Intra-Golgi Formation of IgM–Glycosaminoglycan Complexes Promotes Ig Deposition

Autoantibody production is controlled by immune tolerance (1, 2). When the initial Ig rearrangements encode autoreactive H and L chain pairs, tolerance impairs maturation of B cells with such autoreactive Ig receptors. B cell maturation may proceed following further diversification of autoreactive H/L pairs. The autoreactive receptors are revised, or edited, by expression of new H or L chains that are accessed by additional V gene rearrangements (3). Receptor editing usually effectively prevents or limits autoreactivity. Some new rearrangements fail to completely eliminate autoantigen binding, resulting in receptor dilution (4–6) or the production of receptors with intrinsic affinity for intracellular Ags (7). It is plausible that such incompletely edited autoreactive B cells are the precursors of B cells that express pathogenic autoantibodies.

One mechanism whereby autoreactive Abs lead to autoimmune pathogenesis involves immune complexes (ICs) that form in situ or in circulation. Systemic lupus erythematosus is a classic example of an IC deposition disorder (8). In lupus, IC deposits activate complement and other mediators that lead to tissue damage, most notably glomerulonephritis. One prevalent view holds that ICs composed of antinuclear autoantibodies and nucleoprotein Ags released from dying cells form in the circulation (9, 10) and then lodge along the glomerular basement membrane as subendothelial deposits (11). An alternative view is that ICs form in situ because of binding of autoreactive Ig to Ags that are expressed or deposited in the kidneys (12, 13). In either case, B cells are thought to produce autoreactive Abs that form ICs after binding to Ags elsewhere in the body. However, there is no generally accepted concordance between autoantibody specificity and IC deposition, and alternative mechanisms of IC formation may prove to have a role in pathogenesis.

A broad category of Ig deposition diseases includes those that result from plasma cell dyscrasias and lead to paraprotein deposits and amyloidosis (14, 15). Ig deposits are frequently associated with B cell malignancy (16). Autoantigens may be involved in the stimulation of B cell proliferative diseases, as suggested by the restricted use of particular V genes (17) and the accumulation of somatic mutations (18). Autoantigens may also participate in the formation of Ig deposits, such as in mixed cryoglobulinemia where precipitates reflect Ab–Ag interactions (19). Similarly, amyloid deposits of Ig L or H chains may incorporate glycosaminoglycans (GAGs) into the fibril structure (20).

Our studies on receptor editing uncovered an incompletely edited Ab that was formed by pairing the VH56R anti-DNA H chain with an unusual editor L chain. The VH56R/Vκ38c pair bound DNA and phospholipid self-Ags, yet it was represented in the mature B cell repertoire. We discovered that a substantial portion of the intracellular VH56R/Vκ38c IgM accumulated in the Golgi complex in B cell hybridomas. Additionally, we also observed that the VH56R/Vκ38c-producing B cells exhibited a CD21^hiCD23^int marginal zone phenotype and displayed Ab clusters at the cell surface. The same VH56R H chain with a different editor L chain, Vλx, did not accumulate in the Golgi or form IgM surface clusters (7).

In this study, we explored the basis for VH56R/Vκ38c accumulation in the Golgi and followed the extracellular fate of IgM complexes. We found that VH56R/Vκ38c IgM was released from the cell as soluble hemimers and monomers as well as in large, spherical bodies (spherons), in contrast to VH56R/Vλx IgM that was secreted as polymeric IgM. The spherons contained IgM polymers that incorporated the J chain and acquired endoglycosidase H (Endo H)-resistant glycosyl moieties, arguing that spheron-associated IgM had been processed in the Golgi. Addition of chondroitinase avidin/biotin complex (ABC), an enzyme with broad activity toward GAGs, decreased the average size of spherons, as did treatment of hybridomas with β-d-xyloside, a cell-penetrating inhibitor of glycan biosynthesis. Additionally, the VH56R/Vκ38c IgM isolated from spherons bound heparin and hyaluronic acid, suggesting that IgM binding to repetitive, anionic GAGs contributes to spheron formation. These results indicate that binding of IgM to glycans leads to the production of large, insoluble IgM complexes that could have counterparts in vivo.

The 38C-13 lymphoma (21) was a gift from Prof. J. Haimovich (Tel Aviv University), and WEHI-231 (22) was a gift from Dr. A.K. Yi (University of Tennessee Health Science Center). B cell hybridomas expressing VH56R/Vκ38c IgM, VH76R/Vκ38c IgM, and VH56R/Vλx IgM have been previously described (7).

Rabbit anti–β-coatomer (COP; Affinity Bioreagents, Golden, CO), anti-giantin, and anti–Golgin-95 (23), mouse anti-mannosidase II (Covance, Denver, PA), Griffonia simplicifolia lectin II-AF488 (Invitrogen, Carlsbad, CA), and goat anti-mouse IgM-AF488 or -AF647 (Invitrogen) were used in immunofluorescence, as previously described (7). DNA was visualized with Sytox Orange (Invitrogen) and cell membranes with 3,3′-dioctadecyloxacarbocyanine perchlorate (DiO), CellMask Deep Red (CM-DR), or AF488-labeled annexin V (Invitrogen). To induce Golgi fragmentation, cells were incubated with nocodazole (Sigma-Aldrich, St. Louis, MO) at 2.5 μg/ml for 2 h. To colocalize IgM and Golgi markers, pixels corresponding to anti-IgM were quantified using Carl Zeiss LSM software (release 3.2) and assigned to the Golgi versus the remainder of the cell in optical sections projected in three dimensions. Colocalization with Golgi markers was assessed by dividing pixels showing overlap (yellow) by anti-IgM pixels (red or yellow). At least five cells were analyzed for each marker.

Cells grown on polyester membranes (Corning, Acton, MA) were fixed in paraformaldehyde/glutaraldehyde, incubated in OsO₄, and dehydrated prior to embedding in Spurr’s resin (Electron Microscopy Sciences, Hatfield, PA). Nickel grids holding 75-nm sections were treated with uranyl acetate and lead citrate (Leica, Deerfield, IL). Sections were viewed with a JEOL JEM-2000 EX II electron microscope.

Cells were removed from flasks by pipetting with PBS. Spherons were dislodged with a scraper, taken up in HBSS, filtered through a 5-μm filter (Millipore, Billerica, MA), and pelleted at 25,000 rpm in an SW-31 rotor for 30 min. Spherons were fixed in paraformaldehyde-glutaraldehyde and incubated for 2 h with goat anti-mouse IgM conjugated to 25-nm gold particles (EMS). The spherons were embedded, sectioned, and viewed by electron microscopy as above. Purified spherons were also embedded in agarose, fixed with 10% paraformaldehyde, paraffinized, and sectioned. Samples were stained with colloidal iron. Staining of spherons on cover slips with thioflavin T used standard procedures.

Cells were labeled with 100 μCi ³⁵S-labeled cysteine and methionine (MP Biomedicals, Solon, OH) for 16 h. Cultures were fractionated into cells, clarified supernatants, and spherons. Spherons were collected as above. Ten million cells were separated from media by centrifugation and lysed in buffered saline containing 0.5% deoxycholate/0.5% Nonidet P-40 (NP-40). Cleared lysates were collected after centrifugation and pellets were solubilized by boiling in 2% SDS. Goat anti-mouse IgM (SouthernBiotech, Birmingham, AL) was added to all fractions for 2 h and immunoprecipitates were collected on protein A-Sepharose beads (Sigma-Aldrich). Samples were aliquoted and analyzed by reducing and nonreducing SDS-PAGE. Gels were dried and exposed for autoradiography.

IgM from different cell fractions was untreated or incubated with Endo H or PNGase F (New England Biolabs, Ipswich, MA). IgM was denatured for Endo H as suggested by the manufacturer and for PNGase F in the denaturing buffer (0.5% SDS, 150 mM Tris [pH 7.8]) prior to digestion and analysis by reducing SDS-PAGE.

Proteins were transferred to nitrocellulose and probed with alkaline phosphatase-labeled goat anti-mouse μ, goat anti-mouse κ Abs (SouthernBiotech), or rabbit anti-J chain IgG (Nordic Immunology, Tilburg, The Netherlands). Binding was detected by HRP-conjugated secondary antisera (Sigma-Aldrich).

Exponentially growing cells were depleted of methionine and cysteine for 40 min, pulsed with 25 μCi ³⁵S-labeled methionine/cysteine for 15 min, and chased in complete medium for 0, 2, 4, or 6 h. Cells, supernatants, and spherons were collected at the indicated time points. Cells and spherons were treated with lysis buffer. IgM from cell lysates, supernatants, and spheron fractions was immunoprecipitated and analyzed by SDS-PAGE.

Cells (1 × 10⁵) were grown on glass cover slips for 24 h. The cells were metabolically arrested with 2% sodium azide in RPMI 1640. The cover slips were washed and incubated in 0.2 U/ml chondroitinase ABC (Seikagaku, Tokyo, Japan) or buffer alone for two 24-h digests at 37°C. Alternatively, cells were passaged into media containing 100 μM para-nitrophenyl-β-d-xyloside (Sigma-Aldrich) and incubated at 37°C for 48 h.

Dot blots were used for direct binding of IgM to chondroitin sulfate, heparan sulfate, and heparin (Sigma-Aldrich) on SPC Nytran membranes (Whatman, Piscataway, NJ). Hyaluronic acid-binding protein (Sigma-Aldrich) was coated onto nylon to capture hyaluronic acid followed by incubation with IgM for 2 h. IgM binding was detected by alkaline phosphatase-conjugated anti-mouse IgM.

Supplemental Video 1 depicts spherons at the cell surface in live cells. The composition of IgM from cell lysates and supernatants analyzed by two-dimensional PAGE is shown in Supplemental Fig. 1, binding of single-chain variable fragment (scFv) to glycan microarrays in Supplemental Fig. 2, and histopathology of mouse kidneys in Supplemental Fig. 3.

The distribution of IgM in hybridomas expressing VH56R μ H chain paired with Vκ38c (VH56R/Vκ38c) differs from the IgM distribution in cells expressing VH56R with other L chains in that it appears more concentrated in the Golgi (7). A significant portion of VH56R/Vκ38c IgM localized to the perinuclear region of the cells (Fig. 1A, 1C, hatched circle), whereas IgM composed of VH56R paired with Vλx (Fig. 1B, 1D, hatched circle) was distributed throughout the cytoplasm but was less prominent in the perinuclear region. This observation suggested that intracellular IgM distribution reflects characteristics of the Vκ38c L chain.

To test whether Vκ38c expression promotes a centralized IgM localization when paired with a different μ H chain, we took advantage of 38C-13, an IgM murine B cell lymphoma that expresses Vκ38c and has a secretion defect (24). Again, there was significant accumulation of IgM in the perinuclear region of 38C-13 cells (Fig. 1E), thus suggesting that Vκ38c does not require VH56R to induce this distinct IgM accumulation pattern. When IgM expression was examined in the WEHI-231 lymphoma that expresses an unrelated VH/Vκ pair (25), we found that IgM was predominantly localized to the periphery of the cell (Fig. 1F).

To pinpoint the cellular compartment where VH56R/Vκ38c IgM accumulates, we performed quantitative colocalization of IgM and proteins that have well-characterized functions in the secretory pathway (Fig. 2). We used Abs to β-COP, a protein that plays a key role in vesicular transport between the endoplasmic reticulum (ER) and the cis-Golgi (26), and Abs to giantin and golgin-95, two proteins that provide docking sites in the cis-Golgi for vesicles from the ER (27). We also compared the location of IgM to mannosidase II, an enzyme that processes high mannose N-linked glycans in the medial Golgi (28). Additionally, we visualized the medial Golgi with fluorescent G. simplicifolia lectin II, a protein that binds terminal N-acetyl-d-glucosamine (GlcNAc) sugars (Fig. 2A).

Orthogonal sections were used to outline the compartments identified by the markers. In combination, between 16 and 22% of intracellular VH56R/Vκ38c IgM colocalized with the various Golgi complex markers, whereas <2% of VH56R/Vλx IgM showed colocalization with any of the markers (Table I). Up to a third of the Golgi-associated IgM (22–32%) was colocalized with markers for the ER to Golgi and cis-Golgi compartments (i.e., β-COP, giantin, and golgin-95), and about two thirds (66–71%) localized to the medial Golgi compartment (i.e., mannosidase II and G. simplicifolia lectin II). These data indicated that the VH56R/Vκ38c IgM accumulates upon entry into the Golgi, and that the cis- to medial-Golgi contains approximately a fifth of total cellular IgM.

It is unusual to observe accumulation of IgM in the Golgi compartment (29, 30). To confirm that IgM was associated with this organelle, we took advantage of Golgi fragmentation, a process that is induced by treatment of cells with nocodazole (31). Following nocodazole treatment, Golgi fragments displayed colocalization of golgin-95 and VH56R/Vκ38c IgM (Fig. 2B). In contrast, little to no overlap between golgin-95 and VH56R/Vλx IgM was observed (Fig. 2B). These results indicated that VH56R/Vκ38c IgM is retained in the Golgi even after the Golgi architecture is disrupted by nocodazole. Accumulation of VH56R/Vκ38c IgM in the Golgi compartment could reflect aspects of IgM assembly or IgM interactions with a resident component in the lumen of the Golgi. To distinguish between these alternatives, we analyzed extracellular forms of VH56R/Vκ38c IgM.

Despite the accumulation of VH56R/Vκ38c IgM in the Golgi compartment, we could observe VH56R/Vκ38c IgM clusters at the cell surface (Fig. 1A, 1C). To learn whether the VH56R/Vκ38c IgM clusters detach from the cell surface, we grew the VH56R/Vκ38c hybridomas on cover slips. Z-stack images (Fig. 3A) revealed IgM clusters at the cell surface (Fig. 3B–G) and spherical IgM deposits on the cover slip (Fig. 3H). We refer to these IgM deposits as spherons.

Although IgM spherons ranged up to 2 μm in diameter, and thus could be observed by conventional light microscopy (Fig. 3I), most spherons measured <1 μm (Fig. 3H) and were thus just at or below the resolution limit of light microscopy. To estimate the rate of IgM spheron production, we grew cells on cover slips for 16 h and analyzed them by confocal microscopy. We observed ∼200 spherons near the base of an individual hybridoma cell, suggesting that, on average, a cell produces at least 12 spherons per hour. The association of spherons with the cell membrane and their release from the cells were examined over a period of 20 min by live cell fluorescence microscopy. We observed numerous membrane-attached IgM clusters that were in lateral flux and rarely detached from the cell surface (Supplemental Video 1), suggesting that clusters remain attached to the B cell surface.

Spherons represent unexpectedly large complexes of extracellular IgM. Thus, we asked whether cells package spherons into membrane-bound IgM vesicles. To test whether IgM spherons are contained inside phospholipid membranes, we exposed live cells to anti-IgM Abs (red) and to DiO, a fluorescent dye used to detect phospholipids (green). Additionally, we used Sytox Orange, a dye used to visualize DNA (blue). Anti-IgM reacted with IgM surface clusters, whereas DiO stained membranes beneath and adjacent to the clusters (Fig. 4A–F). On cover slips, spherons did not react with DiO or with CM-DR (Fig. 4G), an additional phospholipid-binding dye, suggesting that spherons do not contain appreciable amounts of phospholipids. These data argue that spherons are not IgM-containing vesicles. Moreover, spherons were clearly distinguishable from apoptotic bodies. In apoptosis, cells package fragments of the apoptotic nucleus in membrane protrusions (blebs) that detach from the remainder of the cell. Apoptotic bodies reacted with CM-DR but rarely with anti-μ (Fig. 4G), whereas spherons did not react with CM-DR.

Transmission electron microscopy of hybridoma cells and isolated spherons (Fig. 4H–K) revealed that the surface IgM clusters have relatively homogeneous electron density and that they do not display an obvious external boundary (Fig. 4L). Anti-μ bound near the periphery of paraformaldehyde/glutaraldehyde-fixed spherons, possibly because the Ab was unable to penetrate the spherons’ core (Fig. 4J). Scraping spherons from cover slips disrupted their structural integrity, thus revealing a fibrous internal structure to which anti-μ was able to bind (Fig. 4K).

Because insoluble Ig aggregates that display a fibrillar ultrastructure are typical of amyloid (32), we exposed spherons to the amyloid-selective dye thioflavin T. Thioflavin T-labeled spherons emitted fluorescence that precisely coincided with the location of spherons (Fig. 4L, 4M), suggesting that spherons consist of amyloid-like material. Additionally, pockets of undisrupted spherons collected by scraping and centrifugation stained for GAGs by colloidal iron (Fig. 4N).

IgM H and κ L chains were the most abundant proteins in the recovered spherons, as assessed by SDS-PAGE followed by Coomassie blue staining (Fig. 4O) and Western blot analysis (Fig. 4P). Spherons were recovered from hybridomas expressing Vκ38c along with VH56R (Fig. 4O, 4P, lane II). In contrast, the VH56R/Vλx hybridoma did not produce IgM spherons (Fig. 4O, lane I). This and following experiments also examined VH76R/Vκ38c (Fig. 4O, 4P, lane III). The VH76R is identical to VH56R except for a serine to arginine replacement at position 76 of framework 3. The VH76R/Vκ38c IgM also accumulates at a perinuclear location (7) and is released from cells as spherons (Fig. 4O, 4P, lane III).

To characterize VH56R (or VH76R)/Vκ38c IgM assembly intermediates and examine their relation to spherons, we labeled cells in culture with [³⁵S]methionine/cysteine. Supernatants were separated from cells and from spherons by centrifugation. Addition of NP-40 detergent to cell pellets yielded soluble lysates and NP-40–insoluble material.

IgM polymers were resolved from hemimers (HL) and monomers (H₂L₂) by nonreducing SDS-PAGE (Fig. 5A). All three hybridoma lysates contained similar ratios of IgM polymers to assembly intermediates, suggesting that initial IgM assembly did not dramatically differ in the three hybridomas. Furthermore, SDS-PAGE under reducing conditions revealed that each fraction exhibited a constant stoichiometry of H to L chains (Fig. 5A). Two-dimensional SDS-PAGE of the IgM isolated from cell lysates (Supplemental Fig. 1) confirmed that VH56R and Vκ38c associate efficiently.

Nonetheless, the two Vκ38c hybridomas differed from the Vλx hybridoma in three remarkable ways (Fig. 5A). First, 10–20% polymeric IgM present in Vκ38c hybridomas was NP-40–insoluble, whereas all of the VH56R/Vλx IgM was extractable by the nonionic detergent. Second, VH56R/Vκ38c and VH76R/Vκ38c secreted IgM mostly as hemimers and monomers, along with small amounts of soluble IgM polymers. In contrast, supernatants from VH56R/Vλx-producing cells revealed that, as expected, IgM was secreted as polymers, with only negligible amounts of hemimers and monomers. Third, the IgM isolated from spherons of Vκ38c hybridomas consisted of polymers with little or no hemimers or monomers (Fig. 5A). The secretion of IgM hemimers and monomers from Vκ38c hybridomas was unexpected because free IgM monomers are usually retained in the ER by ER-to-Golgi quality control mechanisms (33). The incomplete assembly of monomers into polymers and the segregation of soluble IgM hemimers and monomers from insoluble IgM polymers prompted us to analyze H chain glycosylation in Vκ38c hybridomas.

Because IgM monomers assemble into polymers prior to the processing of complex oligosaccharides in the Golgi, we assayed VH56R glycosylation in VH56R/Vκ38c and VH56R/Vλx hybridomas (Fig. 5B). IgM isolated from different cellular fractions was treated with Endo H. Endo H cleaves high-mannose sugars that are added in the endoplasmic reticulum but it does not cleave complex oligosaccharides that are processed in the Golgi. As a control, we exposed cell fractions to PNGase F, an enzyme that cleaves both forms of N-linked sugars.

Consistent with the rapid secretion of IgM following terminal sugar processing, VH56R/Vλx lysates contained only small amounts of Endo H-resistant μ H chain sugars (Fig. 5B). In contrast, VH56R/Vκ38c lysates contained significant quantities of Endo H-resistant μ-chains, suggesting that although complex oligosaccharides were processed on VH56R/Vκ38c, these IgMs were inefficiently secreted. Additionally, VH56R/Vκ38c supernatants and spherons contained mostly Endo H-resistant H chains, indicating that complex oligosaccharides were processed on VH56R/Vκ38c hemimers, monomers, and polymers, and that spheron IgM had passed through the Golgi.

To determine whether the VH56R/Vκ38c polymers represent IgM pentamers, we examined whether J chain was incorporated into IgM. J chain addition is a late step in the assembly of IgM pentamers (34). Western blots of VH56R/Vκ38c spherons and of secreted VH56R/Vλx IgM demonstrated anti-J chain immunoreactivity, whereas the VH56R/Vκ38c culture supernatants showed no detectable reactivity (Fig. 5C). Therefore, the anti-J chain Western blots, combined with the Endo H analysis, indicated that VH56R/Vκ38c cells produce IgM with mature carbohydrate modifications, and that spherons incorporate J chain-containing IgM pentamers.

Pulsed incorporation of radioactive amino acids into IgM assembly intermediates and IgM polymers confirmed that both VH56R/Vκ38c and VH56R/Vλx B cell hybridomas assembled IgM into higher order polymers that were readily detectable by 2 h chase (Fig. 6A, compare lysates on nonreducing SDS-PAGE). However, between the 2- and 6-h chase period, cell-associated IgM polymers declined in the VH56R/Vλx hybridomas, whereas IgM polymers remained fairly constant in VH56R/Vκ38c cells. Additionally, VH56R/Vλx cells efficiently secreted polymers, the usual form of secreted IgM, whereas VH56R/Vκ38c cells primarily secreted IgM monomers and hemimers (Fig. 6A, compare supernatants on nonreducing SDS-PAGE). After 4 h chase, labeled IgM polymers appeared in spherons that were released by the VH56R/Vκ38c hybridoma, but the spherons did not contain detectable IgM hemimers and monomers (Fig. 6A).

The pronounced intracellular accumulation of polymeric IgM in VH56R/Vκ38c cells and the slow integration of this IgM into spherons, along with the intracellular accumulation of Endo H-resistant μ-chains (Fig. 6B), indicated a lag in the release of fully glycosylated VH56R/Vκ38c IgM from the Golgi. The preferential integration of IgM polymers into spherons, in parallel with the secretion of IgM hemimers and monomers, suggested that multivalency of VH56R/Vκ38c and VH76R/Vκ38c IgM contributes to the selective integration of polymers into spherons. Multivalency could favor IgM polymers over IgM hemimers/monomers for integration into spherons if spheron formation involves binding to a multivalent Ag.

IgM accumulation in the Golgi compartment may reflect the binding of IgM to Ags that are synthesized at that location. To test the possibility that the Golgi accumulation reflects IgM binding to GAGs (sugar polymers that are synthesized in the Golgi), we employed chondroitinase ABC, an enzyme that cleaves diverse glycan polymers. Addition of chondroitinase ABC to spherons on a cover slip reduced the average spheron size (Fig. 7A). Following treatment with enzyme, spherons >1 μm were nearly absent and, conversely, the number of smaller spherons increased. Enzyme treatments also increased the concentration of soluble polymeric IgM. These results suggested that chondroitinase treatment may compromise an important structural component of spherons.

Because GAGs are synthesized on serine residues of proteoglycans, we incubated cells with β-d-xyloside, an inhibitor of GAG addition to a protein backbone. In the presence of inhibitor, there were fewer spherons of any size than in the absence of inhibitor (Fig. 7B), suggesting that growth of spherons is sensitive to inhibition of GAG biosynthesis.

To gain more direct information about IgM binding to GAGs, we tested VH56R/Vκ38c and VH56R/Vλx scFv binding to glycan microarrays (Supplemental Fig. 2). The results of these studies showed that both scFv bound to GlcAβ1-3GlcNAcβ-Sp8, whereas VH56R/Vκ38c differed from VH56R/Vλx in that it was also able to bind to a biantennary Lewis A glycan, Galβ1–3(Fucα1–4)GlcNAcβ1–2Manα1–3(Galβ1–3(Fucα1–4)GlcNAcβ1–2Manα1–6)Manβ1–4GlcNAcβ1–4GlcNAcβ-Sp19. Because most GAGs do not have a direct counterpart on the microarray, the results provided only preliminary clues regarding the authentic Ag interactions.

Further insight into the role of GAG synthesis in the assembly of spherons was derived from VH56R/Vκ38c IgM binding to purified glycans. A direct binding assay showed avid binding of VH56R/Vκ38c IgM to heparin, reduced binding to chondroitin sulfate, and no detectable binding to the poorly sulfated heparan sulfate used in our assays (Fig. 7C). VH56R/Vλx IgM did not bind to any of the three glycans.

Heparin and heparan sulfate are close relatives that share a repeating disaccharide structure and differ only in the degree of sulfation: heparin is highly sulfated, whereas heparan sulfate, depending on tissue source, exhibits variable sulfation (35). Heparin is the exclusive product of mast cells (36). In contrast, B cells express the enzymes to produce highly sulfated heparan sulfate (37), and splenocytes indeed synthesize the most highly sulfated heparan of any tissue except the heart (38). The preference of VH56R/Vκ38c IgM for heparin over poorly sulfated heparan sulfate strongly suggested that VH56R/Vκ38c IgM can bind highly sulfated heparan sulfate in the Golgi of B cells.

Our assays also demonstrated VH56R/Vκ38c IgM binding to hyaluronic acid, whereas VH56R/Vλx IgM did not bind (Fig. 7D). Because hyaluronic acid contains sugars that are the building blocks of heparan sulfate (without the added sulfates), the binding of VH56R/Vκ38c IgM to hyaluronic acid suggests a contribution of glucuronic acid to binding. We conclude that GAGs that are synthesized in the Golgi (i.e., heparan sulfate and chondroitin sulfate) and hyaluronic acid that is produced at the cell surface contribute to the formation of IgM spherons.

In mice with the VH56R or VH76R anti-DNA H chain transgenes, the transgenic H chains are expressed in association with a limited set of L chains. The set is composed of L chains whose representation in the repertoire is skewed as a result of receptor editing. Editor L chains such as Vκ20 and Vκ21D use negatively charged aspartates to shield H chain arginines from DNA binding, thus counteracting the propensity of the H chain for DNA binding (39). The Vκ38c L chain is also an editor because it can replace a preceding L chain and thereby become overrepresented in the peripheral repertoire. However, Vκ38c does not edit VH56R binding to DNA or phosphatidylserine (7). Previously, we argued that expression of Vκ38c sequesters the BCR at an intracellular location, thereby reducing BCR surface expression during the crucial time when central tolerance eliminates other anti-DNA B cells (40). We likened Vκ38c-expressing B cells to “Trojan horses” (7) because they enter the peripheral tissues while concealing their pathogenic potential.

In this study, we examined the mechanism of BCR sequestration and observed that up to a fifth of intracellular IgM accumulates in the cis- to medial-Golgi of Vκ38c-expressing B cells (Fig. 2, Table I). This localization may be directly responsible for the inability of this autoreactive IgM to induce further receptor editing. To induce editing, surface BCR signals through phospholipase Cγ (41) and SLP65 (42) to activate the NF-κB pathway (43) and induce the IFN regulatory factor-4 transcription factor (44). SLP65 appears to play a particularly important role because receptor editing in anti-DNA B cells is impaired in SLP65 knockout mice (42). Whereas the SLP65 signaling adaptor is anchored at the cytoplasmic side of the plasma membrane (45), the VH56R/Vκ38c IgM accumulates in the Golgi (Fig. 2) and thus may not efficiently engage SLP65. Therefore, the VH56R/Vκ38c IgM may have a reduced ability to signal via Syk/SLP65 and induce additional rounds of V to J recombination.

The accumulation of VH56R/Vκ38c IgM in the Golgi may be a manifestation of the IgM binding to GAGs. This conclusion is supported by the results of scFv binding to glycan microarrays (Supplemental Fig. 2) and IgM binding to GAGs (Fig. 7C). According to this scenario, the accumulation of VH56R/Vκ38c IgM in the Golgi reflects interactions between GAGs that are synthesized in the Golgi and the IgM that transits the Golgi as part of the normal protein maturation process. Interactions with GAGs may slow the IgM transit, resulting in the IgM accumulation in the Golgi. The likely Ag recognized by VH56R/Vκ38c IgM is highly sulfated heparan sulfate, the B cell analog of heparin (Fig. 7C).

The binding of VH56R/Vκ38c IgM to GAGs in the Golgi is predicted to yield complex multivalent Ab–Ag structures. The avidity of multivalent IgM polymers for repetitive GAG epitopes may account for the fact that fully assembled IgM polymers are rendered largely insoluble upon transit through the Golgi (Fig. 5A). Conversely, the lower avidity of IgM monomers and hemimers for GAGs may result in their preferential secretion as the predominant soluble IgM species from VH56R/Vκ38c cells (Fig. 5A).

Interactions between GAGs and VH56R/Vκ38c IgM may therefore be central to the formation of spherons. This is implied by the reduction in the size of spherons following treatment with chondroitinase ABC, an enzyme that degrades GAGs (Fig. 7A), and by the release of fewer and smaller spherons from cells treated with β-xyloside, a cell-permeable inhibitor of protein glycanation (Fig. 7B). The binding of VH56R/Vκ38c IgM to hyaluronic acid suggests that spherons may increase in size and complexity by incorporating additional carbohydrate polymers while tethered to the B cell surface.

We think that many aspects of VH56R/Vκ38c IgM–GAG interactions that we observe in cell culture also occur in vivo. That spherons form in vivo is suggested by the autoimmune pathology of these mice (46). If the IgM deposits are derived from spherons, we would expect them to accumulate at sites that filter blood and display accessible GAG moieties. Consistent with this prediction, the glomerular basement membrane was an important site of IgM deposits in VH76R IgM mice (Supplemental Fig. 3). Additionally, we have observed that B cells from VH56R transgenic mice release spheron-like IgM complexes after 48 h culture in the presence of LPS (unpublished observations). However, clear evidence for the production of spherons in vivo requires the construction of Vκ38c transgenic mice. These efforts are currently underway.

In closing, we argue that the IgM spherons analyzed in this study represent a form of pathogenic ICs that arise as result of the proper functioning of immune tolerance. As such, spherons, or their in vivo counterparts, could represent ICs that arise in previously non-autoimmune individuals and deposit in tissues. Their Ag binding properties could make spherons into powerful promoters of IC deposition disorders and establish conditions for the further progression of immune pathogenesis.

We thank Profs. Joseph Haimovich and Ae-Kyung Yi for lymphoma cell lines, Drs. David Smith and Jamie Heinsburg-Molinaro for scFv analysis by the Glycan Microarray Core (Emory University, Atlanta, GA), and Kathy Troughton for expert assistance with electron microscopy. We acknowledge Janine Babulski of Applied Precision’s mobile laboratory for assistance with Delta Vision live cell microscopy, the Special Histology Laboratory, Department of Pathology, University of Tennessee Health Science Center, and Tim Higgins for help with the preparation of figures. Manuscript editing was provided by Dr. Cindy Benedict-Alderfer.

The authors have no financial conflicts of interest.