml proteinase K (Boehringer Mannheim) for 90 min at 0°C to generate ␣␤⅐K3. For [ 35 S]Met radioactive sequencing, ␣␤⅐K3 complexes were separated by 12% SDS-PAGE and then transferred to polyvinylidene difluoride membrane (Immobilon P, Millipore Corp., Bedford, MA). The band corresponding to the 18-kDa K3 fragment was visualized by autoradiography, cut out, and subjected to in situ N-terminal Edman degradation. Edman degradation was performed at the Keck Protein and Nucleic Acid Chemistry Facility of the Yale University School of Medicine. Each cycle product was eluted with n-butyl chloride, dried onto a membrane, resuspended with 15% acetonitrile, and analyzed for [ 35 S]Met by scintillation counting.
Immunoprecipitation-Unless otherwise indicated, cells were lysed at 5 ϫ 10 6 /ml in TS buffer containing 2% C 12 E 9 and protease inhibitors for 45 min at 4°C. Lysates were clarified by centrifugation and precleared with normal rabbit serum and protein A-Sepharose (Sigma). Lysates were then incubated with mAb-containing ascites (3 l) or rabbit antiserum (10 l) for 1 h at 4°C. Immune complexes were immunoprecipitated by the addition of 30 l of a 50% (v/v) suspension of protein A-Sepharose or protein G-Sepharose (Pharmacia, Uppsala) and rotation at 4°C for 1 h. Sepharose beads were washed twice in TS buffer containing 0.1% C 12 E 9 and once in H 2 O prior to analysis by SDS-PAGE.
High Pressure Size Exclusion Chromatography (HPSEC)-For HPSEC separation, [ 35 S]methionine-labeled cells (5 ϫ 10 6 ; treated as indicated) were lysed in 125 l of TS buffer containing 2% C 12 E 9 as described above, and the clarified lysates were equilibrated in either 1% n-octyl glucoside or 0.6% CHAPS as indicated. Cell lysates were then injected over tandem KW-804 columns (Waters Associates, Milford, MA) in 50 mM sodium phosphate, pH 7.0, containing either 1% n-octyl glucoside or 0.6% CHAPS at a flow rate of 0.5 ml/min. Fractions were collected at the times and volumes indicated and then immunoprecipi-tated as described above. In some experiments, fractions were chemically cross-linked before immunoprecipitation by the addition of 4.5 l of 1 M Bicine/NaOH, pH 12.9, to bring the pH to 8.2 and then by the addition of 0.2 mg/ml 3,3Ј-dithiobis(propionic acid N-hydroxysuccinimide ester) (DTSP) (Sigma) for 30 min at 4°C. The cross-linking reaction was stopped by the addition of glycine to a final concentration of 20 mM.
O-Glycanase Treatment-The I chain or fragments were analyzed for the presence of O-linked glycans by immunoprecipitation with the PIN.1.1 monoclonal antibody as described above. Immunoprecipitated proteins, bound to protein G-Sepharose, were then washed in 0.05 M sodium acetate buffer, pH 5.5, containing 9 mM CaCl 2 . Proteins were treated with 25 milliunits of neuraminidase (Calbiochem) for 3 h at 37°C. Desialylated proteins were then washed in 50 mM sodium phosphate buffer, pH 7.0, containing 0.1% C 12 E 9 . These proteins were treated with 2 milliunits of O-glycanase (Boehringer Mannheim) overnight at 37°C. Deglycosylated proteins were analyzed by SDS-PAGE as indicated.
Electrophoresis-Samples were resuspended in Laemmli (54) sample buffer with 2-mercaptoethanol and separated by SDS-PAGE (12 or 13% acrylamide) as indicated. Radioactivity was visualized by fluorography using Enlightning (DuPont NEN) and Eastman Kodak X-Omat AR film. In some experiments, radioactive protein bands were quantitated by a PhosphorImager (Molecular Dynamics, Inc., Sunnyvale, CA) and expressed in arbitrary PhosphorImager units.

RESULTS
Digestion of Purified ␣␤⅐I with Proteinase K-Previous studies (46) demonstrated that limited digestion of purified ␣␤⅐I complexes with proteinase K yielded a distinct luminal I chain fragment of ϳ18 kDa referred to as K3 (Fig. 1). The K3 fragment of the I chain remained associated with MHC class II ␣␤ dimers. This fragment contains the N-linked glycans at positions 114 and 120, but not O-linked glycan. The data showed that a short segment of the C terminus was also removed (46). These results suggested that the N terminus of the K3 fragment was near or within the CLIP region (amino acids 81-104) of the I chain. To precisely define the N terminus of K3, metabolically labeled ␣␤⅐I complexes were purified from the HLA-DR3 homozygous B-lymphoblastoid cell line Pala and digested with proteinase K (Fig. 2). At 10 g/ml proteinase K, digestion yields a single I chain fragment of ϳ18 kDa that remains associated with ␣␤ dimers. The ␣ and ␤ chains are unaffected by proteinase K at this concentration. N-terminal sequencing of the K3 fragment, after transfer to polyvinylidene difluoride membrane, gave the results shown in Fig. 3. [ 35 S]Methionine peaks were detected at cycles 3, 12, and 18 that aligned with only one region in the I chain sequence and mapped the N terminus of the K3 fragment to the glycine residue at position 110. From the known N terminus of K3 and from the fact that it has lost a C-terminal segment of the I chain, at least 8 of the 14 methionine residues present in the intact I chain are absent from K3. This accounts for the reduced intensity of K3 compared with the I chain in Fig. 2.
Size Exclusion Analysis of ␣␤⅐K3 Complexes-Purified ␣␤⅐I exists as a nonameric complex (14). To determine whether ␣␤⅐K3 complexes retain the nonameric structure, purified ␣␤⅐I was digested with proteinase K and injected over a HPSEC column in the detergent CHAPS. Fractions were then immunoprecipitated with the monoclonal antibody BU45, specific for a luminal epitope within the I chain. The results shown in Fig.  4 show that ␣␤⅐K3 complexes elute in a broad peak at ϳ17.5-ml elution volume (V e ). This V e is only slightly greater than that determined for ␣␤⅐I complexes in CHAPS (17.25 ml) and suggests that the K3 fragment of the I chain can retain the nonameric association with ␣␤ dimers. The peak at 19.75 ml represents free K3 fragment, no longer associated with ␣␤ dimers. To further demonstrate that the material eluting at 17.5 ml represents a nonameric complex of three ␣␤ dimers and three K3 fragments, proteinase K-digested material was separated by HPSEC and treated with the chemical cross-linker DTSP prior to immunoprecipitation (Fig. 5). Following crosslinking, the material eluting at 17.5 ml has an apparent molecular mass of ϳ200 kDa by SDS-PAGE. This is consistent with ␣␤⅐K3 existing as a nonameric complex. It should be noted that ␣␤⅐K3 cross-links extremely efficiently, as does ␣␤⅐I (14). K3 trimers, however, cross-link poorly compared with I chain trimers (47). Thus, the regions of the I chain lost from K3 are important for efficient cross-linking to itself, but not to ␣ and ␤ chains.
Appearance of the p12 Fragment during Intracellular I Chain Proteolysis-Evidence suggests that once ␣␤⅐I complexes reach the appropriate endosomal/lysosomal compartment, the I chain is proteolytically digested in a stepwise fashion (29,48). Human B-lymphoblastoid cells grown in the presence of the sulfhydryl protease inhibitor leupeptin accumulate a partial proteolytic fragment of the I chain called LIP (29). This fragment is ϳ21 kDa and includes the N-terminal cytoplasmic region of the I chain as well as a portion of the luminal region including the N-linked glycans at positions 114 and 120 (29 -32). Pulse-chase analysis of untreated melanoma cells suggests that the class II-associated LIP fragment is subsequently cleaved by a leupeptin-sensitive protease to a 12-kDa fragment called p12 (49). To confirm this result in the human B-lympho-blastoid cell line, the HLA-DR3 homozygous Pala cells were pulse-labeled for 30 min with [ 35 S]methionine and then chased for 3 h. Immunoprecipitation of detergent-solubilized cell lysate (Fig. 6) demonstrates that anti-I chain antibodies (PIN.1.1 and R.I.EQLP) precipitate ␣␤ dimers in association with the intact I chain. In addition, these antibodies both recognize two smaller fragments of the I chain of ϳ12 and 10 kDa. Immunoprecipitation with the anti-DR monoclonal antibody L243, which recognizes a combinatorial epitope of the ␣ and ␤ chains, demonstrates that only the 12-kDa (p12) fragment of the I chain is associated with MHC class II ␣␤ dimers. L243 does not recognize ␣␤ dimers in association with the intact I chain.
Size Exclusion of ␣␤⅐p12 Complexes-HPSEC was performed to determine whether ␣␤⅐p12 complexes, like ␣␤⅐LIP and ␣␤⅐K3, remain associated with ␣␤ dimers in a nonameric complex. Pala cells were pulsed and chased to enrich for ␣␤⅐p12 as described, and then detergent lysates were injected over  2. Generation of ␣␤⅐K3 complexes. ␣␤⅐I complexes were purified from metabolically labeled Pala cells as described under "Materials and Methods." Purified ␣␤⅐I was treated with (ϩ) or without (Ϫ) proteinase K (10 g/ml) for 90 min at 0°C. After immunoprecipitation with purified BU45 ascites and protein G-Sepharose, the material was dissolved in reducing sample buffer and separated by 12% SDS-PAGE. The positions of the molecular mass standards are indicated in kilodaltons to the right of the gel. The positions of class II ␣ and ␤ chains, the I chain, and the K3 fragment are indicated to the left of the gel.

FIG. 3. Identification of the K3 N terminus by [ 35 S]Met sequencing.
Metabolically labeled ␣␤⅐K3 complexes were generated and separated by SDS-PAGE as described for Fig. 2. Proteins were transferred to polyvinylidene difluoride membrane, and the K3 band was visualized by autoradiography and cut out of the membrane. Sequencing was carried out by in situ N-terminal Edman degradation for 20 cycles, and the product of each cycle was analyzed in a liquid scintillation counter for 35 S cpm. The N terminus was determined to be at position 110 by aligning the radioactive peaks with Met residues in the I chain sequence.
HPSEC columns in CHAPS. Fractions were collected and immunoprecipitated with L243. The results in Fig. 7 show a large peak eluting at 18.75 ml that contains ␣ and ␤ chains and represents mature ␣␤ dimers. In addition, L243 immunoprecipitated a smaller peak of ␣␤ dimers associated with the p12 I chain fragment, which eluted at 17.5 ml. This V e is only slightly larger than that reported for ␣␤⅐I in CHAPS, suggesting that the p12 fragment retains higher order nonameric complexes with ␣␤ dimers. To examine this further, the same experiment was performed using the detergent n-octyl glucoside for HPSEC. In octyl glucoside, ␣␤⅐LIP complexes have been shown to dissociate from nonamers into trimeric complexes containing one ␣, one ␤, and one LIP molecule (42). HPSEC of ␣␤⅐p12 complexes in octyl glucoside (Fig. 8) showed that these complexes also dissociate from higher order nonamers (V e ϭ 18 ml) into complexes eluting at 19.5 ml, which probably represent ␣␤⅐p12 trimers. enrich for ␣␤⅐p12 complexes, were immunoprecipitated with PIN.1.1, and the precipitated material was treated with neuraminidase and then with O-glycanase to remove O-linked glycans. The results in Fig. 9A demonstrate that although the intact I chain drops slightly in molecular mass after O-glycanase treatment, the p12 fragment does not. The two left lanes are a longer exposure of the two right lanes in order to visualize the p12 fragment. The two right lanes are a shorter exposure to visualize the drop in molecular mass of the intact I chain. These data indicate that the p12 fragment does not contain O-linked glycan.
The LIP fragment of the I chain was analyzed in the same fashion as p12. The C terminus of the LIP fragment is Cterminal to the N-linked glycosylation site at position 120 (see Fig. 1). Pala cells labeled in the presence of leupeptin were lysed and immunoprecipitated with PIN.1.1. Immunoprecipitated proteins were treated with neuraminidase followed by O-glycanase. The results in Fig. 9B show a slight drop in molecular mass of the LIP fragment after removal of sialic acid, but no further reduction after treatment with O-glycanase. The reduction in molecular mass after neuraminidase treatment is attributable to removal of sialic acid residues from the N-linked glycans present on the LIP fragment at positions 114 and 120. As a control, CD45 was immunoprecipitated from the same cell lysate and treated with neuraminidase and O-glycosidase. This heavily O-glycosylated protein showed a large drop in molecu- Pala cells were labeled to enrich for ␣␤⅐p12 as described for Fig. 6 and then lysed and equilibrated in 0.6% CHAPS. Lysates were separated by HPSEC in 0.6% CHAPS, and fractions were collected (V e ; indicated along the top of the gel). Each fraction was then immunoprecipitated with L243 ascites. The immunoprecipitated material was analyzed by 13% SDS-PAGE. Shown beneath the gel are traces of ␣ chain, ␤ chain, and p12 fragment signals in each lane of the gel generated by PhosphorImager analysis and reported in arbitrary units. The positions of the ␣ chain, ␤ chain, and p12 I chain fragment are indicated to the left of the gel. The positions of the molecular mass standards are indicated in kilodaltons to the right of the gel. lar mass after treatment (data not shown). These data demonstrate that, at least in human B cells, little or no I chain is modified by O-glycosylation in proximity to the plasma membrane. Combined with the earlier data, this argues that the O-linked glycan(s) of the I chain are close to the C terminus, within the region cleaved from the K3 fragment.

DISCUSSION
The data presented here demonstrate that at least two sites along the length of the I chain interact with the MHC class II ␣␤ dimer. Analysis of natural and in vitro generated fragments of the I chain also reveals that these fragments remain associated with class II molecules in higher order multimers analogous to nonameric ␣␤⅐I complexes. In vitro digestion of purified ␣␤⅐I with proteinase K yields an 18-kDa I chain fragment (K3) containing most of the C-terminal luminal portion of the I chain (46). The N terminus of this fragment was mapped to amino acid 110. Size exclusion (Fig. 4) and cross-linking (Fig. 5) analyses suggest that the K3 fragment remains associated with DR molecules in a nonameric complex. With its N terminus at position 110, the K3 fragment does not include the CLIP region of the I chain extending from positions 81 to 104. Nested sets of I chain peptides from this region have been isolated from mature class II molecules in mouse and human cells (33)(34)(35)(36). Evidence suggesting that CLIP is bound in the class II peptide-binding groove (11,12) has been confirmed by x-ray crystallographic analysis (13), and the CLIP region has recently been shown by deletion analysis to be critical for initial assembly of ␣␤ and I chains (37,38). However, it is clear from the data presented here that once the ␣␤⅐I complex is formed, a site distal to the CLIP region can maintain the interaction with class II molecules. It should be noted that ␣␤⅐K3 complexes are isolated from monensin-treated cells and therefore have immature N-linked glycans. It is conceivable that this may contribute to the stability of the nonameric ␣␤⅐K3 complex.
A second site of I chain interaction with class II molecules is present in the p12 fragment. This is a 12-kDa fragment that is transiently generated within a class II-expressing cell during proteolytic degradation of the I chain (49). It includes the N-terminal cytoplasmic domain and, based on its apparent molecular mass by SDS-PAGE, the transmembrane region. The precise C terminus of p12 is not known, but its apparent size on SDS-PAGE leads to a predicted C terminus around amino acids 100 -110, consistent with its lack of N-linked glycans. If this prediction is correct, then the p12 fragment should include at least a part of the CLIP region (amino acids 81-104) of the I chain. It is possible, however, that because this fragment is relatively small and includes the transmembrane segment, it might bind an aberrant number of SDS molecules, FIG. 8. HPSEC of ␣␤⅐p12 in octyl glucoside. Pala cells were labeled and lysed as described for Fig. 6, and then the lysate was equilibrated in 1% octyl glucoside. The lysate was injected over HPSEC columns equilibrated in 1% octyl glucoside and fractionated by V e (indicated across the top of the gel). Each fraction was then immunoprecipitated with the anti-DR ␣␤ chain mAb L243. The immu- disrupting the charge/mass ratio. An anti-CLIP antiserum failed to react detectably with p12 (data not shown), but the same serum reacted only weakly with the intact I chain, and the presence or absence of all or part of the CLIP sequence in p12 remains unknown. If the CLIP region is associated with the peptide-binding groove of class II molecules, it is likely to be protected from proteolysis. It is therefore tempting to speculate that p12 contains CLIP, while p10, which fails to bind class II molecules (Fig. 6), does not.
This report, in combination with data on the LIP fragment of the I chain (42), has detailed the analysis of fragments that cover nearly the entire length of the I chain. p12 represents a small N-terminal fragment. LIP is also an N-terminal fragment, but includes a portion of the luminal domain including the CLIP region. Finally, K3 is a C-terminal fragment that overlaps LIP, but does not include the CLIP region. None of these fragments contain the O-linked glycan(s) characteristic of the intact I chain. Based on the proposed N and C termini of all the fragments (Fig. 1), it seems likely that the O-linked glycan is close to the C terminus of the intact I chain, a portion of which is lacking in K3 (46). All three of the I chain fragments remain associated with MHC class II molecules in a nonameric structure similar to that reported for ␣␤⅐I complexes (14). Recombinant I chain constructs that include the C-terminal region have previously been shown to be trimeric (50). The data presented here and elsewhere (19) argue that the N-terminal regions of the I chain also interact to form a trimeric structure. This interaction appears to be variably stable in solution depending upon the detergent used. Thus, both ␣␤⅐LIP and ␣␤⅐p12 retain a nonameric structure in CHAPS, but appear to dissociate into simple trimers in n-octyl glucoside (Figs. 6 and 8) (6). This suggests a possible interaction of the I chain transmembrane regions. The requirement for C-terminal interactions to maintain a stable I chain trimer in conventional detergents may explain the observations of a number of groups that C-terminally truncated I chain constructs fail to trimerize, as detected by chemical cross-linking, when expressed in vivo (38,51). Alternatively, the C-terminal region may be a primary site for efficient cross-linking. Data from Wiley and co-workers (50,52) indicate, based on NMR analysis, that the region of the I chain between the membrane and position 118 is extended and that this region is highly susceptible to proteolysis. Thus, a picture is emerging where the I chain interacts at both the N and C termini to generate a trimeric structure, and a class II ␣␤ dimer interacts with each component of the trimer at the CLIP region via the peptide-binding groove and in an unknown fashion with the C-terminal region (53). Whether additional interactions between the class II molecules and the I chain exist, perhaps via the transmembrane region or the cytoplasmic domains, remains unknown.