Identification of Cell Surface Straight Chain Poly-N-Acetyl-Lactosamine Bearing Protein Ligands for VH4-34–Encoded Natural IgM Antibodies

Antibodies encoded by the VH4-34 gene have been studied because of their inherent autoreactivity and unusual regulation. B lymphocytes bearing the VH4-34-BCR are overrepresented in the naive repertoire but counterselected for Ab secretion. Although nearly 5% of naive B cells use the V4-34-BCR circulating Abs are low to undetectable in healthy adults (1, 2). In fact, VH4-34-BCR bearing cells exhibit a surface profile of anergic cells, are actively policed away from germinal centers in healthy adults, and are primarily found in the marginal or mantle zone (3, 4).

Circulating VH4-34 Abs are readily detectable in limited clinical conditions, such as infectious mononucleosis (IM), AIDS, systemic lupus erythematosus (SLE), hepatitis C infection, EBV-associated nasopharyngeal carcinoma and a subset of cold agglutinin disease (5–11). The shared feature among these diverse syndromes is association with lymphotropic viruses and B cell hyperproliferation (6). In IM, a widespread disease caused by EBV, the secreted VH4-34 response is transitory and limited to the multivalent IgM and IgA disappearing with the resolution of the disease (12). No manifestation of long-term clinical autoimmunity is linked with brief secretion of high-avidity VH4-34 Abs approaching 400 μg/ml. Isotype-switched VH4-34 Abs are detected in >50% of SLE patients (13). The operative mechanism that restricts VH4-34-BCR bearing B cells from entering germinal center reactions is blocked, leading to the production of isotype-switched somatically mutated autoantibodies (3). Hence, secretion of VH4-34 Abs has been monitored to study the interplay between acute anti-pathogen response and breakdown in tolerance leading to autoimmunity, particularly in SLE.

In the context of autoreactivity, it is the recognition of a conserved carbohydrate epitope on human adult or fetal RBC and B lymphocytes that sets VH4-34 Abs apart from the typical natural autoantibodies. The germ-line nonmutated framework region of the VH4-34 H chain, independent of the L chain, is accountable for binding linear/straight chain poly-N-acetyl-lactosamine (SC-PNAL) or i-antigen on human fetal/nucleated RBC and human B lymphocytes. Despite the preponderance of circulating i-antigen/SC-PNAL on fetal RBCs, B lymphocytes bearing VH4-34 BCR are positively selected in human fetal immune repertoire. This scenario represents an unprecedented selection of an autoreactive BCR in the presence of abundant autoantigen, making the VH4-34 gene a rare example where expression of surface versus secreted Ig is stringently regulated. In contrast, the branched SC-PNAL chains or i-antigen is present only on adult human RBCs and are not bound by anti–B cell germ line–encoded VH4-34 IgM Abs (14).

The anti-i/anti–B cell SC-PNAL binding VH4-34 Abs, range from high to low binders, with the high binders demonstrating B cell cytotoxicity via membrane perturbations. Such VH4-34 Abs are observed to bind or paint B lymphocytes in patients with IM, SLE, or HIV (15, 16). On RBCs, SC-PNAL is present on both glycolipids and glycoproteins (17). The nature of the biological backbone that carries the sugar epitope on B lymphocytes has not been studied in detail. Earlier work with VH4-34-IgG has identified CD45 protein as a carrier of SC-PNAL (18). In this study, we show that SC-PNAL is present on CD147 and other proteins known to associate with it, such as CD98. This membrane super complex plays a central role in cell proliferation, transport, and energy metabolism. SC-PNAL expression on CD45 is detected only on peripheral blood B lymphocytes and not on human B cell lines, suggesting differential glycosylation patterns. Because glycan moieties have crucial functions from protein folding/adhesion to signaling, differential expression of SC-PNAL on proteins in different B cell types may be linked to the intricate control of VH4-34 Ab secretion.

The Consortium for Functional Glycomics (CFG) mammalian glycan microarray version 4.2 at Emory University School of Medicine (Atlanta, GA) was used for binding studies. Version 4.2 has >500 glycans on the microarray (19). Four mAbs were tested on the array at 100 μg/ml and detected with fluorescently labeled anti-human IgM Ab. Z21 and Z2D2 were two VH4-34 encoded anti–B cell binding IgM mAbs. Z15 was VH4-34 encoded non-B cell binding IgM mAb, and MS2B6 was non–VH4-34 encoded non-B cell binding IgM Ab. Standard CFG procedures and results can be accessed at www.functionalglycomics.org under CFG request number 2112.

Glycans on a ceramide backbone were purified as described (20, 21), separated on TLC, and immunostained as described (22, 23). Plate was air dried and blocked with PBS with 5% BSA for 2 h and exposed overnight with mAb 216 (5 μg/ml). After washing five times with 0.5% BSA/PBS, the plate was incubated for l h with ¹²⁵I-labeled anti-human IgM (Zymed Laboratories, South San Francisco, CA). After 10 washes with PBS, the plate was then air-dried and subjected to autoradiography. Total staining of glycan-ceramides was performed by orcinol-H₂SO₄ as described (20, 21). Glycan-ceramide purification and TLC studies were performed by Dr. Sen-itiroh Hakomori (University of Washington, Seattle, WA).

VH4-34 gene–encoded MAb 216, Z21, Z2D2, and A6H4C5 have been described previously (14). VH4-34–encoded mAb Y2K was prepared in our laboratory by fusing the heteromyeloma SHMD33 with human fetal splenocytes obtained from medically approved pregnancy termination with the approval of the Committee for the Protection of Human Subjects at Stanford University. IGM-55.5 is a recombinant monoclonal VH4-34 encoded IgM engineered to have similar Ag specificity as mAb 216 and restricted polyreactivity (24). All VH4-34 mAbs, incuding Y2K and 55.5, are VH4-34 germline encoded with distinct VH-CDR3 regions. The three human IgM isotype controls used were myeloma protein (Thermo Fisher Scientific, #31146), a VH3 gene–encoded previously described mAb MS2B6 (14, 25) and another VH3 gene–encoded previously described polyreactive mAb, B314/3 (26, 27). This later mAb binds ssDNA, cardiolipin, and other autoantigens but does not bind human B cells. The source and concentration of enzymes and chemicals are tunicamycin (5 μg/ml, #T-7765; Sigma), swainsonine (1 μg/ml, #S-8195; Sigma), benzyl-α-GalNac (5 mM, B-4894), endo-β-galactosidase (Amsbio #100455-1, 10 mU; Seikagako Corporation), cytochalasin-D (σ, #C-8273; 5 μg/ml), mycalolide B (#T123-0020, 0.1 μM; Enzo Life Sciences), Jasplakinolide (#J-7473, 1 μg/ml; Molecular Probes).

d-Threo PDMP (1-phenyl-2-decanoylamino-3-morpholino-1-propanol, HCl) is a ceramide analog that blocks glucosylation of ceramide by inhibiting glucosylceramide synthase reducing cellular synthesis of complex glycosphingolipids; it is a useful tool for studying the effects of cellular glycosphingolipid depletion. PDMP was dissolved in ethanol (sc-280659; Santa Cruz Biotechnology, Santa Cruz, CA). Fumonisin B1 (FB1), a mycotoxin produced by Fusarium moniliforme, is an inhibitor of ceramide synthase (sphingosine/sphinganine N-acyltransferase), blocking cellular formation of ceramides, sphingomyelins, and complex glycosphingolipids. FB1 was dissolved in DMSO (sc-201395; Santa Cruz Biotechnology). Nalm-6 and Daudi cells were incubated at 2.5 × 10⁵ cells per milliliter in IMDM with 10% FCS. FB1 or d-threo PDMP was added at the appropriate concentration on three consecutive days. Ethanol- and DMSO-only controls were set up in parallel. Cells were stained and analyzed by flow cytometry on day 4 as described. Viability as determined by propidium iodide uptake was >70% for PDMP and >80% for FB1-treated cells at the highest concentration of the inhibitor.

Cells were stained at RT with Abs, anti–CD19-APC (#555415, BD Pharmingen, San Diego, CA), anti–CD17-FITC (#166-040; Ancell, Bayport, MN), anti–CD77-FITC (#551353; Becton-Dickinson, San Jose, CA), cholera-toxin-FITC (C-1655, Sigma) according to the manufacturer’s instructions. Cells were analyzed with an LSR II (Becton Dickinson) using Cell Quest for data interface and FlowJo (Treestar, Ashland, OR) software for analysis. For detergent insoluble fraction (DIF), B cell lines were stained with primary Abs, mAb 216, CD19, or anti-human IgM Abs and secondary Abs anti-human λ-FL or anti-mouse IgG-FL using standard protocol and then incubated with 0.5% Triton-X-100 in PBS for 10 min. Cells were washed two times with staining media (PBS plus 3% FCS plus 0.05% NaAz) and analyzed with FACS. Stained control cells not solubilized with Triton-X-100 were run in parallel.

Human cell lines Nalm-6 (28), OCI-Ly8 (29), Reh (30) and Peer (31) were maintained in logarithmic growth phase in IMDM media with 10% FCS (heated to inactivate complement), washed two times with PBS and lysed on ice with M-PER mammalian protein extraction reagent (#78503; Thermo Scientific, Waltham, MA) supplemented with protease inhibitors (Roche Diagnostics #04 693 124 001). Human buffy coats were obtained with the approval of the Committee for the Protection of Human Subjects at Stanford University from Stanford Blood Bank and mononuclear cells were purified using standard Ficoll procedure. B lymphocytes were purified by negative selection using B Cell Isolation Kit II (#130-90-151; Miltenyi, San Diego, CA), and MPER extracts were prepared. B lymphocytes were >90% pure. M-PER extracts were centrifuged at high speed and lysates stored on ice for use within the same day or stored at −20°C until use. CaptureSelect IgM Affinity Matrix (Life Technology, Carlsbad, CA; BAC BV, #289005) was washed three times with PBS and loaded with human mAbs of the IgM isotype (216, Y2K, Z21, Z2D2, 55.5, B314/3, MS2B6, or pierce IgM). Thiry microliters of packed beads were incubated with 300 μg mAb for 30 min at room temperature (RT) on a shaker. The beads were washed five times with PBS and loaded with cell M-PER extract diluted 1:5 in PBS with protease inhibitors for 2 h at RT on a shaker. The affinity matrix was washed five times with PBS and eluted with 3% SDS with 2.5 mM EDTA in PBS. SDS was removed with a detergent removal column (Thermo Fisher Scientific, #87777). The CD147-CD98 complex was not detected when beads were eluted with low-pH glycine buffer.

For the second method of IP, human B cell lines (2 × 10⁷ cells, Nalm-6, OCI-Ly8, Reh) and purified B lymphocytes from human mononuclear cells were washed one time with PBS and incubated on ice for 20 min with VH4-34–encoded mAbs (30 μg, 216, Y2K, 55.5, or Z2D2). Controls with no Ab or isotype control (MS2B6 or Pierce IgM) were set up in parallel. Cells were washed with cold PBS three times before extraction with 0.5% NP40 supplemented with protease inhibitors (Roche Diagnostics, Indianapolis, IN; #04-693-124-001). Extracts were centrifuged at high speed (microfuge at 15,000 rpm), and the detergent-soluble extract was stored on ice until use. The detergent-insoluble pellet was washed one time with PBS and incubated for 45 min at RT in PBS with DNase-I, DNase-II, and RNase A (25 μg/ml; Sigma-Aldrich; DN-25, D-8764, and R-5000 respectively). Solubilization of nuclear cytoskeletal associated proteins by DNase and RNase treatment has been described previously (32–35). The solubilized proteins were recovered by centrifugation (microfuge at 15,000 rpm for 10 min) and termed DIF. Both detergent-soluble and insoluble extracts from VH4-34 Ab treated and control specimens were diluted 1:5 in PBS with protease inhibitors and loaded on prewashed CaptureSelect IgM Affinity Matrix (BAC BV, #289005; Life Technology). Loaded resin beads were rocked for 2 h at RT, washed five times with PBS and eluted with 3% SDS in PBS with 2.5 mM EDTA. SDS from eluted proteins was removed by detergent removal column (Thermo Fisher Scientific, #87777).

IP of CD147 was performed by loading anti-CD147 Abs, 8D6 (sc-21746; Santa Cruz Biotechnology) or HIM-6 (#306206, BioLegend, San Diego, CA) on protein-G-agarose beads (#22851; Thermo Scientific) followed by M-PER extracts of Nalm-6 or Peer. Tomato-lectin agarose beads (Lycopersicon esculentum, AL-1173; Vector Labs, Burlingame, CA) were loaded with protein, rocked at RT for 2 h, and bound proteins eluted as described above.

Extracts were separated by SDS-PAGE on 4–12% gradient gels (Novex NP0322BOX; Life Technologies) under reducing conditions and transferred onto PVDF membranes using standard procedures. Membranes were blocked with 5% milk in PBS for 1 h before immune blotting with primary Abs, followed by secondary anti-mouse IgG HRP diluted 1:2000 in PBS (Cell Signaling #7076S). Blots were developed using ECL reagents (Thermo Scientific #1859674 and #1859675). The primary Abs and reagents were anti-CD147 (F-5, Santa Cruz sc-374101; HIM6, Biolegend #306206), anti-CD45 (H130, Biolegend #304002; 35-Z6, Santa-Cruz sc-1178), anti-CD98 (E-5, Santa Cruz sc-376815), anti-CD298 (Santa Cruz, sc-135998), ASCT2 (V501, Cell Signaling #5345S), and tomato lectin (TL)-biotin (L. esculentum, B-1175; Vector Labs, Burlingame, CA). The ASCT2 was detected using anti-rabbit IgG HRP (Promega, W401B), and TL-bio was detected using avidin-HRP (#18-4100-94; eBioscience, San Diego, CA).

All solution samples were precipitated with 4-fold the volume of −80°C acetone on dry ice for at least 1 h. Following centrifugation at 4°C, 12,000 × g for 10 min, the supernatant was removed, and the samples were speed vacuum-dried for 5 min. The protein pellet was reconstituted in 15 μL 8 M urea and 20 μL 0.02% of acid labile surfactant protease max (Promega), 50 mM ammonium bicarbonate. The samples were reduced using DTT (5 mM) at 50°C for 30 min, followed by alkylation using propionamide (10 mM) for 30 min at room temperature. Sample volume was adjusted to less than 1 M Urea by the addition of 50 mM ammonium bicarbonate (pH 8.0) and trypsin (Promega) was added to a protein to protease ratio of 50:1, respectively. The samples were digested overnight at 37°C, followed by an acid quench using 10 μL 10% formic acid water. Peptides were stage-tip purified (NEST group) on C18 columns, where the eluent was dried and stored at 4°C.

The LC (liquid chromatography) was either a Proxeon easyLC or an Eksigent nanoLC 2D run at 300 or 600 nL per minute flow rates, respectively. The analytical column was packed in house using 3 μM C18 reversed-phase particles (Peeke Scientific) at 15 cm in length. LC acquisition time varied from 60- to 90-min gradients. The mass spectrometer was a LTQ Orbitrap Velos set to acquire in data-dependent acquisition fashion where the top 12 or 15 most intense precursor ions were selected for fragmentation in the ion trap. The .RAW data were converted to mzXML format and searched using SEQUEST. Modifications were propionamide (+71) as static and methionine oxidation (+16), and serine, threonine, and tyrosine phosphorylation (+80) were variable modifications. The tolerances were 20-ppm precursor mass error, searched in a target decoy approach where the false discovery rate (FDR) was set to 1%. Further filtering was done to require a minimum of three peptides per protein assignment, rendering the true FDR to <0.5%.

In Scaffold (version 4.0.4; Proteome Software, Portland, OR), peptide spectral counts were used as a semiquantitative measure with peptide and protein thresholds at 95.0% and 99.0%, respectively, with a three-unique peptide criterion to report protein identification. For all experiments, the data were categorized to include control IgM and VH4-34 IgM IP, and a Fisher exact test was applied to the two relevant categories. Fisher exact test generates a p value based on the nonrandom association between two categorical conditions.

Many previous studies have reported that VH4-34 gene encoded Abs bind human B lymphocytes via SC-PNAL (36–38). Our analysis of 24 independently derived VH4-34 IgM mAbs showed mAb 216 as the strongest binder, with mAb Z21 a close second followed by mAb Z2D2 as one the top B cell binders (14). Binding of SC-PNAL by mAb 216 was confirmed by TLC. MAb 216 does not bind paragloboside and GM3 (Fig. 1B, lanes 1, 2, 7), but binds SC-PNAL and sialyl-SC-PNAL (Fig. 1B, lanes 3, 4). The orcinol-H₂SO₄ staining depicts the total glycan loaded per lane (Fig. 1A). The specificity of the sugar ligand for other representative VH4-34–encoded IgM mAbs, Z21 and Z2D2 was confirmed by glycan array (Fig. 1C). The CFG Array 4.2 containing 501 glycans was used. Number 329, the strongest glycan bound by mAbs Z2D2 and Z21, is sialyl-SC-PNAL; it is identical in structure to the glycan bound by mAb 216, loaded in lane 4 (Fig. 1B). These data show that a minimum of 3 U of N-acetyl-lactosamine is required for good binding by VH4-34–encoded IgM mAbs.

On RBCs, branched PNAL (I-antigen) and unbranched (i-antigen or linear or straight chain/SC-PNAL) are found on both lipids and proteins (39, 40). We tested to determine whether the i-antigen–like SC-PNAL ligand on human B cells was also present on both glycolipids and glycoproteins. SC-PNAL expression was assessed after treatment with inhibitors of N- and O-linked protein glycosylation. Tunicamycin, swansonine (N-linked inhibitors), and Benzyl-GalNac (O-linked inhibitor) significantly downregulated expression of SC-PNAL, whereas papain, a protease, did not alter SC-PNAL expression, suggesting its presence on glycoproteins (Fig. 2A). Removal of the terminal sialic acid by neuraminidase dropped SC-PNAL binding by VH4-34 mAbs but did not abolish it, whereas endo–β-galactosidase, an enzyme that specifically cleaves the β1-4 linkage within SC-PNAL, decreased the binding significantly (Fig. 2A). Next, expression of SC-PNAL was tested on human B cells lines (Nalm-6 and Daudi) treated with methanol, because it solubilizes a subset of lipids and fixes proteins simultaneously. Comparison of methanol-fixed cells and control cells showed no change in expression of SC-PNAL (Fig. 2B), whereas there was decrease in glycolipid ligands, CD77 (neutral glycolipid globotriaosylceramide, Gb3), and CD17 (lactosylceramide).

We then inhibited lipid glycosylation using two specific inhibitors that block independent enzymes involved in glycolipid synthesis. Cells were treated with FB1 (41, 42) and d-threo PDMP (43, 44) as described in 2Materials and Methods and analyzed for expression of glycosylated epitopes with flow cytometry. Fig. 2C shows that expression of three different glycolipid ligands (CD77, CD17, and cholera toxin ligand GM1) was inhibited in a dose-dependent manner by both inhibitors, but expression of SC-PNAL ligand was not altered. Combined, these results suggest that the SC-PNAL ligand bound by VH4-34-IgM anti–B cell Abs is present predominantly on protein carriers.

Many membrane receptors, including surface Ig (BCR), are known to undergo a physical association with the underlying cytoskeleton following ligand or Ab cross-linking (45–50). This interaction has been studied extensively using nonionic detergents known to dissolve membrane proteins preserving a DIF. Resistance to solubilization by mild detergents, such as Triton X-100, is dependent on intact cytoskeletal structures such that depolymerization of actin disrupts the association (51–53). As projected, cross-linking of surface BCR on OCI-Ly8 cells led to 50% BCR to codistribute to the DIF (49, 50). ligands, such as CD19, did not fractionate to DIF. Binding of SC-PNAL on normal peripheral B lymphocytes or human B cell lines leads to partitioning of >95% of the cell bound anti–B cell VH4-34 IgM mAb into the DIF (Fig. 3A). An intact cytoskeleton is essential for the DIF partitioning. When Nalm-6 cells where treated with actin depolymerizing agents, cytochalasin-D or mycalolide B association with DIF was abolished. In contrast, when treated with Jasplakinolide, an actin-polymerizing agent, the association with DIF remained intact (Fig. 3B).

We used this partitioning to trace the SC-PNAL carrier in the DIF. Cells were treated with VH4-34 or control IgMs, and detergent-soluble and insoluble extracts prepared and loaded on to anti-human IgM affinity matrix as described in 2Materials and Methods. Immunoprecipitated material was analyzed by nano-electrospray ionization mass spectrometry (Nano-ESI-MS) and proteins specifically present in the DIF of VH4-34 IgM-treated samples were identified with an FDR < 0.5% (Supplemental Table I). The 22 proteins, pooled from three independent experiments, are shown in Table I. Nine of 22 proteins are membrane proteins, and seven are glycosylated membrane proteins. In samples treated with control IgM or samples receiving no treatment, the CD147-CD98 complex is not detected in the DIF. The complex partitions into the DIF only in VH4-34 IgM-treated samples. In cells not treated with VH4-34 IgM, the CD147-CD98 complex is soluble and fractionates with the membrane fraction.

A standard IP protocol for SC-PNAL carrier identification was also implemented. Anti-human IgM affinity matrix beads were loaded with VH4-34 or control IgM mAbs and then incubated with membrane extracts of Nalm-6 cells. In the prior method, the VH4-34 or control mAbs were on the cells; in contrast, in this procedure the VH4-34 or control mAbs are loaded on the immobilized matrix. Immunoprecipitated material was analyzed by Nano-ESI-MS, and proteins were identified with an FDR < 0.5% (Supplemental Table II). Seventeen of the ninety-eight proteins were identified as membrane proteins or membrane-associated proteins as described by scaffold GO annotation (Table II). CD147 (BSG, basigin), CD98 (4F2, SLC3A2), CD298 (ATP1B3), and ASCT2 (SLC1A5) are glycosylated, whereas CD81 (54) and MCT1 (55) are nonglycosylated proteins. Four proteins—CD147, CD98, ASCT2, and CD298—were identified by both methods of IP. Glycosylated proteins CD71 and class II MHC identified in the DIF fraction (Table I) were absent in the standard IP method (Table II).

Tomato lectin (TL), a protein from L. esculentum, is known to have specific affinity for PNAL sugar residues (56, 57). Proteins immunoprecipitated from DIF were run on PAGE, and Western blot was performed with TL-bio and av-HRP (Fig. 4). The gel band stained with TL was eluted for Nano-ESI-MS (Supplemental Table III). The list of the top 28 proteins identified as membrane proteins or membrane-associated proteins by scaffold GO annotation includes CD147 (BSG, basigin), CD98 (4F2, SLC3A2), and ASCT2 (SLC1A5), but not CD298 (ATP1B3). CD298 has an m.w. of 35kD, and it is not detected with TL Western blot (Table III).

The Nano-ESI-MS analysis was performed with mAb 216 as a representative anti–B cell binding VH4-34-IgM and Nalm-6 as a representative B cell. To confirm similar reactivities, we performed Western blot analysis of immunoprecipitated material by multiple independently derived VH4-34 anti–B cell mAbs. Similar to mAb 216, four other VH4-34–encoded B cell binding IgM mAbs immunoprecipitated CD147 protein, as identified by two independent anti-CD147 Abs (Fig. 5A, 5B). However, the three isotype-control mAbs did not immunoprecipitate CD147. Peer, a γδ+ T-cell line with low to negligible SC-PNAL expression, was used as a negative control (Fig. 5C). Nalm-6 expresses both CD147 and SC-PNAL, whereas Peer expresses only CD147 but not the SC-PNAL glycosylation. CD147 is not immunoprecipitated by VH4-34-IgM mAbs from Peer, despite high expression of the CD147 protein.

When CD147 was immunoprecipitated using anti-CD147 Ab, 8D6, both Nalm-6 (Fig. 6A, lane 3) and Peer (Fig. 6B, lane 3) yielded a high glycoform (HG-CD147) and a low glycoform (LG-CD147) of CD147 as distinguished by m.w. However, mAb 216 immunoprecipitated only the HG-CD147 from Nalm-6. The LG-CD147 moiety from Nalm-6 is not immunoprecipitated by mAb 216. Peer also has both HG and LG-CD147, but none of the two moieties is immunoprecipitated by VH4-34 IgM implying the absence of SC-PNAL on both the CD147 moieties. This shows the specificity of the VH4-34-IgM binding and suggests that the presence of heavily glycosylated CD147 does not necessarily translate to SC-PNAL expression.

To confirm the presence of PNAL, a sequential IP was performed, wherein proteins immunoprecipitated from Nalm-6 DIF were immunoprecipitated again using TL-agarose beads. The double immunoprecipitated material was run on PAGE, and Western blot performed with anti-CD147 and anti-CD98 Abs. Both proteins were present in the double-immunoprecipitated material, confirming the presence of PNAL chains on the two proteins or alternately confirming the association between these two proteins, in which least one protein of the complex bears PNAL chains (Fig. 7A). The CD147-CD98 complex is also immunoprecipitated from other B cell lines, Reh, and OCI-Ly8 besides Nalm-6 (Fig. 7B).

CD45 on human B lymphocytes has been previously described (18) to express PNAL and bind VH4-34–encoded Abs of the IgG isotype. Previous studies failed to immunoprecipitate CD45 using VH4-34 IgM derived from cold agglutinin disease (18). Because Nalm-6 does not express CD45 (58, 59), we tested pre-B cell line Reh and normal human peripheral blood B lymphocytes that express both SC-PNAL and CD45 (Fig. 8A). OCI-Ly8 has low to moderate expression of CD45. All three B cell lines and human CD19⁺ B lymphocytes are SC-PNAL⁺. Similar to Nalm-6, VH4-34–encoded IgMs bind and kill Reh, OCI-ly8, and normal human peripheral blood/splenic B lymphocytes via membrane perturbation and partition into the DIF with mild detergent. CD147-98 complex is immunoprecipitated from both Reh (Fig. 7B) and purified normal B lymphocytes from healthy human buffy coats (Fig. 8B). However, CD45 is immunoprecipitated only from normal B lymphocytes, but not from Reh (Fig. 8B) by VH4-34–encoded IgM mAbs.

In this study, we report the CD147-CD98 complex carrying SC-PNAL as the ligand for VH4-34 anti–B cell IgM Abs using Nano-ESI-MS. Two independent IP techniques identified the same protein complex. In addition, the glycan specificity of the protein complex was confirmed by TL binding. These proteomics studies of endogenous proteins were performed without any chemical or photo–cross-linking of the Ab-ligand pair. Proteomics was validated by Western blot analysis of human B cell lines, OC1-Ly8 and Reh, and human peripheral blood B lymphocytes.

Selective association among the proteins identified in this investigation has been previously reported in multiple studies (60–63). Xu et al. (60) describe the physical association of the CD147-CD98 complex in ectopic-CD147 expressing HT1080 cells with covalent cross-linking of bound Ab accompanied by mass spectrometry. The third most abundant protein identified in their analysis is ASCT2, followed by LAT1, MCT1, Epcam, CD71, ATP1B3, ATP1A1, and integrins (CD29 and CD49). Concerns about nonspecific identification because of protein abundance and ectopic expression of proteins are addressed, showing specificity of the interaction in their study (60). Proteins identified in our analysis without ectopic expression or cross-linking confirm the physical interaction between CD147-CD98 complex in B lymphocytes. Proteins ASCT2, MCT1, CD71, ATP1B3, and ATP1A1 are also identified in our proteomics. B cell–specific proteins such as MHC class II proteins are immunoprecipitated together with the CD147-CD98 complex. Other abundant Nalm-6 proteins, such as CD49 or CD47, although heavily glycosylated, are not immunoprecipitated. Table IV lists the VH4-34-IgM ligands identified by Nano-ESI-MS with CD147, CD98, and ASCT2 detected in all three methods.

The CD147-CD98 complex is normally detected in the membrane soluble fraction when B lymphocytes or B cell lines are incubated with control IgM or with no treatment at all. The complex shifts to the DIF only when cells are treated with VH4-34 IgM. In the DIF procedure, it is conceivable that the CD147-CD98 complexes are brought in proximity by the pentameric VH4-34 IgM; however, the standard IP method also identifies the complex even though the membrane is solubilized prior to IP, confirming coassociation between the two proteins shown by previous studies. It is conceivable that SC-PNAL is carried on multiple proteins or a single protein in this complex. Identification of the specific protein within the complex that carries SC-PNAL has been difficult because VH4-34 IgM Abs do not recognize CD147-CD98 complex on two-dimensional Western blot. Moreover, knockdown of CD147 and CD98 with siRNA by electroporation had an antiproliferative effect and use of lipid-based transfection reagents, such as DMRIE-C or oligofectamine, yielded low transfection efficiency in suspension nonadherent B cell lines (data not shown). The antiproliferative effect of CD147 knockdown has been previously reported and reviewed (64, 65). In addition, the m.w. of immunoprecipitated proteins does not alter significantly following treatment with endo–β-galactosidase, possibly because deletion of terminal SC-PNAL chains does not alter the overall protein mass (data not shown). Nonetheless, the presence of PNAL glycans on CD147 and CD98 proteins has been described in other studies (66, 67). Our studies using TL also confirm PNAL on CD147 and CD98 proteins individually immunoprecipitated by Abs to CD147 or CD98, respectively, using protein-G (data not shown).

SC-PNAL, the molecular target of VH4-34–encoded Abs, was initially identified due to its susceptibility to endo−β−galactosidase, an enzyme that specifically cleaves β1-4 linkage between N-acetylglucosamine and galactose in a lactose unit (37, 68, 69). Since PNAL has been reported to be present on CD45 (LCA, leukocyte common Ag), in particular the 220kD B-cell specific isoform, it was suggested as the target for VH4-34 anti–B Abs (70, 71). Molecular evidence of CD45 as the carrier for SC-PNAL was obtained with VH4-34-IgG Abs from SLE patients (18). In this study, CD45 was identified as a VH4-34-IgM ligand only in human peripheral B-lymphocytes. Interestingly, CD45 was not immuno-precipitated from Reh despite abundant protein expression suggesting absence of SC-PNAL on its CD45; however the CD147-CD98 complex was easily identified in this cell line. Thus differential glycosylation of CD45 may explain some of the difficultly associated with its identification as a VH4-34-IgM ligand. Variation in protein isoform and glycosylation allows CD45, a tyrosine phosphatase, to regulate many signaling events relating to development, maturation, activation, differentiation, and apoptosis of B and T lymphocytes (72–74). In B cell malignancies, such as diffuse large-cell lymphoma, changes in CD45 glycan composition alter its susceptibility to galectin-based toxicity (75). These animal lectins, in particular Galectin-3, like VH4-34 IgM mAbs, recognize polylactosamine-enriched glycoconjugates, including tissue/glycan specific CD147 (76).

CD147, a ubiquitous protein, shows remarkable variation in size based on glycosylation (66). Its predicted m.w. is 28 kD, with glycosylated moieties ranging from 40 to 70 kD. A distinct feature of CD147 from various cells and tissues shows two glycoforms of the protein, HG-CD147 and LG-CD147, based on gel migration (77). Our studies show that only the heavily glycosylated moiety in human B cell lines and normal human peripheral blood B cells is SC-PNAL decorated. Although HG-CD147 is found in many cell types, not every HG-CD147 protein is recognized by VH4-34 IgM Abs, as exemplified by the γδ T cell line Peer, confirming that HG-CD147 is immunoprecipitated only when appropriately glycosylated with SC-PNAL. Three glycosylation sites have been identified within its Ig-like domains and are postulated to serve as a regulatory mechanism of CD147 function (78). Moreover, the nature of oligosaccharides and HG/LG ratio has been shown to affect many diverse physiologic processes central to CD147 from tumor invasion to embryo implantation (79). Our studies show that VH4-34–encoded IgM Abs bound only a particular glycan moiety on CD147; such reagents that distinguish CD147 based on its glycosylation flavor may be useful in unraveling the central role of CD147 in many cellular processes.

CD147-CD98 complex is intrinsically linked to energy metabolism, cell proliferation, and transport mechanisms used by the cell for survival. Although the CD147-CD98 complex is ubiquitous, SC-PNAL expression on this complex is specific to B lymphocytes, B cell malignancies, and certain epithelial cancers. Expression of SC-PNAL on epithelial cancers varies from high to low and may be a cross-reactive epitope emerging from the aberrant glycosylation linked to malignant transformation (79, 80). Non–template-based glycosylation is a coordinated, controlled process that leads to many complex glycan products. In this study, we show coordinated SC-PNAL glycan addition only on B lymphocyte–associated CD147-CD98 complex. The identification of mechanisms that leads to specific glycan patterns and cancer-associated glycosylation changes is crucial and can be exploited for therapeutic interventions. Targeting of glycans present only on cancer cells can lead to minimal toxicity, as exemplified by the phase I trial with VH4-34–encoded IgM 216 (81).

We thank the Stanford Shared FACS Facility for technical support and Dr. Bruce Keyt (IGM Biosciences, Mountain View, CA) for providing the IgM-55.5.

The authors have no financial conflicts of interest.