Owing to the impossibility of reaching the Golgi for secretion or the cytosol for degradation, mutant Ig-μ chains that lack the first constant domain (μΔCH1) accumulate as detergent-insoluble aggregates in dilated endoplasmic reticulum cisternae, called Russell bodies. The presence of similar structures hallmarks many ER storage diseases, but their pathogenic role(s) remain obscure. Exploiting inducible cellular systems, we show here that Russell bodies form when the synthesis of μΔCH1 exceeds the degradation capacity. Condensation occurs in different sub-cellular locations, depending on the interacting molecules present in the host cell: if Ig light chains are co-expressed, detergent-insoluble μΔCH1-light chain oligomers accumulate in large ribosome-coated structures (rough Russell bodies). In absence of light chains, instead, aggregation occurs in smooth tubular vesicles and is controlled by N-glycan-dependent interactions with ER-Golgi intermediate compartment 53 (ERGIC-53). In cells containing smooth Russell bodies, ERGIC-53 co-localizes with μΔCH1 aggregates in a Ca2+-dependent fashion. Our findings identify a novel ERGIC-53 substrate, and indicate that interactions with light chains or ERGIC-53 seed μΔCH1 condensation in different stations of the early secretory pathway.
Secretory and membrane proteins fold and assemble in the endoplasmic reticulum (ER), under the assistance of a vast array of specialized chaperones and enzymes. The ER exerts a stringent quality control on its protein products. Proteins that fail to attain their native state are initially retained in the ER and eventually dispatched to the cytosol for proteasomal degradation (Ellgaard et al., 1999; Goldberg, 2003; Sitia and Braakman, 2003). This ER-associated degradation (ERAD) process is responsible for maintaining homeostasis in the ER lumen, and implies the recognition of misfolded or orphan proteins and their retro-translocation or dislocation across the ER membrane. Trimming of N-glycans is important to discriminate between unfolded and terminally misfolded glycoproteins, which explains how newly made proteins are given sufficient time to attain their three dimensional structure, a task that can take hours for secretory IgM and other complex molecules (Tsai et al., 2002; Ellgaard and Helenius, 2003; Hebert et al., 2005).
Pathology offers us many examples of disorders in the ER protein factory. Many genetic diseases, e.g. cystic fibrosis, are due to loss of function because otherwise functional proteins do not pass ER quality control, and are hence degraded. In other cases, the problem is the accumulation of mutant proteins in the ER, which can cause stress and cytotoxicity. An example of these disorders, collectively known as ER storage diseases (ERSD), is the hepatopathy caused by the accumulation of mutant α1 anti-trypsin (Liu et al., 1997; Carlson et al., 1989). Since not all patients harboring a mutant protein develop liver disease, cellular defense mechanisms must be operating. How these aggregates exert their toxicity on some cells is unknown.
Russell bodies (RB) are dilated cisternae of the ER containing large amounts of mutant immunoglobulins (Ig). RB are frequently detected in multiple myelomas, probably reflecting the intrinsic tendency of Ig genes to undergo somatic mutation. Cells with RB are called Mott cells (Weiss et al., 1984; Alanen et al., 1985). Although they contain massive protein aggregates in the ER, Mott cells remain viable (Weiss et al., 1984; Alanen et al., 1985) and undergo mitosis (Valetti et al., 1991), indicating that, unlike hepatocytes harboring mutant α1 anti-trypsin, myeloma cells can cope with the presence of RB. We could induce RB formation in lymphoid and non-lymphoid cells by over-expressing a mutant Ig-μ heavy chain that lacks the first constant domain (μΔCH1) provided that light chains (L) were also synthesized (Valetti et al., 1991). RB are induced also by other Ig isotypes, when the CH1 domain is missing (Kaloff and Haas, 1995). Their presence correlates with the accumulation of detergent-insoluble Ig, which can neither be transported to the Golgi nor be dispatched to the cytosol for degradation (Schweitzer et al., 1991; Valetti et al., 1991).
During Ig assembly, the CH1 domain pairs with the constant domain of light chains (CL). In the absence of light (L) chains, the CH1 domain strongly binds the immunoglobulin-binding protein (BiP/grp78), an ER-resident chaperone of the hsp70 family (Munro and Pelham, 1986). This results in retention of heavy (H) chains in the ER, until assembly with L chains displaces BiP (Hendershot et al., 1987; Hendershot, 1990).
Although formation of H2L2 subunits is sufficient for secretion of `monomeric' Ig (IgG, IgD and IgE), IgM and IgA must be further assembled into covalent polymers, often containing J chains (Hendershot and Sitia, 2004). Polymerization is developmentally regulated, offering an additional means to control IgM secretion at the post-translational level (Sitia et al., 1990). B lymphocytes are unable to form polymers and as a result they degrade virtually all secretory IgM and IgA they produce, mostly by the ERAD pathway (Sitia et al., 1990; Brewer et al., 1994; Guenzi et al., 1994; Mancini et al., 2000; van Anken et al., 2003). Owing to the inefficiency of polymerization, some IgM is degraded also by plasma cells (Sitia et al., 1987; Fra et al., 1993; Fagioli et al., 2001).
Why is ERAD unable to cope with certain mutant proteins? A simple explanation would be that their rate of synthesis exceeds the capacity of degradation. However, not all transport-incompetent Ig mutants cause RB biogenesis when degradation is inhibited. Therefore, additional features must be present for condensation-aggregation in the ER to take place. One possibility is that certain molecular interactions favor the formation of small aggregates that seed RB formation. Indeed, both assembly with L chains and inter-subunit interactions via cysteine 575 in the μs tailpiece, are important for efficient RB formation (Valetti et al., 1991). Therefore, the RB generated by mutant IgM somehow reflect abortive attempts to form secretion-competent polymers. Also the dislocation of ERAD substrates to the cytosol is likely to play an important role in RB biogenesis. In order to negotiate transport across the ER membrane, ERAD substrates must be unassembled and partially unfolded, with reduction of inter-chain disulfide bonds preceding dislocation (Fagioli et al., 2001). Soluble degradation substrates must be captured and concentrated in the vicinity of the retro-translocation machinery. Mechanisms must therefore be operating that couple disassembly and partial unfolding to the insertion of the substrate into the dislocon channels. The partially unfolded molecules generated in the process are at high risk for aggregating (Kopito and Sitia, 2000). In addition, the strong tendency of H chains lacking the CH1 domain to form insoluble aggregates in the ER can be related to the fact that they bind less BiP than wild-type molecules, thus enjoying less chaperoning.
In view of the biological and clinical interest of RB, we have investigated their biogenesis in lymphoid and non-lymphoid cells. Our results show that accumulation-aggregation ensues when the production of μΔCH1 exceeds the capacity of the ERAD system. In addition, we present evidence indicating that condensation-aggregation can take place in two different sub-regions of the early secretory pathway, depending on the interacting molecules present in the host cell. When L chains are present, aggregation takes place in the rough ER. In their absence, instead, μΔCH1 aggregate in smooth tubular vesicles in association with ERGIC-53, a marker of the ER-Golgi intermediate compartment (ERGIC) (Schweizer et al., 1991).
Induction of RB-like structures in the absence of L chains
We have previously described a stable myeloma transfectant that expresses a mutant μ chain lacking the first constant domain (μΔCH1), but fails to form RB (N[μΔCH1]). Expression of L chains readily induced the appearance of RB, suggesting that assembly with L chains is important for RB formation, possibly by cross-linking polymers in the ER through their unpaired CL domains [see Valetti et al. (Valetti et al., 1991) and below]. Another possibility is that assembly with L chains stabilizes mutant H chains (Sitia et al., 1987; Fagioli et al., 2001), thus slowing down ERAD. In this way, μΔCH1 could reach a threshold concentration in the ER, or within a specialized sub-compartment(s) thereof, sufficient to seed condensation.
To determine whether RB could be induced in the absence of L chains, we modulated the ratio between synthesis and degradation. First, we increased the rate of μΔCH1 production by exposing cells to sodium butyrate. At the concentration of 1 mM, sodium butyrate approximately doubled transgene expression without significantly inducing apoptosis, as assayed by metabolic pulse assays and annexin V-propidium iodide staining (our unpublished results). As shown in Fig. 1A, untreated N[μΔCH1] and N[μ1] cells (the latter expressing the wild-type form of the secretory μ chains, μs) display a diffuse staining of the ER (Valetti et al., 1991). These clones have presumably reached an equilibrium between synthesis and degradation of the orphan μ chains. After 40 hours of sodium butyrate treatment, 30-50% of N[μΔCH1] cells developed RB-like structures, of varying in size and number. In contrast, N[μ1] did not form RB despite increased μ synthesis as identified by a brighter immunofluorescent staining pattern (Fig. 1A) and in western blots (Fig. 1B). μs and μΔCH1 differed also with respect to their detergent solubility (Fig. 1B). Whereas virtually all wild-type μs chains were detergent soluble (s), half or more μΔCH1 accumulated in the insoluble fraction (i).
Therefore, RB can form also without L chains. However, morphological differences were evident in myeloma transfectants depending on the presence or absence of L chains. Whereas L chain-containing RB tended to be fewer in number and more circular (see J[μΔCH1] cells, Fig. 1A), a constant feature in N[μΔCH1] cells was that the structures containing mutant μ chains were smaller and less regular, often polygonal (Fig. 1A, NaBut and Fig. 1C). Their size ranged from 0.2 to 0.6 μm, with a mean value of 0.4 μm (Fig. 1E), hence much smaller that L chain containing RB in J[μΔCH1] (average 1.3 μm). Cryo-EM analyses with anti-μ (Fig. 1D) confirmed that in both cases the structures contained abundant μΔCH1 chains.
Besides confirming the size and shape differences observed by immunofluorescent staining, resin embedding electron microscopy (EM) studies (Fig. 1C) revealed that in cells expressing L chains, RB displayed ribosomes all around their surface. In contrast, the structures present in N[μΔCH1] cells had only occasional ribosomes, often concentrated in discrete areas. We will refer to these latter structures as smooth RB (sRB).
Inhibiting degradation favors smooth RB formation
To test whether a decrease in degradation could also enhance RB formation, we utilized proteasome inhibitors. First, we confirmed that μΔCH1 chains are degraded by proteasomes, by chasing pulse-labeled cells with or without MG132 (data not shown) as described for μs (Mancini et al., 2000; Fagioli et al., 2001). In the course of these experiments, we observed that myeloma cells are extremely sensitive to proteasome inhibitors, the vast majority of them undergoing apoptosis after 6 hours of treatment or less (Cenci et al., 2006). To circumvent this problem, we combined treatments. As shown in Fig. 2, adding MG132 for the last 2 hours of a 24-hour treatment with sodium butyrate, doubled the number of cells containing sRB in N[μΔCH1] cells, indicating that aggregation-prone μΔCH1 chains condense in the ER in a detergent insoluble state when their synthesis exceeds disposal.
Induction of RB in non-lymphoid cells
In view of the possibility that sodium butyrate exerted additional effects, we generated inducible HeLa cells (H[μΔCH1]) in which μΔCH1 synthesis is driven by a Tet-Off promoter (Gossen and Bujard, 1992). These cells formed RB-like structures within a few days after removal of tetracycline (Fig. 3A), in correlation with the accumulation of detergent-insoluble μΔCH1 chains (Fig. 3B). In the first 2 days of induction, when few μΔCH1 are detectable in the Triton X-100 (TX100) soluble fraction (Fig. 3B), a weak reticular and peri-Golgi staining was evident by immunofluorescence microscopy. At day 4, small, bright dots became evident in 20-40% of μΔCH1 producing cells (see inset). These dots, usually one in each cell, probably represent the first morphological sign of μΔCH1 aggregation. With time, they increased progressively in size, and this correlated with an increase in the insoluble pellet (Fig. 3B). These structures, present in 50-70% of cells after day 5, were irregular in shape (see also below) and were generally positioned close to the microtubule-organizing center (MTOC) area, as shown by their proximity to γ-tubulin (see Fig. 3C).
Dose-response experiments confirmed that a critical μΔCH1 concentration must be reached for triggering condensation-aggregation in the ER (summarized in Fig. 3D). At concentrations of tetracycline of 0.25 ng/ml, when synthesis of the transgene was minimal, the majority of μΔCH1 remained in the soluble fraction. At 0.12 ng/ml, insoluble μΔCH1 chains became preponderant, and increased progressively at lower tetracycline concentrations. The soluble fraction remained essentially constant, possibly reflecting the chaperoning power of the cell.
Distinct sites of accumulation in the presence or absence of L chain in the inducible H[μΔCH1] clone
The co-expression of mutant μ and λ chains dramatically influenced the shape of RB structures in non-lymphoid cells also (Fig. 4). Instead of the irregularly shaped structures concentrated around the MTOC area (Fig. 4a), numerous spherical vacuoles, rather regular in shape and dispersed throughout the cell, accumulated in λ producing cells (Fig. 4b). These structures were stained by Ac38 anti-idiotypic antibodies (Reth et al., 1979; Valetti et al., 1991), confirming proper pairing between the μ and λ variable (V) domains (Fig. 4b). As in myeloma transfectants (compare Fig. 1, J[μΔCH1] cells), these structures contained an electron dense material surrounded by ribosome-coated membranes (Fig. 4h). In contrast, smaller vacuoles and tubules, generally devoid of ribosomes, accumulated in cells expressing μΔCH1 chains alone (Fig. 4g).
Consistent with the presence of ribosomes on their membranes, rough RB (rRB) containing λ chains were stained by anti-PDI (data not shown), but lacked ERGIC-53 (Fig. 4e). In contrast, sRB containing μΔCH1 alone were intensely stained by anti-ERGIC-53 antibodies (Fig. 4d), suggesting that in the absence of L chains, condensation occurs in distal regions of the early secretory compartments, including ERGIC.
How could the presence of λ chains induce condensation in rRB? In the absence of CH1, the constant domain of L chains, CL, does not find its natural counterpart. Like most Ig domains, CL tends to form dimers (Leitzgen et al., 1997). This could favor aggregate formation, by linking different (μΔCH1-λ)n polymers in a three-dimensional mesh (Valetti et al., 1991). To verify this model, we deleted the CL domain from murine λ chains. These truncated variable domains (VL) assembled with μΔCH1 chains, as demonstrated by Ac38 staining (Fig. 4c), but did not alter the distribution of μΔCH1, which accumulated within ERGIC-53-positive sRB (Fig. 4f). Therefore, the unpaired CL domains seem to play an important role in favoring rRB formation.
ERGIC-53 is enriched in sRB, in the absence of L chains
After 7 days in culture without tetracycline, the majority of the ERGIC-53 staining was concentrated in the region occupied by μΔCH1 (Fig. 5e). This clearly contrasted with the distribution observed in cells cultured in the presence of tetracycline (Fig. 5a). At day 3 of induction, an intermediate phenotype was observed, indicating that ERGIC-53 was recruited in the μΔCH1-containing structures (Fig. 5c). The distribution of GM-130 was not altered, although sRB were generally adjacent to the Golgi complex (Fig. 5b,d,f).
We further characterized sRB by EM. In the absence of L chains, μΔCH1 accumulated in tubular structures, somehow reminiscent of a dish of spaghetti or a ball of yarn (Fig. 5g-i). The interlaced tubules were generally devoid of ribosomes (Fig. 5i-j) and intensely stained by anti-μ and anti-ERGIC-53 antibodies (Fig. 5g and h, respectively). Many examples of continuity between these tubules and normal ER cisternae were present (see Fig. 5j).
Altogether, these results indicated that μΔCH1 can accumulate in different compartments of the early exocytic route. When L chains are present, mutant proteins condense in the rough ER, whereas in their absence aggregation occurs in tubular vesicles enriched in ERGIC-53. L chain expression superimposes a more regular shape to rRB. An important role is played by the unpaired CL domain, which could trigger condensation by linking μΔCH1 polymers or perhaps favoring polymerization in the ER (see Fig. 4).
ERGIC-53 is important for sRB biogenesis and binds μΔCH1 in a Ca2+-dependent manner
To determine whether the co-localization of μΔCH1 and ERGIC-53 reflected specific interactions between the two molecules, we exploited the finding that glycoprotein binding to the lectin requires Ca2+ (Itin et al., 1996; Appenzeller-Herzog et al., 2004). Thus, H[μΔCH1] cells were incubated with thapsigargin or cyclopiazonic acid (CPA), two drugs known to induce Ca2+ release from the ER in an irreversible or reversible manner, respectively (Doutheil et al., 1997). When Ca2+ levels were lowered from intracellular stores with CPA (Fig. 6A, top right panel), or thapsigargin (not shown), μΔCH1 remained confined to sRB, whereas ERGIC-53 regained its normal dispersed distribution (Fig. 6A, top right). ERGIC-53 re-localized into sRB upon removal of CPA and culture in Ca2+-containing medium for 15 and 45 minutes (Fig. 6A, bottom left and right). These findings suggest that sRB do not represent a dead end station, but are still connected with vesicular traffic.
Further evidence for a Ca2+-dependent interaction between μΔCH1 and ERGIC-53 was obtained by biochemical studies. In non-induced H[μΔCH1] cells, virtually all ERGIC-53 was detergent-soluble (Fig. 6B, lanes 5 and 6). After tetracycline removal, when cells accumulated abundant detergent-insoluble μΔCH1 chains, a considerable fraction of ERGIC-53 was present in the insoluble fraction (lanes 3 and 4). However, if cells were lysed in the absence of Ca2+, ERGIC-53 was found exclusively in the soluble fraction (lanes 1 and 2). Therefore, the sub-cellular localization of ERGIC-53 is altered by the presence of μΔCH1 aggregates in a Ca2+-dependent fashion.
To further investigate the interactions between μΔCH1 and ERGIC-53 we performed cross-linking and co-immunoprecipitation studies (Fig. 6C,D). Anti-μ antibodies co-precipitated a considerable fraction of ERGIC-53 wt and the localization-defective KKAA mutant (Fig. 6C, lane 1) but only traces of the binding-deficient N156A mutant. The interaction was clearly inhibited when intracellular Ca2+ stores were depleted (Fig. 6C, compare lanes 1 and 3). Treatment with CPA significantly reduced the amount of ERGIC-53 that could be precipitated with anti-μ antibodies in cells expressing wt or KKAA molecules, but had no effect on the binding-deficient N156A mutant (Fig. 6D), confirming that ERGIC-53 binds μΔCH1 chains in a Ca2+-dependent way via its lectin domain (Appenzeller et al., 1999). Despite our efforts, we failed to co-immunoprecipitate endogenous ERGIC-53 with anti-μ antibodies. This may reflect technical limitations, and also the transient nature of the interactions between ERGIC-53 and substrates. Over-expressed ERGIC-53 molecules could be more easily co-immunoprecipitated because of their localization in the ER (L.M. and C.V., unpublished results), where cargo capture prevails. Nonetheless, the Ca2+-dependent delocalization assays shown in Fig. 6A, are consistent with an interaction of μΔCH1 with endogenous ERGIC-53.
Mannose trimming of μΔCH1 is required for sRB biogenesis
The above data established that in the absence of L chains μΔCH1 chains condense in sRB in association with ERGIC-53. If interactions with this lectin molecule played an active role in condensation, perturbing N-glycan processing could have an effect on sRB formation in cells lacking L chains. In view of the role of mannose residues in targeting misfolded proteins to degradation (Helenius and Aebi, 2004) these experiments could also shed light on the relationships between ERAD and RB formation. We have previously demonstrated that the mannosidase I inhibitor kifunensine prevents degradation/dislocation of orphan wild-type μs (Fagioli and Sitia, 2001) and μΔCH1 chains (our unpublished data), as efficiently as proteasome inhibitors. Therefore, if the concentration of an aggregation prone molecule were sufficient to determine condensation, kifunensine should favor the biogenesis of RB in both myeloma and HeLa transfectants, by causing accumulation of μΔCH1 chains in the ER. If in contrast, glycan-dependent interactions with intracellular lectins were important for condensation, mannosidase inhibitors could prevent sRB formation.
In a first series of experiments, N[μΔCH1] cells were treated with either sodium butyrate or kifunensine for 40 hours. As expected, both treatments considerably increased the amount of μΔCH1 found in the cells lysates (Fig. 7A,B lanes 3-6). However, neither the formation of sRB structures (Fig. 7A) nor the accumulation of detergent-insoluble molecules (Fig. 7B lanes 5 and 6) was induced by kifunensine.
Also in H[μΔCH1] cells, kifunensine prevented sRB formation (Fig. 7C). After 5 days of tetracycline removal, less than 10% of the cells in which mannosidase I was inhibited with kifunensine contained sRB, compared with over 60% in controls. Most cells displayed a diffuse, ER staining pattern when decorated with anti-μ (Fig. 7C). In both kifunensine-treated myeloma and HeLa cells μΔCH1 chains displayed slower mobility because of their higher mannose content and accumulated largely in the detergent-soluble fraction (panels B and D).
Inhibition of mannose trimming also inhibits the interactions between μΔCH1 and ERGIC-53 (Fig. 7E,F). When cells were treated with kifunensine the quantity of ERGIC-53 co-precipitated with anti-μ antibodies was significantly reduced (Fig. 7E, lane 5) to a level similar to that obtained with CPA (see lane 3) as quantified in Fig. 7F and Fig. 6D.
Kifunensine had minor if any effects on rRB formation in cells synthesizing L chains (data not shown), indicating that mannose removal was not essential for detergent insolubility per se. Altogether, these data suggest a crucial role for mannose processing in determining the fate of aggregation-prone μΔCH1 chains, and imply that, in the absence of L chains, a key factor in this process could be ERGIC-53.
ER storage disorders (ERSD) are a recently defined class of diseases characterized by the accumulation of aberrant proteins in the ER. They are `conformational diseases' (Kopito and Ron, 2000), a definition that nicely pinpoints the pathogenic role of proteins that owing to mutations or lack of cofactors assume aberrant conformations. These are recognized by the ER quality control system, and cannot proceed to the Golgi. If these transport-incompetent molecules are not efficiently routed to the cytosol for proteasomal degradation (dislocated), accumulation in the ER lumen inevitably ensues. Not all proteins that fail to reach the Golgi, however, accumulate in the ER. Why do certain proteins, such as for instance, PiZα1AT or μΔCH1, fail to dislocate to the cytosol? ERAD could be inefficient in their capture, unfolding, dislocon insertion, and/or delivery to the cytosol, but it seems clear that these proteins have an intrinsic tendency to form oligomers, probably too big to negotiate export (Carrell and Lomas, 1997).
Our results reveal important facets of the mechanisms that determine aggregation of μΔCH1 chains and hence RB formation. First of all, by manipulating the rate of synthesis and degradation of the condensation-prone μΔCH1 chains, we could demonstrate that RB formation ensues when the synthesis exceeds the degradative capacity. In both lymphoid and non-lymphoid cells, detergent-insoluble μΔCH1 chains accumulate when synthesis is increased (Figs 1, 3) or degradation inhibited (Fig. 2). It seems therefore that cells can cope with limited amounts of mutant μ chains. This is evident in HeLa cells, where intracellular detergent-soluble μΔCH1 reach a plateau (Fig. 3B,D), only insoluble chains increasing with time. Since CH1 domains of all Ig classes bind BiP with high affinity, formation of detergent-insoluble aggregates by mutants that lack them could reflect reduced interactions with the main ER chaperone.
The threshold for aggregation varies for different cell types and also depends on interacting molecules. For example, the fraction of insoluble μΔCH1 is much higher in cells synthesizing L chains. The latter have notable consequences also on the shape and location of RB. In their absence, μΔCH1 condensation occurs in irregular masses of tubular vesicles and vacuoles surrounded by smooth membranes. These sRB are enriched in ERGIC-53, a lectin involved in intracellular glycoprotein transport (Itin et al., 1996; Ben-Tekaya et al., 2005). In contrast, in the presence of L chains, RB adopt circular and regular shape, contain PDI and are surrounded by a membrane rich in ribosomes (Figs 1, 4), suggesting that μΔCH1 condensation occurs near the site of synthesis.
How do L chains favor rRB formation? Two non-mutually exclusive mechanisms can be envisaged. First, the association with L could slow down μΔCH1 degradation, imposing a further requirement for unfolding-dissociation before dislocation (Fagioli et al., 2001). This would increase concentration in the ER lumen, indirectly promoting condensation-aggregation. Second, because of the tendency of most Ig constant domains to form dimers, the unpaired CL associated to μΔCH1 via VL-VH interactions could link different polymers, thus promoting condensation (Valetti et al., 1991). In favor of the latter model is the observation that L chains lacking the CL do not favor rRB formation. Despite proper VH-VL pairing, demonstrated by reactivity with anti-idiotypic antibodies, condensation-aggregation occurred in ERGIC-53-positive sRB, as in the absence of L chains.
The role of N-glycan processing in RB formation
The data presented in Figs 1, 2, 3 are consistent with a simple model, in which RB form when the production of a transport-incompetent molecule exceeds the degradation capacity. However, the results obtained with kifunensine, an inhibitor of ER mannosidase I that blocks μ chain degradation (Fagioli and Sitia, 2001) imply that a mere increase in concentration is not sufficient to cause RB formation. Despite kifunensine-treated NSO and HeLa transfectants accumulating abundant μΔCH1, neither formed sRB (Fig. 7). By contrast, kifunensine did not interfere with RB formation in cells expressing L chains, suggesting an important role for mannose trimming in triggering μΔCH1 condensation in sRB.
Interactions between μΔCH1 and ERGIC-53
ERGIC-53 is a Ca2+-dependent lectin thought to concentrate selected secretory glycoproteins in forward transport vesicles budding from the ER (Itin et al., 1996; Appenzeller-Herzog et al., 2004; Ben-Tekaya et al., 2005). At steady state, ERGIC-53 is enriched in the ERGIC, owing to the presence of a KKFF motif in its cytosolic tail. Our results identify μΔCH1 as a novel substrate of ERGIC-53. They also indicate that interactions between μΔCH1 and ERGIC-53 play an important role in sRB biogenesis in the absence of L chains. ERGIC-53 looses its normal distribution and re-localizes in sRB, in association with μΔCH1. Part of ERGIC-53 is found in the detergent-insoluble fraction and can be cross-linked to μΔCH1 in a Ca2+ dependent manner. Ca2+ was necessary for association, and the N156A mutant was inactive, implying specific interactions through the lectin domain. The observation that kifunensine inhibited binding of μΔCH1 chains to ERGIC-53 was somehow unexpected, since another inhibitor of ER mannosidases (deoxymannojiirimycin) did not prevent binding of cathepsin Z (Appenzeller et al., 1999). Persistence of the terminal mannose may favor binding of μΔCH1 to calnexin and calreticulin (Mancini et al., 2000) (L.M., C.V. and R.S., unpublished results), thus preventing interactions with ERGIC-53. Another intriguing possibility is that subtle structural features that are dependent on the interactions between N-glycans and the surrounding polypeptide backbone determine the specificity of binding (Appenzeller-Herzog, 2005; Cals et al., 1996; de Lalla et al., 1998).
In the absence of L chains, therefore, μΔCH1 chains condense in tubular vesicles, in association with ERGIC-53. This network is largely devoid of ribosomes and contains a dense material intensely stained by anti-μ antibodies. Some of these tubules are clearly continuous with ER cisternae. Condensation of μΔCH1 could take place at the junctions, which often appear as if constricted by a ring. These tubular vesicles could correspond to ER exit sites, in which ERGIC-53 and other molecular complexes concentrate cargo for Golgi transport. It is also possible that junctions or sRB correspond to sites of the early secretory pathway specialized in dislocating ERAD substrates across the ER membrane for cytosolic delivery (Kamhi-Nesher et al., 2001).
Protein-protein interactions dictate the shape and localization of RB
Our results are consistent with an active role of ERGIC-53 in seeding μΔCH1 aggregation in a sugar- and lectin-dependent way. Replacing Cys575 or treating cells with 2 mercapto-ethanol inhibits RB formation (Valetti et al., 1991), indicating that formation of disulfide bonds linking different [μΔCH1]2 dimers is essential for efficient condensation. As discussed above, L chains could mediate the association of polymers, or favor their formation in the ER (Bornemann et al., 1995; Cals et al., 1996). Likewise, ERGIC-53 could seed condensation, perhaps exploiting its property to form hexamers. The results presented in Fig. 6 indicate that once formed, detergent-insoluble complexes do not require the continuous interaction with ERGIC-53. They also imply that these aggregates are not sequestered but remain in dynamic contact with the exocytic pathway, since ERGIC-53 can re-localize in sRB upon re-addition of Ca2+.
In conclusion, our findings reveal that protein aggregation in the ER depends on several cooperating events: abundant synthesis and reduced degradation of proteins that have an intrinsic tendency to form aggregates because of reduced chaperone binding and interactivity with ancillary proteins that favor their concentration and seed condensation. The equilibrium between synthesis, processing, and degradation of aberrant species is subject to delicate and continuous adjustments, operated by the ER quality control machinery.
Materials and Methods
Anti-γ-tubulin monoclonal antibodies, cyclopiazonic acid, dithiobis(succinimidylpropionate, geneticin, N-ethylmaleimide, sodium butyrate, tetracycline, thapsigargin and Triton X-100, were from Sigma-Aldrich; dithiobis(succinimidylpropionate) (DSP) was from Pierce, hygromycin B from Invitrogen and kifunensine from Toronto Research Chemicals Inc.
Polyclonal anti ERGIC-53 (Spatuzza et al., 2004) was a kind gift from Stefano Bonatti (University of Naples, Italy). Polyclonal anti-mouse μ was purchased from Zymed; HRP anti-λ, HRP anti-mouse IgG, TRITC anti-μ were from Southern Biotechnology Associates; HRP anti-rabbit was from Dako; FITC anti-mouse γ chains were from Jackson Laboratories. Anti-Myc (9E10) and anti-ERGIC-53 (G1/93) mAbs were described elsewhere (Evan et al., 1985; Schweizer et al., 1988), and anti-GM-130 was purchased from Transduction Labs.
NSO (Cowan et al., 1974), J558L (Oi et al., 1983) and the derived stable transfectants N[μ1], N[μΔCH1] and J[μΔCH1] were obtained and maintained as described previously (Sitia et al., 1987; Sitia et al., 1990; Valetti et al., 1991). HeLa Tet-Off (Clontech Laboratories) were maintained in DMEM as recommended by the supplier.
Plasmids and vectors
cDNA encoding μΔCH1 was prepared by RT-PCR from total RNA isolated from J[μΔCH1] cells. The first strand cDNA was synthesized by incubation with a μ-specific primer (GGTTAGTTTGCATACACAGAG) and Moloney murine leukemia virus reverse transcriptase RNase H Minus Point Mutant (Promega) for 1 hour at 42°C. The resulting cDNA was used for PCR amplification with primers containing unique restriction sites: EcoRI for the forward primer (GGAATTCGCACACAGGACCTCACC) and XbaI for the reverse primer (AAATCTAGACTGGTTGAGCGCTAGCATGG). PCR products were digested and cloned in pTRE (Clontech Laboratories Inc.) and in pCDNA3.1(+) (Invitrogen, Life Technologies).
pcDNA3.1 encoding λ1 chain was described previously (Fagioli et al., 2001). The truncated variant lacking the constant domain (VL), was obtained by PCR amplification with primers containing unique restriction sites, EcoRI (GAATTCATGGCCTGGATTTCACTT) and XbaI (CTCGAGCTATAGGACAGTCAGTTTGGT) for the forward and reverse primers, respectively. The latter contains a stop codon before the XbaI site. PCR products were cloned in pCDNA3.1(+).
Vectors encoding Myc-tagged, glycosylated human wild-type and mutant ERGIC-53-KKAA and ERGIC-53-N156A (Appenzeller et al., 1999) were kind gifts from Hans-Peter Hauri (Biozentrum, University of Basel, Switzerland).
Establishment of Tet-inducible HeLa cells
HeLa Tet-Off cells were transfected with pTRE-μΔCH1 (50 μg) and pTK-Hyg (5 μg, Clontech Laboratories). Clones were grown in the presence of 3 μg/ml tetracycline, 400 μg/ml hygromycin and 100 μg/ml geneticin. Hygromycin-resistant clones were screened by immunofluorescence 72 hours after tetracycline removal.
Cell lysis and western blotting
Cells were washed and lysed at the concentration of 1×104 cells/μl in buffer A (0.2% TX100, 50 mM Tris-HCl pH 7.5), 150 mM NaCl, 5 mM EDTA, 1 mM N-ethylmaleimide and a cocktail of protease inhibitors (Roche). The TX100-insoluble fraction was separated by centrifugation at 3,400 g for 10 minutes, washed twice in buffer A and further solubilized in lysis buffer B (1% SDS, 50 mM Tris-HCl pH 7.5) for 10 minutes at room temperature (RT), diluted in 50 mM Tris-HCl pH 7.5, 0.2% TX100, to keep the volume of the soluble and insoluble fractions equal, and sonicated for 10 seconds. Immunoprecipitation and western blotting were performed as previously described (Valetti et al., 1991). EDTA was omitted from buffer A, and 1 mM CaCl2 added in co-immunoprecipitation assays involving ERGIC-53. The intensity of the relevant bands was quantified by the software package, IPLab spectrum V3.2 (Scanalytics, Rockville, MD).
Cells were washed three times with Hanks and incubated for 10 minutes at RT with 1 mM DSP on a rocking platform. The reaction was quenched by rinsing the cells twice with 50 mM Tris-HCl pH 7.5 followed by incubation for 5 minutes at RT with the same buffer.
Lymphoid cells were incubated on poly-L-lysine-coated coverslips for 20 minutes at RT whereas HeLa Tet-Off cells were grown directly on 10-mm2 coverslips. Cells were fixed in 3% paraformaldehyde and permeabilized with 0.1% TX100; for γ-tubulin staining, cells were fixed in methanol at -20°C. In cells expressing μΔCH1, pre-adsorbed γ-chain-specific rabbit anti-mouse antibodies were used to avoid cross-reactions. Samples were inspected on an Olympus inverted fluorescence microscope (model IX70). Images were recorded with a digital CCD camera (model C4742-95, Hamamatsu), using the IPLab spectrum V3.2 software package (Scanalytics) and processed with Adobe Photoshop 7.0 (Adobe Systems).
For cryosectioning, cells were fixed with 2% paraformaldehyde and 0.2% glutaraldehyde in PBS for 2 hours, washed with PBS containing 20 mM glycine, scraped off the dish, centrifuged and embedded in 12% gelatin in PBS. Small blocks of embedded cells were incubated overnight with 2.3 M sucrose at 4°C, mounted on aluminum pins and frozen in liquid nitrogen. 60 nm ultrathin cryosections were cut at -120°C, using a cryo-ultramicrotome (Leica-Ultracut EM FCS), and picked up with 1% methylcellulose in 1.15 M sucrose. Cryosections were then incubated with primary antibodies and revealed with protein A gold, according to previously described protocols (Slot et al., 1989).
For Epon embedding, cells were fixed with 0.1 M cacodylate buffer, pH 7.4, containing 2.5% glutaraldehyde at RT for 10 minutes. After washing with 0.1 M cacodylate buffer pH 7.4, and post-fixing for 10 minutes at RT with 0.1 M cacodylate buffer, containing 1% OsO4 w/v. Cells were then washed with 0.1 M cacodylate buffer, and processed for Epon embedding (Polybed 812, Polysciences). Sections were analyzed with a Philips CM-10 electron microscope (Einhdoven, the Netherlands).
This work is dedicated to the memory of César Milstein, who raised our interest in Russell bodies. We thank S. Bonatti (University of Naples), C. E. Grossi (University of Genoa), H.-P. Hauri (University of Basel) and M. Otsu (DiBiT-HSR, Milan), for providing excellent reagents and helpful suggestions; K. Cortese, M. C. Gagliani and M. Bono for help with electron microscopy. This work was supported through grants from the Associazione Italiana per la Ricerca sul Cancro (AIRC), Ministero della Sanità, MIUR (CoFin and Center of Excellence in Physiopathology of Cell Differentiation) and Telethon. Electron microscopy studies were performed at the Telethon Facility for Electron Microscopy (Grant no. GTF03001). T.A. was a recipient of a fellowship from the Federazione Italiana Ricerca sul Cancro.
↵‡ These authors contributed equally to this work
- Accepted March 13, 2006.
- © The Company of Biologists Limited 2006