How J-chain ensures the assembly of immunoglobulin IgM pentamers

We first analyzed JC insertion in IgM in murine myeloma transfectants that express it alone or with Ig λ light (L) and/or secretory µ chains (Fig. 1C). In J[µ1]L cells (Sitia et al, 1987, 1990), IgM secreting transfectants of the J558L myeloma that express B1-8-derived μ chain as well as endogenous L and J chains, µ and L chains efficiently assembled into intermediate complexes (µL and µ₂L₂) and secretion-competent polymers recognized by anti-µ and anti-L antibodies (lanes 5 and 9, Pols). Only the latter were decorated by anti-JC (lane 1). In an NS0 myeloma transfectant expressing µ chain as well as endogenous JC, anti-µ antibodies decorated two sharp bands corresponding to µ and µ₂ dimers, and a smear of higher molecular mass bands (lane 7). Also, anti-JC reacted with this smear (lane 3), suggesting that it likely consisted of heterogeneous, covalent assemblies of JC and µ chains (and possibly other molecules). The smear might reflect attempts to form stable polymers, doomed to fail in the absence of L chains (Hendershot et al, 1987; Sitia et al, 1990). In all cell types, including J558 cells (Oi et al, 1983) expressing J and L (lane 2) and NS0 (Galfrè and Milstein, 1981) cells expressing only JC (lane 4), anti-JC stained monomeric JC, and IgM polymers in J[µ1]L.

Pulse-chase experiments with ³⁵S-labeled aminoacids showed that N[µ1] cells lacking L degrade JC after an initial lag (Fig. 1D), confirming previous results (Fagioli and Sitia, 2001). Interestingly, at all time points tested, monomeric JC migrated with a more diffuse pattern in nonreducing conditions (left panel) than in reducing conditions (right panel). JC collapsed into a sharp band upon DTT reduction, suggesting that JC may adopt multiple redox isoforms through the different arrangements of its intra-chain disulfide bridges, the nature of which is of particular interest in this study.

To simplify the experimental system and bypass the requirement for L chain assembly, we exploited a Halo-tagged truncated µs variant lacking the variable and first constant domains (H-Cµ234tp) (Giannone et al, 2022). This construct and a variant lacking C575 were expressed in HEK293T cells in the presence or absence of JC (Fig. 2A). Cell lysates were resolved by nonreducing SDS–PAGE, and blots were first stained for the Halo ligand and then decorated with anti-JC antibodies to identify the components of the various assemblies. H-Cµ234tp chains polymerized efficiently in HEK293T, yielding a ladder with growing numbers of subunits, from H-Cµ234tp₂ ‘monomers’ to (H-Cµ234tp₂)₆ “hexamers”. The slowest migrating band, consisting of secretion-competent hexamers (lane 1, indicated as 6X), was no longer detectable in cells co-expressing JC (lane 2, left panel). Anti-J antibodies decorated mainly two high molecular assemblies (lane 2, right panel). Based on its size, immunoreactivity, and fate, the most intense of the two likely consists of mature, secretion-competent JC-containing pentamers (dark blue asterisk in panel A). Noteworthy, lower order H-Cµ234tp assembly intermediates contained only traces of JC (light blue asterisks), indicating that also in non-lymphoid cells, JC forms stable covalent bonds mainly with pentamers. As expected, JC did not bind a H-Cµ234tp mutant lacking C575 (Fig. 2A, lanes 3-4, right panel). Very similar results were obtained with myc-tagged JC (Fig. EV1A).

Taken together, the above results suggested that JC covalently binds to nascent pentamers, outcompeting a sixth subunit. Alternatively, JC could bind to intermediates of IgM, subsequently recruiting other subunits to complete a hexameric structure. If this were the case, we would observe a change in the assembly pattern of IgM over the course of polymerization when JC is present compared to when it is not. This could eventually lead to the formation of covalent intermediates containing JC over time. To address this issue, we performed time-course experiments exploiting the covalent binding of the Halo-tag to its ligands (Tempio et al, 2021). Cells expressing H-Cµ234tp with or without JC were incubated with excess green R110 ligand to label preexisting molecules, washed, and cultured in the presence of red TMR ligand to follow the fate of newly made proteins (see scheme in Fig. EV1B). As expected, molecules present at time 0 were only green (compare Figs. 2B and  EV1C): newly made, TMR-labeled H-Cµ234tp chains increased with time (Fig. 2B). Confirming our previous data, JC prevented the formation of hexamers, was mainly detected in high molecular weight polymers and bound H-Cµ234tp assembly intermediates very poorly. Clearly, JC had minor, if any, effects on the formation of dimers of H-Cµ234tp dimers or other intermediates, confirming that JC binds preferentially to nascent pentamers, outrunning H-Cµ234tp2 or µ₂L₂ subunits in completing a hexameric structure (Fig. 2C). Taken together, the above findings suggested that features of pentameric IgM promote stable interactions with JC.

Next, we reconstituted the process in the test tube using purified JC and the µ chain C-terminal domain, including the tailpiece (Cµ4tp), previously shown to be the minimal unit forming hexamers in vitro (Pasalic et al, 2017). In Cµ4tp, only C575 in the µtp is available for intermolecular bonds. Thus, no covalent polymers but only Cμ4tp₂ species are detectable in nonreducing SDS–PAGE (Pasalic et al, 2017). However, non-covalent assemblies of six Cμ4tp₂ dimers ((Cμ4tp₂)₆) formed in vitro can be readily detected by analytical size exclusion chromatography (SEC). (Cμ4tp₂)₆ eluted as a sharp peak (Fig. 3A, peak 1). Additionally, some Cμ4tp oxidation side products resulting in larger covalent oligomers were detected (Fig. EV2A, black insert boxes). Importantly, the addition of JC to the reaction caused a dramatic assembly shift: (Cμ4tp₂)₆ were no longer formed, but slower eluting species appeared whose size and composition are consistent with (Cµ4tp₂)₅ pentamers bound to a single JC (Fig. 3A peak 2). Titration experiments (Fig. EV2B) revealed that JC induced pentamer formation already at stoichiometric concentrations (0.1 mg/mL). JC caused the formation of side products (peak marked with *) if present in excess (>0.1 mg/mL). Assemblies with aberrant stoichiometries may form at the high concentrations reached in test tubes or in non-lymphoid transfectants. For further experiments, we chose ratios of Cµ4tp and JC, resulting in the highest amount of (Cµ4tp₂)₅-JC, while keeping amounts of hexamer and shoulder side products as low as possible (Fig. EV2B). JC alone eluted in a broad peak indicative of an ensemble of conformers (Fig. EV2C), consistent with its behavior in cells (Fig. 1D). SDS–PAGE analyses of individual fractions obtained after preparative HPLC-SE (Fig. EV2A) detected exclusively Cµ4tp₂ dimers in the hexamer fractions (1–2), while the pentamer fractions (3–5) contained a species with the mobility of a complex consisting of a JC linked to two Cµ4tp subunits (Fig. EV2A, lanes 3–5, red box). Reduction of all fractions (Fig. EV2D, top and bottom panel) resulted in Cμ4tp monomers. As expected, JC was found in fractions 3–5 (pentamers), fractions 5–8 (* side products), and fraction 9 (unbound JC co-eluting with Cµ4tp dimers). Taken together, these results established Cµ4tp and JC as the minimal requirements for pentamer formation. They also recapitulated the results obtained in living cells, where JC shifted IgM polymerization toward pentamers at the expense of hexamers, confirming that JC is responsible for pentamer formation. Kinetic experiments (Fig. 3B–D) revealed that the presence of JC accelerated the formation of IgM complexes in vitro (Fig. 3B). In the absence of JC, Cµ4tp monomers transition to dimers and then to hexamers (Fig. 3C). Addition of JC (Fig. 3D) did not cause the accumulation of Cµ4tp-JC intermediates, confirming the results obtained in cells (Fig. 2A,B). These data highlight the striking similarity of polymerization in vivo and in vitro, confirming that JC selectively binds to oligomeric IgM. Interestingly, when added at the final stage of Cµ4tp assembly, JC did not significantly decrease the amount of preformed (Cµ4tp₂)₆ hexamers (Fig. EV2E), emphasizing the importance of timing in IgM assembly. The few pentamers detected at the end of the incubation were most probably formed by the remaining unassembled Cμ4tp dimers. Thus, the interaction of JC with nascent IgM complexes is timed to occur specifically before hexamer formation.

Next, we performed limited proteolysis experiments (Fig. 3E,F) to determine whether JC integration into IgM assemblies is accompanied by conformational changes that result in partial protection against proteolytic cleavage. To this end, Cµ4tp monomers and JC as well as HPLC-SE fractions containing hexamers or pentamers (fractions 2 and 4 from Fig. EV2A, respectively) were incubated with proteinase K (PK) for different times and the effect of proteolysis was analyzed by SDS-PAGE. Confirming previous results in living cells (Mancini et al, 2000), JC was rapidly and completely digested (Fig. 3E, see asterisks). Similarly, Cµ4tp monomers were cleaved, yielding a resistant fragment. In contrast, Cµ4tp hexamers (peak 1) displayed a higher PK resistance than the monomers (Fig. 3F). Cµ4tp₂, derived from non-covalent hexamers, were still detectable after 90 min of incubation with PK, and slowly converted into monomers thereafter. A similar effect was observed in the presence of JC (Fig. 3F). Cµ4tp₂ and Cµ4tp₂JC, the additional species present in the pentamer (Fig. EV2A, red box), were slowly processed by PK. These data indicated that JC is vulnerable to PK before assembly but becomes protected as part of the ß-sheet core that stabilizes IgM polymers (Fig. 3G).

To gain insight into the structure and disulfide bonding of recombinant JC, we performed circular dichroism (CD) and cross-linking mass spectrometry (MS). The CD spectrum of JC is consistent with that of a largely unstructured protein (Fig. 4A). Analysis by cross-linking MS yielded a 100% peptide coverage for JC and revealed a mixture of non-native and native disulfide bonds (Fig. 4B). A few cysteines were shown to be involved in homomultimeric links, which proves the presence of oligomeric JCs. Thus, it cannot be excluded that a portion of the remaining identified bonds stems from multimeric JC. However, the amount of multimeric JC species should be small compared to the monomer as SE-HPLC analysis of JC (Fig. EV2C) clearly shows that the shoulder of higher molecular weight species is by far less intense than the broad monomeric JC peak. Similarly, nonreducing SDS-PAGE shows mostly monomeric species of JC (Appendix Fig. S2B). Interestingly, the numbers of peptide-to-spectrum matches (PSM counts) (Fig. 4C) are the highest for the crosslink between Cys92 with Cys102, which could be due to susceptibility of this site to protease cleavage. These results are consistent with their smeary appearance in nonreducing SDS gels (Fig. 1D, lane 5 compare top and bottom panel). Together with the results obtained in cells (Mancini et al, 2000), the above in vitro findings confirm that, on its own, JC does not adopt a unique structure with a defined disulfide bond pattern.

In living cells, ERp44 patrols IgM hexamerization, recognizing non-native disulfides (Anelli et al, 2007; Giannone et al, 2022). To test whether ERp44 controls JC assembly into IgM pentamers as well, we reconstituted this process in ERp44^KO HeLa cells (Fig. 5). Unexpectedly, JC was secreted by these cells. Under nonreducing conditions, secreted JC migrated as a continuous smear (Fig. 5A, lanes 4 and 7), which collapsed into a single band upon reduction with DTT (Fig. EV3A), suggesting that it consists of an ensemble of JC molecules containing a variety of intra- and inter-chain disulfide bonds. The expression of wild-type Halo tagged ERp44 (H-ERp44), but not of a mutant lacking the RDEL motif, which is abundantly secreted (Fig. 5C), efficiently restored JC retention (Fig. 5A, compare lanes 4-5-6). Upon expression of WT ERp44, JC accumulated in the lysates of ERp44^KO HeLa, concentrating in numerous bands under nonreducing conditions (Fig. 5B, lanes 5 and 8). Most of these bands, but not all, contained ERp44 (Fig. 5D, see asterisks). Thus, ERp44 prevents JC secretion and induces its binding to other proteins, that are presently not identified. ERp44 bound JC via C29 in its active CRFS site: upon C29S mutation, JC no longer formed complexes with ERp44 (Fig. EV3B). As expected, the secretion of unpolymerized IgM subunits by ERp44^KO HeLa was prevented by ERp44 overexpression. In addition, fewer JC-containing pentamers were secreted (Fig. EV3C,D), consistent with the role of ERp44 in IgM-JC quality control (Giannone et al, 2022; Anelli et al, 2003).

Taken together, these findings indicate that JC can escape primary quality control in the ER of HeLa cells and form an ensemble of variants with different disulfide bonding patterns. ERp44 attacks non-native disulfide bonds surveying and rectifying IgM assembly.

The finding that JC stably binds only pentamers raised questions as to how JC discriminates between these and smaller assemblies, and what triggers JC insertion. Hints came from the structural asymmetry and special arrangement of the five μ₂L₂ subunits in the cryo-EM structure of IgM pentamers (Li et al, 2020a; Lyu et al, 2023). Curiously, the µ₂L₂ subunit opposite to the JC (subunit 1 in Fig. 1A; Appendix Fig. S3A) displays its two µtps arranged in antiparallel orientations, each belonging to one distinct β-sheet of the core sandwich structure. In subunits 2–5, instead, the µtps are arranged in a parallel fashion and are part of either the top or the bottom β-sheet. In mature IgM pentamers (Li et al, 2020a), JC caps the ß-sandwich structure formed by the µtps in growing pentamers, adopting an arrangement that mimics subunit 1, with two β-strands (β3 and β4), pointing in opposite directions, that are interacting with the µtps of subunits 4 and 5 (Appendix Fig. S3A). We noted that the exposed hydrophobic core of the µtps in the IgM pentamer includes residues V564, L566, and M568 from subunits 4 and 5 (Fig. 6A; Appendix Fig. S3B). JC binding at this site leads to the extension of the β-sheet hydrogen bonding pattern, and the formation of hydrophobic contacts with several JC residues. Three highly conserved residues, namely I40 from JC strand β3 and F61 and Y63 from strand β4, which are located centrally in the interaction surface, display shape complementarity to form a tightly packed hydrophobic core (Fig. 6A). Therefore, I40, F61 and Y63 might be important in the encounter complex between JC and the core of nascent pentameric IgM. To test this, we co-expressed JC mutants in which these residues were mutated to alanine (I40A, F61A and Y63A, Fig. 6B–F) or serine (I40S, F61S and Y63S, Fig. EV4A,B) with H-Cµ234tp in HEK293T cells (Figs. 6B,C and EV4A,B) or in HeLa-ERp44^KO cells (Figs. 6D,E and EV4C,D). JC binding to IgM complexes was monitored by the ability to prevent hexamer formation. IgM assembly in the presence of the three JC mutants generated pentamers which were secreted poorly, although for different reasons. Mutating I40 prevented JC insertion (Fig. 6B,C, compare lanes 7-8). The phenotype was more pronounced for I40S than for I40A, in line with the notion that hydrophobic packing stabilizes the core structure (Fig. EV4A,B). Conversely, the F61A and Y63A mutants formed abundant polymers (Fig. 6C, lanes 4-5), which, however, were largely retained by HEK293 cells (Fig. 6C, lanes 9-10). The failure of JC hydrophobic mutants to correctly bind pentameric IgM leads to the secretion of hexamers only (Figs. 6B and EV4A, lanes 9-10). However, pentamers containing JC hydrophobic mutants were secreted by HeLa-ERp44^KO cells (Fig. 6D,E, lanes 3, 4, and 5). Retention was rescued by ERp44 expression (Figs. 6D,E, lanes 7-8-9 and  EV4C,D), confirming its pivotal role in IgM biogenesis. These results point to stringent cellular quality control by ERp44 on these misassembled polymeric IgM.

In vitro, SEC analyses of interaction yielded results comparable to those obtained in cells (Fig. 6F). The I40A JC mutant was unable to shift the (Cµ4tp₂)₆ hexamer towards pentamer formation. A small hexamer to pentamer shift was induced by the F61A and Y63A mutants: however, both mutants formed wider peaks with earlier elution times than those containing WT JC, suggesting altered structures (Fig. 6F). These results strengthen the observation made in living cells, in which F61 and Y63 containing polymers were formed but retained by ERp44. Taken together, these findings suggest that I40, F61 and Y63 are essential for the initial and correct interactions of JC with the core of the nascent pentameric structure. F61 and Y63 may promote the conformational changes needed to build secretable JC-containing pentamers.

The above results highlighted the key roles of the hydrophobic encounter complex in JC insertion, and of ERp44 in patrolling assembly. Based on the above results, IgM binding is likely initiated by interactions of I40, F61, and Y63 with the µtps of subunits 4 and 5 exposed by nascent IgM pentamers in opposite directions. Upon docking, JC needs to covalently bind the two µtps via C15 and C69 and adopt its final structure with three intra-chain disulfides (C72–92, C13–102, and C110–135). To investigate the underlying mechanisms, we generated cysteine replacement mutants and tested their capability to form pentamers in cellula (Figs. 7A,B and  EV5A) and in vitro (Figs. 7C,D and  EV5B).

In cells, both C15S and C69S variants yielded secretion-competent pentamers, albeit to a lower extent for C15S than WT JC (Fig. 7A lanes 3 and 6). Surprisingly, also their simultaneous replacement did not completely abolish the covalent binding of the double mutant C15-69S to IgM (Fig. 7A, lane 7), suggesting that other JC residues might interact with C575 in the µtp. While the C13S single mutant was partially incorporated into secreted IgM, the C13-C15S double mutant was not (Figs. 7A, lane 8-9 and  EV5A lanes 8-9). Accordingly, cells expressing C13-15S secreted abundant hexamers (Fig. 7B), confirming their poor insertion into IgM. In contrast, the mutants C72S and C69-C72S were inserted in polymeric IgM and partly secreted, with a lesser degree for the latter one (Fig. 7A lanes 4-5). Taken together, these results point to a pivotal role of the C13-C15 couple in triggering IgM binding.

In vitro, the mutants C13S and C15S were able to integrate to some extent into IgM pentamers, albeit displaying varying distribution of different species (Fig. 7C). Shoulder formation or shifts towards the hexamer peak in SEC analyses indicate the formation of a mixture of hexamers, pentamers in different ratios, while tailing hints at the presence of other undefined assemblies. The tendency of the C15S mutant towards the hexamer shoulder implies a pivotal role of this residue. More intermediates eluting after the pentamer were formed upon mutation of either C69 or C72 (Fig. 7D). The double mutants C69-72S (Fig. 7D) or C13-15S (Fig. 7C) formed larger amounts of hexamers than the single point mutants did. In the former, the pentamer peak is visible as merely a slight shoulder.

These results imply that in vitro both the C13xC15 and C69xxC72 motifs are important for pentamer assembly. However, the removal of both cysteines, which bind to IgM (C15 and C69), had only mild effects (Fig. 7D). A pronounced hexamer shoulder was evident, but despite a slight shift, still plenty of pentamers were formed in the presence of the C15S-C69S double mutant, confirming our in cellula results (Fig. 7A,B). In conclusion, in vitro, at least one cysteine of both the CxC and CxxC motifs needs to be present for JC binding to nascent IgM complexes.

We also analyzed two JC deletion mutants lacking hairpins that establish van-der-Waals contacts with the Cµ4 domains (Li et al, 2020b; Kumar et al, 2020). The C108-C135 disulfide stabilizes a loop crucial for IgM interaction with the pIgR and the secretory component but not for binding IgA (Johansen et al, 2001). Accordingly, a mutant in which this region was deleted (ΔHP3, Δ108-139) retained most of its capability to form pentamers in vitro (Fig. EV5B). The presence of a hexamer shoulder, however, shows that the region contributes to efficient assembly with pentamers. In contrast, the mutant ΔHP2 (Δ72-92) formed large amounts of aberrant assemblies. We conclude that JC incorporation strongly relies on cysteines 13, 15, 69, and 72, but the contacts between JC hairpins and the Cµ4 domains are of additional significance, as their absence induces the increased formation of either side products (ΔHP2) or hexamers (ΔHP3) in vitro (Johansen et al, 2001).

In summary, our data suggest that an ordered sequence of events dictates the biogenesis of JC-containing IgM pentamers in cellula as well as in vitro. Once the hydrophobic cues (revealed by the I40, F61 and Y63 mutants) have placed JC in the proper orientation to extend the amyloid core, disulfide rearrangements are triggered that lead C15 and C69 to form stable bonds with the two µtp C575, properly positioned in nascent IgM pentamers. Since the simultaneous absence of C13 and C15 was almost fatal for pentamer formation both in cellula and in vitro, and intra-chain bonds between these residues were found by MS in vitro (Fig. 4B), we surmise that the two cysteine residues in the CxC motif be necessary for starting the covalent binding of JC into IgM nascent pentamers. Once the process is initiated, the native intra-chain bonds as well as interactions with the hairpins 2 and 3, likely stabilize the final JC conformation.