Exon 6 is essential for invariant chain trimerization and induction of large endosomal structures.

Invariant chain (Ii) is a transmembrane type II protein that forms a complex with the major histocompatibility complex (MHC) class II molecules in the endoplasmic reticulum (ER). The membrane proximal luminal region of Ii is responsible for the non-covalent association with MHC class II molecules. Chemical cross-linking in COS cells was used to study the effect of luminal and cytoplasmic deletions on trimerization of Ii. We demonstrate that trimerization of Ii is independent of the cytosolic tail of Ii, whereas residues 162–191 (the sequence encoded by exon 6) in the luminal part of Ii are essential for trimer formation. Immunofluorescence studies of the transfected luminal deletion constructs show that the amino acids encoded by exon 6 of Ii are also essential for the induction of large endosomal vesicles. The data suggest that Ii must be in a trimeric form to modify the endosomal pathway.

The major histocompatibility complex (MHC) 1 class II molecules are polymorphic heterodimers, consisting of an ␣and a ␤-chain. Invariant chain (Ii) is a type II transmembrane protein that associates with the major histocompatibility complex (MHC) class II molecules in the endoplasmic reticulum (ER) (1,2). The formation of the ␣␤Ii complex occurs by the sequential addition of one ␣and one ␤-chain to a pre-existing core of trimeric Ii molecules (3,4). This results in a stoichiometric 9-subunit complex (5). Following subunit assembly in the ER, the (␣␤Ii) 3 complexes traverse the Golgi complex. Approximately 1-2 h after biosynthesis, Ii molecules are sorted to and proteolytically processed in acidic endosomal compartments (6 -8), and MHC class II molecules with bound peptide are routed to the plasma membrane (9).
Invariant chain is composed of three domains: an N-terminal cytosolic tail of 30 amino acids, a membrane spanning region between residues 31 and 56, and a luminal C-terminal domain of 160 residues (159 in mouse) (10 -14). An alternative form of the human Ii results from translation initiation at an upstream start codon, extending the cytosolic tail by 16 residues. Both human and mouse Ii contain an alternatively spliced exon in their luminal domains. These factors, together with various types of post-translational modifications create numerous forms of the molecule (reviewed in Ref. 15).
Invariant chain contributes to the function of the MHC class II molecules in several ways. Invariant chain is responsible for efficient transport of MHC class II molecules out of ER into the endosomal pathway. This function is mediated by sorting signals in the cytosolic tail of Ii that direct the ␣␤Ii multimer to the endocytic pathway (16 -19). The extra 16 N-terminal amino acid residues in the cytosolic tail of Ii include a strong ER retention signal (20), which also mediates retention of the associated MHC class II molecules. The biological significance of the retention of Ii in the ER remains to be elucidated. The cytosolic tail of wild-type Ii also contains information essential for the formation of large endosomal structures (7,21). Induction of these large endosomal vesicles may be related to a delay in anterograde transport of fluid phase markers, as observed in the endosomal pathway of cells transfected with Ii (21,22).
The region of Ii responsible for association with MHC class II molecules has been assigned to residues 81-109 in the Ii proximal membrane region (23). This association allows Ii to function as a chaperone for MHC class II molecules in the ER (24,25) and to prevent premature peptide binding (26), thereby mediating transport to endosomal compartments and antigen presentation (18,21,27,28).
In this study, we have investigated, by chemical cross-linking, how the multimerization of Ii depends upon the different parts of the molecule. By studying the intracellular distribution of various deletion mutants of Ii, we found that amino acids encoded by exon 6 of Ii were essential for the trimerization of Ii and the formation of large endosomal structures.

MATERIALS AND METHODS
Expression vectors and plasmid constructs-The pSV51L vector is a late replacement expression vector with an SV40 promoter, which gives high transient expression of several proteins in simian cells (16,29). pSV51L Ii (p33 hIi) and ⌬20 hIi have been described earlier (16). p35* hIi, in which the second methionine has been mutated, was a gift from E. Long, National Institutes of Health, Rockville, MD (30); ⌬30 hIi (⌬N) and ⌬26 hIi (⌬E), lacking amino acids 1-30 and 1-26, respectively, were a gift from J. Lipp and B. Dobberstein, Heidelberg (16). These three constructs were subcloned into the pSV51L vector.
Antibodies-BU45 is a mouse monoclonal antibody that recognizes a conformation-dependent epitope within the luminal part of human Ii (31). Purified BU45 was purchased from The Binding Site (San Diego, CA). In1 is a monoclonal rat antibody (IgG2b) recognizing an epitope in the cytosolic tail of mouse Ii (13). In1 does not bind well to protein A-Sepharose beads, and an additional monoclonal mouse antibody against the constant region of rat antibody chain, mouse anti-rat (MAR 18.5) (32), was used for immunoprecipitation. In immunofluorescence microscopy experiments, In1 and MAR 18.5 were used with goat anti-mouse FITC (Zymed Laboratories, San Francisco, CA) as the third antibody.
Cells and Cell Culture-COS cells are Green monkey kidney cells, originating from CV1 cells by stable transfection with an origin-defective mutant of the SV40 vector, coding for the wild-type T antigen (33) necessary to initiate replication of the SV40 vector (34). COS cells were obtained from ATTC (CCL70). COS cells are reported to be devoid of detectable mRNA coding for either MHC class II or Ii (35). The cell lines were maintained in Dulbecco's modified Eagle's medium (DMEM) supplemented with 10% fetal calf serum (FCS), 30 units of penicillin/ml and 30 g of streptomycin/ml. Media, serum, and antibiotics were purchased from Life Technologies, Inc.
Transient Expression in COS Cells-Transient expression of proteins in COS cells was performed essentially as described in Huylebroeck et al. (29). Briefly, 80% confluent cells were split 1:5 into 35-mm wells the day prior to transfection. 0.5-1.0 g of DNA in 100 l of DMEM with 10% NuSerum (Collaborative Research) added dropwise to 1 ml of DMEM with 10% NuSerum, containing 400 g/ml DEAE-dextran and 100 M chloroquine, were used per 35-mm well. Cells were washed twice in 2 ml of PBS (pH 7.4) prior to the addition of DNA. After 3-4 h of incubation at 37°C in 6% CO 2 , the medium was removed, and the cells were exposed to 10% Me 2 SO in PBS (0.5 ml/well) for 1.5 min. Incubation for another 1-2 days in DMEM/FCS permitted expression of the proteins.
Metabolic Labeling and Chemical Cross-linking-COS cells in 35-mm wells were metabolically labeled two days after transfection. The cells were starved in methionine/cysteine-free DMEM for 30 min prior to labeling in 0.5 ml of the same medium containing 150 Ci/ml [ 35 S]methionine/cysteine cell labeling mix (DuPont NEN) for 0.5-3 h. After labeling, the cells were placed on ice, washed twice in ice-cold PBS, and extracted at 2 ϫ 10 6 cells/ml in lysis buffer (1% C 12 E 9 , 0.13 M NaCl, 0.02 M bicine (pH 8.2)), or 0.5 ml of lysis buffer per well, for 20 min at 4°C. Cross-linker in Me 2 SO was then added to the extracts to a final concentration of 0.5 mM DSP, 50 mM DTSSP (Pierce), 50 mM sulfo-EGS (Pierce), or 4 mM DGS (Pierce) and incubated for another 30 min. Cross-linking was terminated by the addition of 1 M glycine in PBS (10 mM final concentration) and a mixture of protease inhibitors ( 4 g/ml phenylmethylsulfonyl fluoride, 2 g/ml antipain, 2 g/ml leupeptin, and 1 g/ml pepstatin final concentrations). After 10 min on ice, samples were centrifuged to remove cell debris and nuclei (10,000 ϫ g for 10 min at 4°C). The supernatants were either used directly for immunoprecipitation or frozen at Ϫ80°C.
Immunoprecipitation-The cell lysates were precleared with protein A-Sepharose beads (Pharmacia Biotech Inc.) for 1 h at 4°C, and after centrifugation, the supernatants were used for the immunoprecipitation. Immunoprecipitation was performed by adding 1-2 l of antibody serum or ascites to 300 -500 l of lysate and incubating for at least 2 h at 4°C. 35 l of protein A-Sepharose beads in a 1:1 slurry in wash buffer (0.1% C 12 E 9 , 0.13 M NaCl, 0.02 M bicine (pH 8.2)) were then added to each sample, and the tubes were incubated for 1-2 h at 4°C. The beads were washed 4 times in wash buffer and boiled for 5 min with 25 l of Laemmli sample buffer (36). In samples used for cross-linking studies, the dithiothreitol was omitted in the sample buffer since the crosslinker is non-functional under reducing conditions. The samples were then analyzed by SDS-PAGE on 6 -12% or 12% polyacrylamide gels using the Laemmli procedure (36). Fluorography of the gels was performed using Amplify (NAMP 100) according to the manufacturer instructions (Amersham Life Science, Inc.).
Immunofluorescence-COS cells, grown on coverslips, were fixed in 3% paraformaldehyde and permeabilized with Triton X-100 as described earlier (16). The cells were incubated for 30 min with a primary antibody, washed three times in PBS, and incubated with the secondary antibody for 30 min (these steps were repeated when more than two antibodies were used). The coverslips were washed in PBS and mounted in Mowiol. The cells were analyzed with a Nikon microphot EPI-FL3 microscope equipped for immunofluorescence.
Labeling with Endocytic Marker-Transiently transfected COS cells were incubated for 1 h at 37°C in DMEM/FCS containing 1 mg/ml albumin-Texas Red (Molecular Probes). The cells were washed three times in PBS before different chase periods in DMEM/FCS at 37°C. The cells were fixed and labeled with antibody as described above. All chemicals were purchased from Sigma except where noted.

Multimerization of Human Ii and Human Ii N-terminal Deletion Constructs-Human
Ii (hIi) has been shown to make trimers in the mutant B-lymphoblastoid T2 cells lacking MHC class II molecules (3). Trimerization of Ii has been studied in vitro, and it was found that residues 163-183 are essential (37). These data do not exclude, however, the possibility that other parts of the molecule might be important for trimerization. To study the influence of the Ii cytosolic tail on multimerization, COS cells were transiently transfected with p33 hIi, p35* hIi, or N-terminal deletion mutants of hIi (Fig. 1), and complexes were stabilized by chemical cross-linking. The transfected cells were metabolically labeled for 1 h and chased for 30 min prior to chemical cross-linking, immunoprecipitation, SDS-PAGE, and fluorography. The results are shown in Fig. 2. All hIi proteins containing N-terminal deletions migrated as trimers after cross-linking ( Fig. 2A, broad bands of approximately 90 kDa) and as higher molecular weight complexes, whereas no monomers were detected (not shown). Resolved cross-linked complexes consist exclusively of the different hIi constructs (Fig. 2B) as compared with the non-cross-linked samples (Fig.  2C). The deletion constructs are also able to dimerize, including ⌬30 hIi which lacks the cytosolic cysteine. This shows that disulfide binding by this cysteine is not essential for the dimerization of Ii. Under these conditions, we also find the trimeric form of Ii in the absence of cross-linker in cells transfected with p33 hIi, p35* hIi, and ⌬20 hIi, demonstrating that the crosslinking procedure does not induce trimers but can stabilize them. We have thus confirmed that Ii may form a stable but non-covalent trimer and that this process does not require an intact Ii cytosolic tail.
Multimerization of Mouse Ii and Mouse Ii with Deletions in the Luminal Domain-A series of deletion mutants of mIi (Fig.  3) were used to study the importance of the luminal parts for trimerization. All constructs, except mcIi, were made using a genomic clone. The DNA was transfected into COS cells, which were metabolically labeled and exposed to cross-linker as described under "Materials and Methods." Four different crosslinkers were used, producing similar results. However, shorter cross-linkers gave a higher fraction of multimers in the cross- linked samples. Full-length mIi, like hIi (Fig. 2), multimerized as a complex of about 90 kDa, corresponding to a trimer (Fig.  4A, lane 1). The mIi cDNA (mcIi) gave identical results (Fig.  4A, lane 2). Transfection of mIi ⌬81-127, mIi ⌬110 -130, mIi ⌬110 -161, and mIi ⌬192-212 resulted in bands corresponding to the molecular weight of trimers (Fig. 4A, lanes 3, 4, 5, and 9), whereas no trimers were visible for mIi ⌬126 -215, mIi ⌬153-208, mIi ⌬153-215, and mIi ⌬128 -215 (lanes 6, 7, 8, and 10). Constructs that do not encode exon 6B also trimerize. The exon 6B-encoded sequence is therefore not essential for the selfassociation of Ii, and we can conclude that the essential sequence for Ii trimer formation is contained within residues 153-191. To further investigate the trimerization region, mcIi was mutated to encode a stop codon at position 163. This gave a construct encoding aa 1-162 of mIi (Fig. 4, A and B, lane 11). Cross-linking of this construct did not produce any trimer, suggesting that the Ii trimerization site lies within residues 163-191, which are encoded by exon 6. The non-trimerizing group of constructs, lacking exon 6, ran as monomers, dimers, and higher molecular weight complexes, probably aggregates. Monomers were also detected for C-terminal deletion constructs that make trimers (except for mIi ⌬192-212 (Fig. 4A,  lane 9)), indicating that trimerization of these deletion constructs was not complete. When the multimers were resolved by reducing the cross-linking agent DSP by dithiothreitol (Fig.  4B), the main bands migrated as expected for monomers.
To summarize, four of the nine deletion constructs make trimers (Fig. 4A, lanes 3, 4, 5, and 9), whereas no trimers were detected for the other constructs. The common denominator for the "trimerization-group" is that all constructs contain exon 6 (residue 162-191), whereas the others do not, strongly indicating that sequences encoded by exon 6 are essential for Ii trimer formation in intact cells.
Intracellular Localization of mIi and Deletion Constructs-Ii contains, in its cytosolic tail, signals for sorting to endosomes (16,17,19). Furthermore, high expression levels of full-length Ii induce large endosomal structures in a subpopulation of the transfectants (7,21). The large endosomes are characterized by size and by the fact that the N-terminal 11 amino acids of Ii are essential for their induction (7,21). To study the influence of the luminal part of Ii on its intracellular distribution, immunofluorescence microscopy studies were performed on COS and CV1 cells transfected with mIi and mIi C-terminal deletion constructs. All the resulting proteins were located in vesicular structures (indicated by long and short arrows in Fig. 5). Cells transfected with constructs that contain exon 6 showed, in addition, large endosomal structures (Fig. 5, long arrows). To assure that the large vesicles had the same characteristics as those described earlier (7, 21), we allowed some of the transfected cells to internalize albumin-Texas Red for 1 h at 37°C.  1, 2, 3, 4, and 5 represent COS cells transfected with p35* hIi, p33 hIi, ⌬20 hIi, ⌬26 hIi, and ⌬30 hIi, respectively, and immunoprecipitated with BU45. A, samples cross-linked with DSP, 6 -12% SDS-PAGE, non-reducing conditions. B, samples cross-linked with DSP, 12% SDS-PAGE, reducing conditions. C, samples not cross-linked, 12% SDS-PAGE, non-reducing conditions.
FIG. 3. Map over mouse invariant chain and the C-terminal deletion constructs. Different exons are indicated by separate boxes, their length being relative to their amino acid content. Dark gray areas correspond to expressed amino acids, and white areas represent deleted and/or not expressed amino acids in the protein formed. Light gray areas represent the amino acid sequence of exon 6B, which might or might not be expressed due to alternative splicing. mgIi is the p33 form of mouse Ii from genomic DNA, and mcIi is the corresponding Ii from cDNA. The deletion constructs are named with the missing amino acids. In mIi ⌬153-208 and mIi ⌬192-212, the amino acid numbers include exon 6B and will extend the 215 amino acids that Ii normally contains. The albumin-Texas Red accumulated in large vesicular structures and in smaller endosomes after 1 h (Fig. 6, A and B, C and  D, and E and F). After a 6-h chase period, the albumin-Texas Red was still detectable in the large vesicles together with mIi ⌬81-127 (Fig. 6, G and H) and with other constructs inducing large vesicles (not shown). In smaller endosomes, the endocytosed marker was not seen after 6 h. The large endosomes have an increased volume and contain more material, and probably also fluid phase marker, than smaller endosomes. It is therefore not possible to compare the different sized compartments since the material maintained in the small endosomes may be below detection level. The detectable level of fluid phase marker co-localizing with Ii in the large endosomes after 6 h do, however, suggest that the fluid phase marker is retained, as endocytosed material normally reaches the lysosomes after 1-3 h (38,39). As no constructs lacking exon 6 induced changes in endosomal morphology, our results thus suggest that the exon 6-encoded sequence is a prerequisite for the formation of large endosomal structures. DISCUSSION The ability of Ii to make trimers in a cell-free system has been assigned to residues 163-183 (37). An earlier hypothesis of Elliot et al. (40) suggested involvement of a predicted amphipathic helix, between residues 146 -164 in Ii self association. NMR studies performed by Jasanoff et al. (41) show that a luminal trimeric part of the invariant chain, expressed in E. coli, contains a protease resistant, compact, and well structured region (aa 118 -193). This indicates that the luminal region involved in trimerization might go beyond aa 163-183, even though these aa are essential for initiation of the trimer formation. In our study, we have expressed various cytosolic and luminal deletion mutants in COS cells and found that the Ii luminal segment encoded by exon 6 (residues 162-191) is essential for the formation of Ii trimers. The cytosolic tail and all the luminal domains of Ii, except the sequence encoded by exon 6, were not essential for trimer formation although the fraction of detected trimers was lower for some of the deletions. This shows that trimerization is not induced by the N-terminal cytosolic tail although NMR studies have shown that a peptide representing the tail is found as a homotrimer. 2 Various chemical cross-linkers were generally needed to stabilize the multimers, but trimers of p35* hIi, p33 hIi, and ⌬20 hIi were also detected under nonreducing conditions without the use of cross-linker. This might indicate that parts of the cytosolic tail of Ii have a stabilizing effect on the trimer after its induction. This is supported by the work by Amigorena et al. (42) where they showed that a leupeptin-induced invariant chain peptide of 10 kDa was still in a trimeric conformation even though it only contained aa 1-98.
From our data and those of Bijlmakers et al. (37), we have concluded that the region responsible for the formation of trimers is located at exon 6. The study of Bijlmakers et al. (37) was limited to the in vitro ER events since microsomes were used to study the interactions, whereas we analyzed the trimerization of Ii in transfected cells. The study of Newcomb and Cresswell (31)showed that the trimeric complex was maintained throughout the biosynthetic pathway of Ii, most likely including endosomes. Degradation of Ii occurs from the C terminus, and by inhibiting degradation by adding leupeptin, they got an Ii-degradation product of approximately 140 amino acids, called leupeptin-induced protein (LIP). The nonamer (MHC class II ␣and ␤-chain-LIP) 3 is no longer kept together under conditions where the full size (␣␤Ii) 3 complex is maintained, hence the (␣␤Ii) 3 complex is disturbed upon partial Ii degradation. This suggests that Ii remains in a trimeric form at least until residues encoded by exon 6 are degraded and maybe even longer (42). Our data show that the full-length Ii, without the MHC class II, are found to be exclusively trimeric after a 1-h labeling and 30-min chase. At this time point, a significant fraction is expected to have passed the Golgi apparatus, imply- Some of the constructs encode exon 6B (mIi ⌬153-208 is fused to exon 6B and will, therefore, always express it (see Fig. 3)), which can give an extra band of higher molecular weight. Ii contains in its membrane spanning region a signal peptidase recognition site (50) that, occasionally, when used, can give a smaller product than the original. Molecular weights of the major bands of each construct correspond well with their amino acid content. Lanes 1-11 correspond to constructs 1-11 in Fig. 5 and 1-10 in Fig. 3. A, samples cross-linked with DGS, 6 -12% SDS-PAGE, non-reducing conditions. B, samples cross-linked with DSP, 12% SDS-PAGE, reducing conditions. ing that Ii remains trimeric until it reaches the endosomal pathway.
The more detailed sequence requirements for the trimerization of Ii are not known. Residues encoded by exon 6, which we find to be the essential region for trimerization of Ii, are 70% identical in mouse, rat, and human, whereas with chicken, the identity is about 30% (Fig. 7). In the sequence of human, mouse, rat, and chicken Ii, there are three well conserved sequences, one in the N-terminal tail, one in the transmembrane region, and the last in exon 6. The conserved region encoded by exon 6 is a largely unpolar sequence, Phe-Glu-Ser2Thr-Trp-Met-His2Lys-Trp-Leu-Leu-Phe-Glu-Met, and this stretch of residues is thus a potential candidate for the core sequence required for the initiation of the Ii homotrimer formation.
Ii is detected in endosomal vesicles in antigen-presenting cells as well as in transfected cells, with or without MHC class II molecules (16 -18, 43). The transfected mouse Ii deletion constructs with an intact cytosolic tail were all transported to endosomal vesicles, either by direct transport or by internalization from the plasma membrane. Ii with exon 6 deleted exists mostly in a monomeric form but is still sorted to endosomes. The chimeric protein INA (a tetrameric protein) containing the Ii cytosolic tail fused to the transmembrane and luminal domains of neuraminidase (NA), is also sorted to endosomes (19). We may therefore conclude that trimerization of Ii is not essential for endosomal sorting. Such a conclusion is supported by the data of Arneson and Miller (44), showing that Ii complexes with monomeric tails also mediate transport to the endosomal pathway.
Romagnoli et al. (21) showed that high expression levels of Ii induced large endosomal structures in a subpopulation of transfected COS cells. Cells with these enlarged endosomal vesicles retained fluid phase markers in the large endosomes for a prolonged period of time before reaching the lysosomes. Large vesicular structures were also found in the human fibroblast M1 cell line (22,45) and canine Madin-Darby canine kidney cells, 3  native antigen presenting cells, they are observed in Langerhans cells (46). We might speculate that the phenotypic changes induced by Ii and the delay of anterograde endosomal transport of fluid phase markers might be advantageous for antigen presentation, as MHC class II molecules and antigen could co-reside in the endosomes for a prolonged period of time. Studies performed on cells expressing these enlarged endosomal structures show that these organelles may be considered as both early endosomes and late endosomes (7,21,45), but except for the delay in anterograde transport of fluid phase markers, they are functionally similar to normal endosomes. Ii lacking the first 11 amino acids does not induce large vesicular structures (7,21). Further elucidation of the signal mediated by the first 11 amino acids has revealed that N-terminal acidic charges and the first di-leucine based signal in Ii were both essential for the induction of the large endosomes (47). However, as shown here, such signals may induce large vesicular structures only in combination with a trimeric Ii. The study by Pond et al. (47) showed that N-terminal acidic residues and di-leucine-based internalization signals from other proteins can mediate the induction of large endosomal structures when transferred to Ii but not in their original surroundings. This and our earlier results with Ii fusion proteins (7,19) indicate that there is more than one requirement in Ii that has to be fulfilled to induce enlargement of the endosomes.
Ii is associated with MHC class II molecules, and the trimer of Ii is a prerequisite for the formation of the transport competent (␣␤Ii) 3 nonameric complex (4). Ii, or more specifically trimeric Ii, is probably necessary for the correct folding of MHC class II molecules as MHC class II molecules expressed in cells lacking Ii have an altered conformation (48). For human Ii, it has been suggested that the trimer of Ii is necessary to retain the ␣, ␤ and Ii molecules in the ER to allow the correct functional conformation to occur (4). Another theory, presented by Roche et al. (5), proposed that the trimer is necessary to prevent the binding of peptides to MHC class II molecules in the ER. Arneson and Miller (44), showed that the localization signals encoded by residues 2-17 in p33 hIi must be in a dimeric or trimeric form to mediate a direct transport to the endosomal pathway. This may also be an important function of the trimer. Recent results of Bertolino et al. (49) indicate that the function of trimerization is of importance for antigen presentation, as the region of Ii essential for the presentation of some peptides from HEL (hen egg lysozyme), residues 131-191, also covers the region of Ii essential for multimerization. These theories for the function of trimeric Ii may all be true; however, since the mechanisms involved are not known, we may not yet fully understand the significance of the trimerization of Ii for the endosomal transport process or for antigen presentation.