Cysteines in CH1 Underlie Retention of Unassembled Ig Heavy Chains*

Conformation, structure, and oligomeric state of immunoglobulins not only control quality and functional properties of antibodies but are also critical for immunoglobulins secretion. Unassembled immunoglobulin heavy chains are retained intracellularly by delayed folding of the CH1 domain and irreversible interaction of BiP with this domain. Here we show that the three CH1 cysteines play a central role in immunoglobulin folding, assembly, and secretion. Remarkably, ablating all three CH1 cysteines negates retention and enables BiP cycling and non-canonical folding and assembly. This phenomenon is explained by interdependent formation of intradomain and interchain disulfides, although both bonds are dispensable for secretion. Substituting Cys-195 prevents formation not only of the intradomain disulfide, but also of the interchain disulfide bond with light chain, BiP displacement, and secretion. Mutating the light chain-interacting Cys-128 hinders disulfide bonding of intradomain cysteines, allowing their opportunistic bonding with light chain, without hampering secretion. We propose that the role of CH1 cysteines in immunoglobulin assembly and secretion is not simply to engage in disulfide bridges, but to direct proper folding and interact with the retention machinery.

Conformation, structure, and oligomeric state of immunoglobulins not only control quality and functional properties of antibodies but are also critical for immunoglobulins secretion. Unassembled immunoglobulin heavy chains are retained intracellularly by delayed folding of the C H 1 domain and irreversible interaction of BiP with this domain. Here we show that the three C H 1 cysteines play a central role in immunoglobulin folding, assembly, and secretion. Remarkably, ablating all three C H 1 cysteines negates retention and enables BiP cycling and non-canonical folding and assembly. This phenomenon is explained by interdependent formation of intradomain and interchain disulfides, although both bonds are dispensable for secretion. Substituting Cys-195 prevents formation not only of the intradomain disulfide, but also of the interchain disulfide bond with light chain, BiP displacement, and secretion. Mutating the light chain-interacting Cys-128 hinders disulfide bonding of intradomain cysteines, allowing their opportunistic bonding with light chain, without hampering secretion. We propose that the role of C H 1 cysteines in immunoglobulin assembly and secretion is not simply to engage in disulfide bridges, but to direct proper folding and interact with the retention machinery.
Immunoglobulins (Igs) 1 are key molecules of the immune response, and, being oligomeric secretory proteins, Igs are excellent models to relate fidelity of folding and chain assembly to intracellular trafficking. Not only do conformation, structure, and oligomeric state control quality and functional properties of antibodies, but they are also critical for Ig secretion. Two heavy (HC) and two light (LC) chains form disulfide-bonded HC 2 LC 2 molecules. Each Ig chain consists of several similar compact domains, each comprising two twisted ␤ sheets stabilized by a single intradomain disulfide (1). IgG assembly is stabilized by an interchain disulfide between C L , the LC constant domain, and C H 1, the first HC constant domain (2). The C H 1 domain confers retention in the endoplasmic reticulum (ER) on unassembled HCs, and despite its high degree of similarity to other constant domains, only C H 1 is retarded in folding and is unable to cycle from the ER chaperone BiP (3). Only upon LC expression is BiP displaced from C H 1, which completes its folding, allowing assembly of HC 2 LC 2 and secretion (4 -7).
Although BiP binds transiently a wide variety of proteins (8), its associations with Ig HC and LC are among the best characterized. In vivo, BiP binds transiently to nascent LC via the unfolded V L domain and is released when this domain folds (9 -12). Non-secreted LC mutants display more persistent association with BiP (13)(14)(15)(16)(17)(18)(19). BiP also binds HC, where in addition to its interaction that seems irreversible with C H 1, it binds transiently other domains. In vitro studies defined specific heptameric sequences throughout LC and HC, which serve as BiP binding sites (18,20). The affinity for binding sites in C H 1 is not obviously different from that for other heptamers, so presumably, BiP binding sites in the incompletely folded C H 1 remain exposed, enabling continued interaction with BiP and ER retention (3,7,21). It has been suggested that C H 1 is unique in its propensity to expose BiP binding sites following C H 2 and C H 3 homodimerization, because C H 1 is the only constant domain unable to homodimerize (7,20).
Recognizing the unique folding status of C H 1, we undertook to identify structural elements within this domain, which control folding, chaperone interactions, and retention of unassembled HC. Several lines of evidence focused our attention on cysteines in C H 1. Contribution of exposed thiols to ER retention is well documented (22)(23)(24)(25)(26)(27). Thus, in the absence of LC, C H 1conferred retention could have resulted from continued exposure of the LC-interacting cysteine. However, even without such a cysteine, a truncated ␥1 HC containing only V H and C H 1 was still not secreted (21). Although this construct could no longer form the C L -C H 1 disulfide, its C H 1 oxidative folding, reflected by intradomain disulfide bond formation and consequent secretion, still required LC (21). Hence, possible contribution of the LC-interacting cysteine to C H 1-conferred retention should be in the context of the two additional C H 1 cysteines, which form the intradomain disulfide.
In this work we studied possible interrelations between all three C H 1 cysteines, and, as shown below, substituting each of them with serine, singly or in combinations, yielded surprising structure-function relationships with respect to folding, assembly, and secretion. Remarkably, a mutant chain lacking all three C H 1 cysteines was secretion-competent. We provide a mechanistic explanation for this phenomenon by showing that, although in C H 1 neither the intradomain nor the interchain native disulfide was essential for secretion, these bonds affect each other's formation. In addition to providing a molecular basis for the coordinated folding of C H 1 and its assembly with LC (7), we discuss a model in which C H 1 cysteines employ thiol-mediated retention, and their interplay maintains the stable BiP binding of unassembled Ig HC.

MATERIALS AND METHODS
Plasmid Construction-All mutant ␥ constructs were based on pJDEV NP C␥ 2b , containing genomic DNA encoding full-length murine ␥ HC with a known constant region (Ref. 28; accession number j00461). The C H 1 exon, flanked by intronic sequences, was excised with EcoRI and PvuII. A NotI-containing linker was ligated to the vector at the EcoRI and PvuII sites, generating the C H 1-less pJDEV NP C␥ 2b⌬ C H 1. Exons C H 1, C H 2, or C H 3 were PCR-amplified, using primers (Table I) and pJDEV NP C␥ 2b as a template, and subcloned into pBluescript (pBS, Stratagene) at EcoRI and XbaI sites, to generate pBS-C H 1, pBS-C H 2, and pBS-C H 3, respectively. Site-directed mutagenesis analyses of C H 1 were performed using pBS-C H 1 as a template and either forward mutagenic primer (Table I) or the complementary (reverse) mutagenic primer. The mutated C H 1 was cut by EcoRI and XbaI and inserted into pBS. Double and triple mutations in C H 1 were introduced successively. All DNA constructs were sequenced. Wild-type and mutated C H 1, as well as C H 2 and C H 3, were excised from the respective pBS constructs and re-introduced into pJDEV NP C␥ 2b⌬ C H 1 with EcoRI and NotI. Amplification and mutagenesis of I was performed by PCR (see primers in Table I) on pTM1-I cDNA (29) and cloned into pCDNA3 (InvitroGen) at the EcoRI and NotI sites.
Cell Culture and Transfection-COS-7 cells were grown in Dulbecco's modified Eagle's medium supplemented with 10% fetal calf serum, penicillin, and streptomycin at 37°C in humidified 5% CO 2 atmosphere. Cells (ϳ50% confluent) were transfected using the calcium phosphate method (30), with 5 g of the various ␥ constructs, with or without equimolar amounts of the pJDEI vector encoding genomic full-length murine I LC (14) or cDNA of wild-type or mutant . Where indicated, 8 g of constructs encoding HA-tagged wild-type or T19G BiP (31) was co-transfected. Cells were analyzed 42-48 h post-transfection.
Metabolic Labeling and Steady-state Analyses-Secretion was monitored by re-feeding cells with 1.5 ml of fresh medium that was collected after 3-5 h. For metabolic labeling, cells were starved for 1 h in methionine-deficient medium and then either pulse-labeled for 30 min with [ 35 S]methionine (250 Ci/ml) and chased in complete medium, or labeled for 3-4 h. In all experiments, cells and media were separated by centrifugation at 4°C, cells were washed with ice-cold phosphate-buffered saline and lysed (300 l/2 ϫ 10 6 cells) in ice-cold NET buffer (50 mM Tris-HCl (pH 7.5), 150 mM NaCl, 5 mM EDTA, 0.5% Nonidet P-40) supplemented with 2 mM phenylmethanesulfonyl fluoride (Sigma) and 100 units/ml aprotinin (Roche Applied Science). For samples analyzed by non-reducing SDS-PAGE, the phosphate-buffered saline and NET buffers included freshly added N-ethylmaleimide (20 mM, Sigma). Cell lysates were centrifuged (20,000 ϫ g, 20 min, 4°C), and supernatants were collected.
Immunoprecipitation, Treatment with Glycosidases, and Electrophoresis-Protein A-Sepharose (Repligen) was added to cell lysates or media to quantitatively precipitate ␥ and associated proteins. To promote BiP release, 10 mM Mg 2ϩ -ATP was added to the lysates, whereas otherwise apyrase (10 units/ml, Sigma) was included to prevent the release of BiP (3). Samples were rotated for 4 h at 4°C, washed three times with borate wash buffer (7), and once with phosphate-buffered saline. Where indicated, immunoprecipitated proteins were treated with endoglycosidase H (endo H) or peptide: N-glycosidase F (New England Biolabs), according to the manufacturer's instructions. Proteins were resolved by SDS-PAGE under reducing or non-reducing conditions and electroblotted onto nitrocellulose. For detection of metabolically labeled proteins, blots were autoradiographed and quantified by densitometry. Subsequently, blots were probed with the following antibodies: horseradish peroxidase (HRP)-conjugated anti-mouse ␥2b (SouthernBiotech), HRP-conjugated anti-mouse I (SouthernBiotech), biotin-conjugated anti-BiP (SouthernBiotech) followed by HRP-conjugated avidin (Jackson), anti-HA (clone 12CA5) followed by HRP-conjugated anti-mouse IgG (Jackson). The HRP was visualized by enhanced chemiluminescence (ECL).
Limited Proteolysis-Limited proteolysis of intracellular HCs was carried out on Protein A precipitates under optimized time, temperature, and enzyme concentrations. Papain (Sigma) was added at 5 g/ml to precipitates resuspended in 100 mM sodium acetate (pH 5.5)/50 mM cysteine/1 mM EDTA and incubated at 37°C. Precipitates resuspended in phosphate-buffered saline were either incubated at 30°C with 5 g/ml of trypsin (Difco) or at 37°C with 0.1 g/ml elastase (ICN).
Homology Modeling and Sequence Alignment-The potential BiP binding sites were based on alignment of high BiP-binding score heptapeptides (32) with C H 1 sequence of ␥2b (28). Structural alignment of C H 1 of ␥2b was carried out via the Swiss-Model web site and using as template the sequence of an identical C H 1 whose structure is available (Protein Data Bank ID: 1MAM (33)).

RESULTS
The coordination between C H 1 oxidative folding and its covalent assembly with C L (7) and the well documented contribution of exposed thiols to ER retention raised the possibility that the 3 C H 1 cysteines are key determinants in these processes. This hypothesis was tested by Cys-to-Ser substitutions of relevant cysteines, singly, or in combinations. We disrupted formation of either intradomain or interchain disulfides and investigated in vivo the interrelations between these bonds. Based on conserved cysteine positions in multiple Ig constant domains (1, 33), we anticipated that in C H 1 of mouse ␥2b HC (designated ␥), Cys-140 and Cys-195 formed the intradomain disulfide, whereas Cys-128 formed the interchain bond with LC (see Fig. 6, below). Each mutant was expressed with or without I LC (designated ) in non-lymphoid COS-7 cells, where its fate could be assayed in the absence of endogenous HC and LC.
We first validated COS-7 cells as a system to study the intracellular fate of ␥, based on retention of wild-type HC (WT; These experiments confirm and extend previous reports, that, despite their similar fold, neither C H 2 nor C H 3 can confer retention even when grafted in place of C H 1. Importantly, all secreted HC consisted mostly of ␥ 2 , indicating that C H 1 absence did not hamper HC homodimerization. Secretion of WT was restored in the presence of LC (Fig. 1, B and C). Because was not directly precipitated by Protein A (Ref. 12 and Fig. 1D), its co-precipitation reflected covalent assembly into the secreted species ␥, ␥ 2 , and ␥ 2 2 (Fig. 1C,  lanes 1, 13, and 25). In addition, escorted the ␥ 2 intermediate non-covalently, as shown by co-precipitation of monomeric Primers were designed based on the published genomic sequence of murine ␥2b (accession No. j00461). Designed restriction sites are italicized and underlined. Primers to amplify C H 1, C H 2, and C H 3 were designed to include the relevant exon flanked by 50 -100 bp of intronic sequences to ensure functional splicing. The reverse primer for C H 3 was designed such that the stop codon was skipped over and the C H 1 donor splice site was added. All primers were synthesized by Invitrogen.

Forward primer
Reverse primer FIG. 1. Secretion of the various HC mutants in the absence or presence of LC. COS-7 cells were transfected with an empty vector (mock) or with genomic constructs encoding the indicated HC mutants in the absence (-; A) or presence (ϩ; C) of -encoding genomic construct. Cells were metabolically labeled for 4h with [ 35 S]methionine, and ␥ was precipitated by Protein A-Sepharose (IP: Protein A) from the medium. Precipitated proteins were resolved by SDS-PAGE under non-reducing conditions (PAGE: NR) and electroblotted, and blots were autoradiographed ( 35 S; lanes 1-12). Blots were then probed with anti-␥2b antibody (IB: anti-␥; lanes 13-24) and re-probed with anti-antibody (IB: anti-; lanes [25][26][27][28][29][30][31][32][33][34][35][36]. The data are representative of seven independent experiments. B, for comparative secretion, the seven experiments were quantified by in addition to a 125kDa species (Fig. 1C, lanes 1 and 25). This conclusion was verified by size-exclusion chromatography, where all secreted WT species were ϳ150-kDa tetramers, regardless of their covalent assembly status (Supplemental Fig. S1). HCs lacking C H 1 (either deleted or replaced) did not interact with co-expressed (Fig. 1C, lanes 26 -28) nor was their secretion improved by (Fig. 1, A-C), corroborating their LC-independent secretion. The association of BiP with intracellular WT, but not when C H 1 was either deleted or replaced (Fig. 1E), confirmed that the C H 1-conferred retention was due to BiP binding. BiP was displaced from WT by co-expression of , and to a lesser extent by ATP (Fig. 1E, lanes 1-3; Fig. 1F). The only partial dissociation of HC and BiP by ATP (Fig. 1F) likely reflects their rebinding as nucleotide is degraded (see Ref. 18).
Ablation of All Three C H 1 Cysteines Negates C H 1-conferred Retention-Substituting each of the 3 C H 1 cysteines with serine, singly or in combinations, yielded surprising structurefunction relationships with respect to folding, assembly, and secretion. When expressed without LC, these mutants could be divided into two categories. The majority resembled WT and were retained intracellularly, whereas a few mutants were secreted, resembling HCs that lacked C H 1 altogether. The most remarkable example of a mutant that was secreted efficiently in the absence of LC was the triple mutant C128S/C140S/ C195S (Fig. 1A, lanes 11 and 23; Fig. 1B). In this mutant all 3 C H 1 cysteines, which normally stabilize the conserved Ig fold, were ablated. Moreover, C128S/C140S/C195S did not interact with LC, even non-covalently (Fig. 1C, lane 35), and neither its interaction with BiP ( Fig. 1F) nor its secretion (lanes 11 and 23 in Fig. 1, A and C; Fig. 1B) were affected by LC.

Mutation of Cys-195 Hampers Interchain Disulfide Formation, Causes Misfolding, and Prevents Secretion-
The secretion of C128S/C140S/C195S may appear as a hierarchical contribu-tion of exposed thiols to HC retention. As shown in Fig. 1A and quantified in Fig. 1B, when all three cysteines were present, HC retention was complete (lanes 1 and 13); substituting Cys-128 (C128S) enabled only slight LC-independent secretion (lanes 5 and 17); substituting both Cys-128 and Cys-140 (C128S/C140S) led to more significant secretion (lanes 8 and 20), and ablating all 3 C H 1 cysteines resulted in efficient LCindependent secretion (lanes 11 and 23). However, this apparent hierarchy did not apply to every cysteine in C H 1. Disruption of the C H 1 intradomain disulfide by substituting both Cys-140 and Cys-195 did not alleviate retention (Fig. 1B). Moreover, co-expression of LC allowed us to use the ability of the various mutants to assemble with , dissociate from BiP, and be secreted as a criterion to further divide the non-secreted HCs into two categories: the properly folded ones that resembled WT and were secreted, and the misfolded ones that were not secreted even in the presence of LC. In this respect, the C H 1 domain of the C140S/C195S was severely misfolded, because this mutant was unable to assemble with , either covalently or non-covalently (Fig. 1C, lane 34), even though Cys-128 was intact (see Fig. 6).
We next asked whether disrupting the intradomain disulfide by singly mutating Cys-140 or Cys-195 was equivalent to the pairwise substitution in C140S/C195S. Evidently, a mutant lacking only Cys-195 resembled the double mutant C140S/ C195S, because C195S was neither secreted (Fig. 1B) nor assembled with (Fig. 1C, lane 31), and what appeared as LCfacilitated secretion was negligible and hardly contained ␥-only species not associated with LC (Fig. 1, B and C (lanes 7, 19, and  31)). Intracellular C195S seemed to be as misfolded as C140S/ C195S, because even in the presence of it associated with BiP in an ATP-dependent manner (Fig. 1F) and did not bind LC (Fig. 1C, lane 31) even though Cys-128 was intact. Indeed, ablation of Cys-128 in addition to Cys-195 (C128S/C195S) resulted in a phenotype very similar to that of C195S (Fig. 1, A-C; see also Fig. 6).
When the intradomain disulfide bond was disrupted by mutating Cys-140 alone, the resulting C140S was as efficiently retained as C195S or C140S/C195S (Fig. 1, A (lanes 6 and 18) and B). However, unlike these mutants, C140S responded to the presence of LC like WT. C140S assembled reasonably efficiently into ␥ 2 and ␥ 2 2 (Fig. 1C, lanes 18 and 30), dissociated from BiP (Fig. 1F) and was secreted (Fig. 1, B and C (lanes 6,  18, and 30)). We conclude that a C H 1 domain lacking Cys-140, one of the two cysteines that make up the native intradomain disulfide bond, is not misfolded, whereas C H 1 lacking its partner Cys-195 is (see Fig. 6).
Mutation of Cys-128 Retards C H 1 Oxidation and Allows Opportunistic Interchain Disulfides without Hampering Secretion-The failure to form the interchain disulfide in the absence of Cys-195, which forms the intradomain disulfide, suggested interrelations between these two bonds. We therefore performed a reciprocal experiment and disrupted the indensitometry, monitoring ␥ secreted in the absence (white bars) or presence (black bars) of LC. The calculated secretion is relative to 100% set for secretion of ⌬C H 1 and is presented as averages Ϯ S.D. Note the comparable levels of intracellular ␥ (Figs. 1E, 2A, 3A, and 5B). D, COS-7 cells were transfected with an empty vector (Ϫ) or with an excess (30 g of DNA) of genomic constructs encoding (ϩ). Cells were incubated for 4 h in fresh medium, and proteins were precipitated from identical samples of medium (m) and lysed cells (c) by either goat antiantibody followed by Sepharose-conjugated mouse anti-goat antibody (anti-) or by Protein A-Sepharose (PA). Precipitated proteins were resolved by SDS-PAGE under non-reducing conditions (PAGE: NR) and electroblotted, and blots were probed with an antiantibody (IB: anti-). E, COS-7 cells transfected with an empty vector (mock) or with constructs encoding WT, ⌬C H 1, (C H 2) 2 or (C H 3) 2 , with (ϩ) or without (Ϫ) -encoding construct, were labeled for 4 h with [ 35 S]methionine, ␥ was precipitated by Protein A-Sepharose (IP: Protein A) from lysed cells, and BiP was co-precipitated. Precipitated proteins were resolved by SDS-PAGE under reducing (PAGE: R) conditions and electroblotted, and blots were autoradiographed ( 35 S). Where specified, ATP was added to lysates. F, for all ␥ mutants, bands corresponding to intracellular ␥ and co-precipitated BiP detected by autoradiography (see Fig. 1E for WT) were quantified by densitometry and BiP/␥ ratio was calculated relative to the value 1 set for cells expressing the respective HC alone (see Fig. 1E, lane 1 for WT). BiP/␥ ratio from four independent experiments is presented as averages Ϯ S.D. for lysates incubated with ATP prior to precipitation (white bars) or for lysates from cells expressing ␥ϩ (black bars). ␥, HC monomers; ␥ 2 , homodimers; ␥, ␥ 2 , ␥ 2 2 , assembly species; , LC monomers precipitated by anti-or co-precipitated by protein A as non-covalently bound LC; 2 , LC homodimers; BiP, co-precipitated BiP.
terchain disulfide by substituting Cys-128. Without LC, C128S was secreted only marginally (5% of ⌬C H 1 secretion; Fig. 1B). However, C128S was not nearly as misfolded as C195S, because it resembled WT (Fig. 3A). In the presence of LC, C128S dissociated from BiP (Fig. 1F) and was secreted efficiently (Figs. 1B, 1C (lanes 5, 17, and 29), and 3A). As expected, the major secreted covalent form of C128S was ␥ 2 rather then ␥ 2 or ␥ 2 2 (Figs. 1C (lanes 17 and 29) and 3A). Because WT ␥ 2 was hardly secreted (Fig. 3A, lane 4), it suggested that when a LC-interacting cysteine was absent, the secreted ␥ 2 species of C128S was associated non-covalently with two LCs, as corroborated by the abundant free LC co-precipitated by Protein A (Fig. 3A, lane 16). Interestingly, a -containing species of ϳ47kDa, which was not formed when was expressed alone (Fig. 1D), was also co-precipitated from cells and media of C128S-expressing cells (Fig. 3A, lanes 15 and 16). This species is most likely disulfide-bonded 2 homodimers that are formed only when Cys-128 is absent, as discussed elsewhere. 3 A minor, yet significant, fraction of C128S was surprisingly assembled into covalent ␥ 2 2 and ␥ 2 , as revealed by probing with anti-␥ (Fig. 3A, lanes 7 and 8) and anti- (Fig. 3A, lanes 15  and 16) antibodies. Because LC was not disulfide-bonded to HCs that lacked C H 1 (see Fig. 1C), this result indicated that, when Cys-128 was absent, C H 1 cysteines that usually formed the intradomain disulfide could form opportunistic interchain disulfides with , precluding intradomain disulfide between Cys-140 and Cys-195. Indeed, consistent with altered oxidative folding (34), all species containing covalently linked C128S and (␥ 2 2 , ␥ 2 , and ␥) exhibited altered electrophoretic migration in nonreducing gels (Fig. 3A, lanes 4, 8, 12, and 16). Formation of opportunistic interchain disulfides is probably allowed by inefficient engagement of Cys-140 with Cys-195 when Cys-128 is absent, as shown for a small fraction of intracellular HC monomers. These assembly species, which have not yet covalently bound to either or ␥, could be resolved into oxidized and unoxidized forms (Fig. 3A, lanes 1-8, upper and lower panels). Whereas WT monomers were fully oxidized (lane 3), C128S monomers appeared to be mostly reduced (lane 7), yet only the oxidized monomers were secreted (lane 8). That the distorted and retarded oxidative folding was due to the mutant C128S HC and not to its opportunistically bonded LC was indicated by substituting Cys-214 in , the residue that interacts with HC (Fig. 3B). This mutation abolished not only the normal disulfide with WT, but also the opportunistic interchain disulfides with C128S. Nonetheless, WT HC monomers remained oxidized and C128S monomers remained reduced (Fig. 3B, lanes 1-12, upper and lower panels), indicating that only Cys-214 could engage a HC cysteine, either the native Cys-128 or an opportunistic partner (see Fig. 6).
Taken together, this detailed analysis of single, double, and triple cysteine substitutions shows that the effect on folding and secretion of mutating one cysteine is not equivalent to mutating the other, even when the effect on disulfide bond formation is the same. Hence, when either interchain or intradomain disulfide bond in C H 1 is disrupted, formation of the other disulfide bond is affected, and thus formation of these two disulfide bonds in C H 1 is interdependent.
The C128S/C140S/C195S Heavy Chain Exhibits Unusual Folding and Assembly-The triple mutant C128S/C140S/ C195S was secreted independently of LC and in addition ex- hibited a unique assembly pattern. First, the secreted molecules were mostly typical 100-kDa ␥ 2 homodimers, indicating that this Cys-less C H 1 did not interfere with HC homodimerization. Yet, a significant fraction migrated as an unusual ϳ200-kDa species (Fig. 1A, lanes 11 and 23), which contained HC exclusively, as even in cells co-expressing , they were reactive only with anti-␥ but not with anti-antibodies (Fig.  1C, lanes 11, 23, and 35). Excising both assembly species from non-reducing gels and individually subjecting them to reducing SDS-PAGE resolved the 100-kDa species into the expected 50-kDa ␥ monomers, whereas the 200-kDa species gave rise to a 100-kDa species (Fig. 4A). The latter probably represented atypical dimers (␥ 2 *) held together via SDS-resistant interactions. Two such ␥ 2 * molecules appeared to undergo conventional homodimerization into 200-kDa ␥ 4 * tetramers (see Fig. 6).
The secretion of the Cys-less C H 1 mutant suggested that its C H 1 domain folded into a distinct structure that passed the scrutiny of the cellular quality control. To test this possibility, intracellular HCs of the secretion-competent C128S/ C140S/C195S and secretion-incompetent WT were each subjected to limited proteolysis with elastase or trypsin. Partial elastase digestion yielded a ϳ85-kDa intermediate only from C128S/C140S/C195S (Fig. 4B, lane 5, arrow). An additional ϳ70-kDa elastase product was generated from both WT and C128S/C140S/C195S, but was more pronounced in C128S/ C140S/C195S (lane 5) than in WT (lane 2). Limited proteolysis with trypsin also revealed differences between intracellular C128S/C140S/C195S and WT. Although two major fragments, ranging between 80 and 90 kDa, were generated from the mutant (Fig. 4C, lanes 8 -12, arrows), only a single  [13][14][15][16][17][18][19][20][21][22][23][24] antibodies. The lower panels represent longer exposures of the monomeric ␥ section (␥ ox , ␥ red ) from blots probed with the anti-␥2b antibody. The remnants of ␥ 2 (panel B, lanes 16 -24) probably reflect incomplete stripping from the abundant ␥ 2 prior to re-probing, because antidoes not recognize ␥ when used first. The data are representative of four independent experiments. ␥, HC monomers; ␥ ox , ␥ red , monomers with oxidized or reduced C H 1; ␥ 2 , homodimers; ␥, ␥ 2 , ␥ 2 2 , assembly species; , co-precipitated non-covalently bound LC; arrows, the altered migration of C128S-containing ␥ 2 and ␥ 2 2 . major fragment was generated from WT (lanes 2-6). To address the conformation of the secreted C128S/C140S/C195S, it was compared with the secreted (C H 3) 2 , because both HCs could be collected from the medium independently of LC expression. Papain digestion yielded a 50-kDa anti-␥-reactive product in both (C H 3) 2 and C128S/C140S/C195S, but only C128S/C140S/C195S gave rise to a predominant ϳ70-kDa product (Fig. 4D, lanes 8 -12, arrow). The altered proteolytic sensitivity demonstrated that the fold of intracellular C128S/ C140S/C195S was distinct from that of intracellular WT, and the fold of secreted C128S/C140S/C195S was distinct from that of secreted (C H 3) 2 . An alternative interpretation to the altered proteolytic sensitivity is that C128S/C140S/C195S could interact with a unique set of auxiliary proteins. Either way, the data points to an alternate conformation of the Cys-less C H 1 mutant that is correlated with unusual assembly into ␥ 4 * and is evidently compatible with secretion of HCs that are not assembled with LC.
BiP Cycles from the Cys-less C H 1 to Allow Secretion of C128S/ C140S/C195S-As established previously and corroborated in this study, C H 1-conferred retention of HC is mediated (at least in part) by BiP. Yet, none of the 3 C H 1 cysteines was included in the two identified clusters of potential BiP binding sites (Fig.  5A), obtained by multiple alignment of ␥2b C H 1 sequence (28) with likely BiP-binding heptapeptides (32). Therefore, we investigated whether and how BiP-mediated retention operated on the Cys-less C H 1. For WT, the relatively constant levels of co-precipitated BiP during the chase, detected either radioactively or with an anti-BiP antibody (Fig. 5B, middle or upper panels, respectively, lanes 8 -11), presumably reflected recycling of long-lived BiP on and off the nascent HC. Consistent with Vanhove et al. (3), the diminished levels of co-precipitated BiP in the presence of (Fig. 5B, lanes 8 and 9; see also Fig. 1E,  lanes 1 and 3) were interpreted as a transition from an irreversible to a reversible mode of HC binding to BiP. Interestingly, BiP association with C128S/C140S/C195S in the absence of LC resembled BiP association with WT in the presence of LC (Fig. 5B, lanes 1-4 and 9 -11), suggesting that, by ablating all its cysteines, C H 1 binding to BiP became reversible. This was supported by overexpression of an ATPase-defective BiP mutant (T19G), known for prolonged association with its substrates, which inhibits their secretion (10). As shown in Fig. 5C, this mutant BiP decreased the secretion of C128S/C140S/ C195S by more than 1.6-fold, from 56% to only 34% secreted C128S/C140S/C195S. Apparently, under these conditions, ␥ 2 * formation was also hindered, resulting in secretion of a smaller fraction of ␥ 2 * (Fig. 5C), suggesting that BiP displacement, a prerequisite for native Ig assembly, was also a prerequisite for the atypical assembly of Cys-less C H 1 mutant into ␥ 2 *. DISCUSSION The ER retention of unassembled HC is attributed to inherently retarded folding of C H 1 and its irreversible binding to BiP (3). However, what sets C H 1 apart from the structurally similar C H 2 and C H 3 is not fully understood. This work provides three important and surprising findings that help explain the unique nature of C H 1. First, the retarded folding is due to delayed oxidation of the 3 C H 1 cysteines, because formation of both intradomain and interchain disulfide bonds is interdependent. Second, formation of neither disulfide is obligatory for secretion. Third, the retention can be overcome if all 3 C H 1 cysteines are substituted, resulting in efficient LC-independent secretion of a HC harboring a full-length C H 1. These findings demonstrate a link between two well established quality control mechanisms, thiol-mediated and BiP-mediated retention. Furthermore, they highlight the plasticity of the Ig fold.
Cysteine residues can affect Ig folding, assembly, and secretion in a number of ways. Often they participate in disulfide bonds, a characteristic of Ig superfamily proteins. Usually, disulfides serve to stabilize the fold rather then to catalyze its formation and can be replaced if the loss of folding stability is at least partially compensated for by other stabilizing mutations (35,36). In most Ig domains, a disulfide links two ␤ sheets and is buried in the core of the domain, with conserved, hydrophobic side chains packed against it. If this intradomain disulfide fails to form, exposed thiols in this hydrophobic core adversely affect folding, leading to continued exposure of BiP binding peptides and consequent retention (18). This can clearly be the explanation for the compromised folding of all mutants lacking Cys-195, which cannot be rescued by LC expression. However, the non-reciprocal nature of the mutations described here indicates that C H 1 cysteines must also have other effects. Cysteine substitutions other than Cys-195, that disrupt the same intradomain disulfide (C140S; C128S when opportunistically engaged with ; Fig. 6), can be rescued by LC, at least partially, as judged by BiP displacement and consequent secretion. Moreover, leaving a surface-exposed cysteine such as Cys-128 reduced, or replacing it with serine, would not normally be expected to have a dramatic destabilizing effect. Nonetheless, C128S displays defects in oxidative folding that are manifested even when LC is present, contrary to expectation. It seems, therefore, that Cys-128 is essential for C H 1 folding and not just for linking C H 1 to LC. A third effect of cysteines may be via thiol-mediated retention (23,25), where unpaired cysteines can interact with ER proteins (37)(38)(39). This mechanism may operate in WT retention, where cysteines and Cys-128 in particular are unpaired, or in some of the single mutants that form opportunistic disulfides and have promiscuously exposed thiols.
We propose that C H 1 is unique because its folding requires interdependent and concurrent formation of intradomain and interchain disulfides. This is in contrast to other Ig domains with multiple cysteines, such as C L , where the intradomain disulfide forms with high fidelity, leaving unpaired the third cysteine responsible for the interchain disulfide. The interdependent oxidation of C H 1 is shown in a number of ways. Cys-128 fails to form interchain disulfide with LC when intradomain disulfide cannot form due to Cys-195 substitution. Reciprocally, in HC whose LC-interacting Cys-128 is absent, intradomain cysteines oxidize inefficiently, even allowing opportunistic engagement of Cys-140 or Cys-195 with LC. Importantly, this inefficient oxidation is inherent to HC and does not occur in LC. Therefore, one possible interpretation is that as long as a fourth cysteine provided by LC is missing, the 3 C H 1 cysteines cannot adopt the proper configuration to form intradomain disulfide and HC cannot dissociate from BiP. However, the finding that LC that lacks this cysteine can still promote folding and secretion of WT HC (as well as of some mutants) demonstrates that the role of C H 1 cysteines in Ig folding, assembly, and secretion is not simply to engage in disulfide bridges, but rather to affect C H 1 folding, including mediating possible interactions with thiol-sensing ER proteins. When LC is provided, a cascade of interdependent folding and assembly events juxtaposes the two pairs of cysteines. Once the positioning of interacting surfaces is accomplished, disulfides form, but only to stabilize the Ig fold. BiP displacement and subsequent secretion are achieved concurrently with proper C H 1 folding and this can be accomplished in the presence of C L as a folding template.
A mechanistic link between BiP-mediated and thiol-mediated retention of HC is quite probable. The capacity of BiP to FIG. 6. A model of IgG2b and the functions of C H 1 cysteine. ␥2b HC in IgG2b and C H 1 in the detailed C H 1-C L models of the various mutants are represented by black lines. I LC in IgG2b and C L in the detailed C H 1-C L models are represented by gray lines. Cys-128 is represented by light gray circles, Cys-140 by dark gray circles, Cys-195 by black circles, and Cys-214 of I by dark gray circles. In IgG2b these cysteines are also indicated by numbers, and all other cysteines are represented by open circles. In the detailed C H 1-C L models, substituted cysteines are represented by open circles, and C L that fails to associate with C H 1 is indicated by dotted lines. Disulfides (S-S) that fail to form are encircled by dotted ellipses and opportunistic disulfides of C128S are formed with either Cys-140 or Cys-195. In tetrameric C128S/C140S/ C195S typical homodimers formed by C H 2 and C H 3 interactions and disulfides in the hinge region are either gray or black, and 2 Cys-less C H 1 domains (heavy lines) exhibit a unique association.