Cullin 3 Is Crucial for Pro-B Cell Proliferation, Interacts with CD22, and Controls CD22 Internalization on B Cells

B lymphocytes, which are part of the adaptive immune system, develop from pro-B to immature B cells in the bone marrow and leave through the blood stream into the periphery. There, mature B cells are then activated during an immune response. The activation step via the BCR is important to promote B lymphocyte proliferation and further differentiation (1). However, regulation of B cell activation is also crucial for a balanced immune system, as hyperactivation and dysregulation can lead to autoimmunity. To protect B cells from hyperactivity, different regulatory proteins and mechanisms exist. For example, inhibitory membrane receptors, like CD22, are involved in preventing B cell hyperactivation (2, 3). However, their expression has to be controlled and regulated as well to ensure a balance of activation and inhibition in B lymphocytes.

CD22 (Siglec-2) belongs to the sialic acid binding Ig type lectin (Siglec) family, which consists of several transmembrane proteins. These are mostly expressed on immune cells and share some structural features. CD22 is almost exclusively expressed on B lymphocytes and is present on the cell surface from the pre-B cell stage until mature stages (4). CD22 is known to be an important inhibitor on conventional (B2) B cells and negatively regulates BCR signaling (5–10). Its extracellular domain is responsible for binding the ligand, α2,6-linked sialic acid, via a conserved arginine residue (R130) (11, 12). The cytoplasmic tail of CD22 contains six conserved tyrosines (Y), of which three (Y783, Y843, and Y863) are located in ITIMs (13–15). After BCR activation, the inhibitory function of CD22 is mediated via phosphorylation of these ITIMs by the kinase Lyn (14). This in turn leads to SHP-1 binding and activation (16). Thus, SHP-1 inhibits the signaling complex of PLCγ2 (17) and eventually reduces calcium release from the endoplasmic reticulum (5, 7, 8). Additionally, CD22 is involved in activating the plasma membrane Ca²⁺ ATPase (PMCA) that further removes calcium from the cytosol (18, 19). Besides its inhibiting function, the endocytic capacity of CD22 has often been the focus of studies. On a molecular basis, it has been shown that anti-CD22 directed Ab-binding results in fast CD22 internalization via the clathrin-dependent pathway with rather slow recycling (20, 21). In this scenario AP50, a part of the clathrin-associated AP-2 adaptor complex, binds to the nonphosphorylated tyrosines of YXXØ (Ø is a hydrophobic residue) clathrin-internalization motif within the cytoplasmic tail of CD22 (Y843 and Y863) (21, 22). Moreover, in the steady-state CD22 internalization probably leads to lysosomal degradation (23). This knowledge was then transferred to B cell leukemia treatment by coupling immunotoxins to anti-CD22 Abs (24–26) or sialosides (27–30). Application of these treatments induces internalization of CD22 together with the bound toxin and results in the killing of leukemic cells. Because understanding the mechanism of CD22 internalization and recycling is important we were interested in finding interactors that might regulate this function. In this study, we searched for new CD22 interacting proteins to further decipher CD22 signaling and the underlying molecular mechanisms. We identified cullin 3 as a CD22-associated protein after anti-IgM treatment that is involved in CD22 internalization.

Cullin 3 belongs to the cullin protein family with seven different family members in mammals that function as E3 ubiquitin ligases (31). Ubiquitination is a posttranslational modification and is defined as the attachment of one ubiquitin molecule or an ubiquitin chain to a protein. This is directed by a guided reaction of ubiquitin-activating enzymes (E1), ubiquitin-conjugating enzymes (E2), and ubiquitin ligases (E3) (32, 33). The attachment of already one ubiquitin moiety (also referred to as monoubiquitination) is sufficient to induce receptor internalization (34). The receptors are then transported to the early endosome (sometimes referred to as sorting endosome) and from there either further to the lysosome for degradation or are recycled back to the cell surface after deubiquitination (35–37).

E3 ubiquitin ligases can be subdivided into two classes, the HECT-type E3 and the RING-type E3 ligase complexes (38, 39). Cullin-based complexes belong to the latter one. The covalent attachment of a Nedd8 moiety, an ubiquitin-like protein, activates cullin proteins (40, 41). Cullin 3, like other cullin members, operates as a scaffold protein to form a cullin 3–RING–ubiquitin ligase complex (CRL3). This multisubunit complex consists of a cullin 3, a RING protein, an E2 ubiquitin ligase, and a substrate adaptor that is connected via its BTB/POZ domain to the cullin 3 N terminus (38, 42). The active form of CRL3 is homodimeric, and the dimerization is mediated via the BTB/POZ-substrate adaptor proteins (43–45). Protein-interaction domains at the C terminus of the substrate adaptor bind the substrate and thereby recruit it into the CRL3 complex for ubiquitination. Those protein–protein-interaction motifs of BTB/POZ domain proteins are mostly either Kelch, PHR, or zinc finger domains (46). One protein family that often functions as a substrate adaptor in CRL3 is the Kelch-like family (KLHL), with KLHL11 or KLHL6 as examples (45, 47). Ubiquitination of the target protein can have different effects, such as proteasomal or lysosomal degradation or influences on protein localization (48). The function of cullin 3 and its targets were first studied in 1999 and revealed a regulatory role for cullin 3 in Cyclin E expression and therefore controlling S-phase entry in mammalian cells (49). Since then, many studies analyzed the function of cullin 3 in various organisms and cell types. Thus, cullin 3 is important to, for example, degrade MEI-1/Katanin in meiosis-to-mitosis transition in Caenorhabditis elegans or to target Cyclin E to maintain quiescence in murine hepatocytes (50, 51). The important role of cullin 3 is underlined by the fact that a total cullin 3 deletion is embryonic lethal (51), indicating an important developmental role.

During B cell development, cullin 3 is already expressed in the common lymphoid progenitor. The expression level is moderate and constant during B cell development, however the highest expression was observed in germinal center and plasma cells (ImmGen database, http://rstats.immgen.org/Skyline/skyline.html). So far, only one group has studied the function of cullin 3 in B lymphocytes. Using a CUL3^flox/flox mouse bred to a CD19^Cre mouse revealed hardly any alterations in B lymphocyte development (52). However, the marginal zone B cell population in the spleen and the B1 B cell population in the peritoneal cavity were impaired. They also demonstrated that cullin 3 associates with the BTB-zink finger transcription factor BCL6, which is important to direct germinal center B cell and T follicular helper cell development (52). In this study, we did not only focus on the role of cullin 3 with respect to CD22 but expanded our investigation to analyze general cullin 3 functions in B lymphocytes. Our studies revealed that cullin 3 is not only regulating CD22 expression and internalization after BCR activation but is also important to control pro-B cell proliferation and B cell survival.

Cul3^flox/flox mice (51) were purchased from Jackson laboratory (JAX stock no. 028349) and crossed to Flp mice to remove the frt-flanked PGK-neo cassette. Afterwards, Cul3^floxΔneo mice were mated with mb1^CRE mice (53) to obtain B cell–specific cullin 3 deficient, mice which were further named CUL3^fl/fl mb1^CRE/+ mice. Control mice (CUL3^fl/fl mb1^+/+ or CUL3^WT/WT mb1^CRE/+) were age- and sex-matched littermates with the same genetic background. No differences were observed between the two used controls, CUL3^fl/fl mb1^+/+ and CUL3^WT/WT mb1^CRE/+. Mice were originally generated from 129S4/SvJaeSor-derived AK7 embryonic stem cells (51) and were backcrossed to C57BL/6 for seven generations. CD22^−/− mice (5) and CD22-Y2,5,6F mice, which have point mutations in the tyrosines of ITIM 2,5,6 (12), were generated as described. Experiments were performed in accordance with the German law for protection of animals, after approval by the animal welfare committee.

Single-cell suspensions of bone marrow, spleen, lymph nodes, or isolated peritoneal cavity were prepared in PBS (Life Technologies) supplemented with 5% FCS (PAN Biotech). Blood was collected in PBS containing 1% heparin (Ratiopharm) from either tail vein or heart. Erythrocytes were depleted using 1× ammonium–chloride–potassium (ACK) lysis buffer, and afterward, cells were washed and stained with Abs conjugated to FITC, AlexaFluor488, PE, PE-CF594, PerCP-Cy5.5, PE-Vio770, allophycocyanin, allophycocyanin-Cy7, BV421, efluor506, or biotin and diluted in PBS containing 0.1% BSA (Carl Roth), 2 mM EDTA (Carl Roth), and 2 mM sodium azide (Sigma-Aldrich). The following Abs were used for cell surface staining: anti-B220 (clone RA3-6B2; eBioscience), anti-CD4 (clone GK1.5; BD Biosciences), anti-CD5 (clone 53-7.3; eBioscience), anti-CD8 (clone 53-6.7; eBioscience), anti-CD19 (clone 6D5; BioLegend), anti-CD21 (clone 7E9; own hybridoma), anti-CD22 (Cy34.1; BD Biosciences), anti-CD23 (clone B3-B4; BD Biosciences), anti-CD24/anti-HSA (clone M1/69; BD Biosciences), anti-CD25 (clone PC61.5; eBioscience), anti-CD43 (clone BV510; BD Biosciences), anti-CD86 (clone GL-1; eBioscience), anti-CD93 (clone AA4.1; ebioscience), anti-CD95/anti–Fas-R (clone Jo2; BD Biosciences), anti-CD117/anti–c-kit (clone 2B8; BioLegend), anti-CD127/anti–IL-7-Rα (clone A7R34; BioLegend), anti-CD268/anti–BAFF-R (clone 7H22-E16; BioLegend), anti-IgD (clone 11-26c; BD Biosciences), anti-IgM (clone II/41; eBioscience), anti-κ–L chain (clone R5240; Pharmingen), anti–MHC class II (MHCII) (clone 25-9-17; BD Biosciences), anti–Siglec-G (clone SH1; BD Biosciences), and α2-6-Sialic acid-PAA [6´SLN(Gc)-PAA, 1MD, kindly provided by N. Bovin]. Fc-block (2.4G2; hybridoma generated in house) was added to prevent unspecific binding via the Fc-part of the Abs to the cell’s Fc receptors. Cells were stained for 30 min at 4°C. To stain dead cells, fixable viability dye (eBioscience) or DAPI (Sigma-Aldrich) was used. Biotinylated Abs were detected by an additional staining of the cells with streptavidin-PE-Vio770 (Miltenyi Biotec). Afterwards, cells were washed. In case of intracellular staining, cells were fixed and permeabilized with Cytofix/Cytoperm kit (BD Biosciences) or by using 4% paraformaldehyde (PFA; Carl Roth) and 70% methanol in PBS (Carl Roth). The following Abs were used for intracellular staining: anti-active caspase3 (clone C92-605; BD Biosciences), anti-pY551/pY511-Btk (clone 24a/BTK; BD Biosciences), and anti-pY759-PLCγ2 (clone K68-689.37; BD Biosciences). In case of cell cycle analyses, DNA staining was performed with 10 μg/ml DAPI (Sigma-Aldrich) in PBS (Life Technologies)/0.1% Triton X-100 (Carl Roth). Afterwards, cells were washed. Data were acquired via FACSCalibur (BD Biosciences) or Cytoflex S (Beckman Coulter) and analyzed using FlowJo Software (Tree Star). Total cell numbers of living cells were determined via trypan blue staining.

Approximately 1–2 × 10⁷ bone marrow cells were resuspended in 0.7 ml RPMI 1640 media (Life Technologies) containing 5% FCS (PAN Biotech) and loaded with 0.7 μM indo-1-AM pluronic acid F-127 (Molecular Probes) by shaking for 25 min at 30°C. Afterwards, 0.7 ml RPMI 1640 media (Life Technologies) containing 10% FCS (PAN Biotech) was added, and cells were incubated by shaking for a further 10 min at 37°C. Cells were washed twice and stained with anti-B220 and anti-HSA as described above. Cells were washed and resuspended in Krebs–Ringer solution, and calcium mobilization was measured at LSR II (BD Biosciences). Basal Ca²⁺ level was determined for 50 s, then BCR was stimulated using anti-IgM [F(ab)₂] (Jackson ImmunoResearch), and Ca²⁺ mobilization was measured up to 3 min. Indo-1 loading efficiency was determined by stimulation with ionomycin (Sigma-Aldrich). Obtained data were analyzed with FlowJo software (Tree Star). In case of calcium mobilization assays of DT40 cells, 5 × 10⁶ cells were used without additional surface staining. BCR stimulation was performed with anti-chicken IgM (clone M4, kind gift from Prof. J. Wienands, University of Göttingen).

Bone marrow cells from CUL3^fl/fl mb1^CRE/+ and CUL3^fl/fl mb1^+/+ mice were isolated as described above. Living pro-B cells were sorted (B220^pos c-kit^pos DAPI^neg), and 25,000 pro-B cells per well were plated into a 96-well plate in RPMI 1640 supplemented with 5% FCS, 1,2 mM l-glutamine, 50 μM 2-ME, 100 U/ml penicillin/streptomycin (Pen/Strep), 1 mM sodium pyruvate, and 1× nonessential amino acids (Life Technologies). Cells were either treated with 10 ng/ml recombinant mouse IL-7 (BioLegend) or left untreated and cultivated for 6 d at 37°C and 5% CO₂. Three to five technical replicates were analyzed per condition. Each day, living cells were manually counted using trypan blue (Life Technologies) staining and Neubauer chamber. At day 4, an additional flow cytometry staining with anti-B220, anti-IgM, anti-CD25, and anti-CD117/anti-c-kit was performed to determine the percentage of pro-B cells.

To isolate splenic B2 cells, single-cell suspensions were erythrocyte depleted using 1× ACK solution and cells were incubated with supernatants of anti-CD4, anti-CD8, and anti-Thy1 hybridomas (generated in our house) for 30 min at 4°C. Afterwards, T cells were depleted via incubation of cells with baby-rabbit complement (Cedarlane) at 37°C for 45 min.

B cells were isolated from the spleen as described above and resuspended in RPMI 1640 supplemented with 5% FCS, 1,2 mM l-glutamine, 50 μM 2-ME, 100 U/ml Pen/Strep, 1 mM sodium pyruvate, and 1× nonessential amino acids. Cells were split in two cell culture dishes and either treated with 1 μM cullin inhibitor MLN4924 (Enzo) or DMSO as solvent control. MLN4924 is a specific small-molecule inhibitor of NEDD8-activating enzyme E1 and acts by covalent binding to NEDD8. Therefore, the covalent attachment of NEDD8 to members of the Cullin family, which is crucial for their ubiquitin ligase activity, is prevented (54). Further tests showed no difference in the percentage of living cells between MLN4924 treated and solvent control cells (Supplemental Fig. 1A). Cells were incubated (5% CO₂, 37°C) for 10 h. Afterwards, 1.4 × 10⁷ cells per condition were washed with PBS and resuspended in 0.35 ml RPMI 1640 containing 10% FCS. Ten micromolar bromoenol lactone (BEL; Cayman Chemical) was added to both conditions, an additional 1 μM MLN4924 was added to the overnight inhibitor-treated sample, and cells were incubated on ice for 10 min. BCR stimulation was introduced by adding 10 μg/ml anti-IgM [F(ab)₂] (Jackson ImmunoResearch). Cells were fixed using 2% PFA (Carl Roth) in PBS at different time points and stained with anti-B220, anti-CD22 and anti-κ–L chain as mentioned above. Data were acquired via FACSCalibur (BD Biosciences) and analyzed using FlowJo Software (Treestar).

To generate the CD22-TST-DT40 cell line, chicken DT40 cells (immature B cell line) were transfected by electroporation (260 V) with 10–15 μg pcDNA3.1zeo plasmid containing murine CD22 with a C-terminally fused Twin-Strep-tag (55). Wild-type (WT) DT40 cells, which do not express CD22, were used as control. Cells were cultured at 37°C, 5% CO₂ in RPMI 1640 supplemented with 10% FCS (PAN Biotech), 1% chicken serum (Sigma-Aldrich), 2 mM l-glutamine (Life Technologies), 50 μM 2-ME (Life Technologies), 100 U/ml Pen/Strep (Life Technologies), and 0.3 mg/ml Zeocin (only for transfected cells; Thermo Scientific). CD22-TST-WEHI-231 cells were generated by retroviral transduction using Plat-E packaging cell line (56) and pMSCV-puro plasmid containing murine CD22 with a C-terminally fused Twin-Strep-tag (55). Cells were cultured at 37°C, 5% CO₂ in RPMI 1640 supplemented with 10% FCS (PAN Biotech), 2 mM l-glutamine (Life Technologies), 50 μM 2-ME (Life Technologies), 100 U/ml Pen/Strep (Life Technologies), and 0.5 μg/ml Puromycin (Thermo Scientific).

CD22-TST-DT40 and WT DT40 cells were metabolically labeled using the stable isotope labeling by/with amino acids in cell culture (SILAC) method (57). For this, cells were cultivated for 10 d in SILAC RPMI 1640 (Life Technologies) supplemented with 10% dialyzed FCS (PAN Biotech), 2 mM l-glutamine, 50 μM 2-ME, 100 U/ml Pen/Strep, 200 mg/l l-arginine, and 40 mg/l l-lysine. Arginine and lysine were either supplemented as natural isotopes (light) or heavy amino acid isotopes (¹³C₆¹⁵N₄-arginine, ¹³C₆-lysine; heavy). Light and heavy labeling was switched for CD22-TST-DT40 and WT DT40 cells in the different experiments. A total of 1 × 10⁸ cells of each CD22-TST-DT40 and WT DT40 were starved for 1 h, washed twice with PBS (Life Technologies), and resuspended in PBS. If applicable, BCR was stimulated with 2 μg/ml anti-chicken IgM (clone M4, kind gift from Prof. J. Wienands, University of Göttingen) for 5 min at 37°C. Afterwards, cells were pelleted by centrifugation (1300 rpm, 5 min, 4°C) and lysed in ice-cold lysis buffer (138 mM NaCl, 50 mM Tris [pH 7.8], 10% glycerine, 1 mM sodium-orthovanadate, 0.5 mM EDTA [pH 8], 2% NP40; freshly supplemented with leupeptin, sodium-orthovanadate, PSMF, and aprotinin) for 30 min on ice. CD22-Twin-Strep-tag and bound proteins were purified from cell lysate via Twin-Strep-tag using 180 μl Strep-Tactin Type3 XT bead suspension (IBA Lifesciences). The enrichment was performed according to the manufacturer’s instructions, and proteins were eluted in 12 μl Rotiload (Carl Roth) at 96°C. Eluates of the samples to be compared (e.g., unstimulated CD22-TST-DT40 and unstimulated WT DT40) were combined and separated via SDS-PAGE. Colloidal Coomassie staining of the proteins was performed with PageBlue protein staining solution (Thermo Scientific).

Each gel lane was excised and cut into 9–10 gel slices. Slices were washed three consecutive times with 10 mM ammonium bicarbonate (ABC) and 5 mM ABC/50% EtOH (v/v). Protein reduction and alkylation was performed with 10 mM DTT at 56°C for 30 min, followed by incubation with 50 mM iodoacetamide at room temperature for 30 min in the dark. Gel slices were then successively washed and dehydrated with 10 mM ACN followed by 100% EtOH three times each. Slices were dried in vacuo for complete dehydration. Tryptic digestion (sequencing grade, 1:50 enzyme to protein ratio; Promega) was performed overnight in 10 mM ABC at 37°C. For peptide extraction, two consecutive times 0.05% (v/v) trifluoroacetic acid in 50% (v/v) acetonitrile were added to each slice and incubated in an ice-cold ultrasonic bath for 10 min. Extracted peptides were dried in vacuo and stored at −80°C until liquid chromatography–mass spectrometry (LC-MS) analysis.

LC-MS analyses were performed with an Ultimate 3000 RCLCnano HPLC system coupled online to a QExactive Plus mass spectrometer (both Thermo Scientific). The ultrahigh performance liquid chromatography system was equipped with two C18 μ-precolumns (Ø 0.3 × 5 mm; PepMap, Thermo Fisher Scientific) and an Acclaim PepMap analytical column (Ø 75 μm × 500 mm, 2 μm, 100 Å; Dionex LC Packings/Thermo Fisher Scientific). Resuspended peptides (0.1% trifluoroacetic acid, v/v) were loaded and washed for 5 min on a precolumn. Peptide separation was performed at 40°C and a flow rate of 250 nl/min. Peptides were separated using a binary solvent system consisting of solvent A (2% DMSO [v/v] /0.1% formic acid [v/v]) and solvent B (86% acetonitrile [ACN] [v/v], 2% DMSO [v/v], 0.1% formic acid [v/v]). A linear gradient was applied from 3 to 39% solvent B in 30 min. Afterwards, the gradient increased from 39 to 95% solvent B in 5 min and was kept constant at 95% for 5 min. Subsequently, the system was re-equilibrated at 3% solvent B for 14 min.

The QExactive Plus instrument was operated in the positive mode and externally calibrated using standard compounds. Peptides were ionized with a Nanospray Flex ion source (Thermo Scientific) and either stainless steel (Thermo Scientific) or fused silica emitters (New Objective) with a spray voltage of 1.8 and 1.5 kV, respectively. The temperature of the transfer capillary was set to 200°C. Full scans were measured in a scan range between 375 and 1700 m/z and with a resolution of 70,000 (at 200 m/z). The automatic gain control target was set to 3 × 10⁶ ions with a maximum ion collection time of 60 ms. For tandem MS measurements, the top 12 most intense ion signals exceeding the intensity threshold of 5.8 × 10³ were selected and subjected to higher-energy collisional dissociation fragmentation with a normalized collision energy of 28. Fragment spectra were measured with a resolution of 35,000 (at 200 m/z), an automatic gain control target of 1 × 10⁵ ions, and a maximum ion collection time of 120 ms. Unassigned and singly charged ions were excluded from fragmentation. The dynamic exclusion time for previously fragmented precursor ions was set to 45 s.

The Andromeda/MaxQuant software pipeline (version 1.5.3.30) was used for LC-MS raw data analysis (58, 59). The UniPort ProteomeSet of Gallus Gallus including isoforms (17,902 entries, downloaded in February 2015) was extended by the three mouse CD22 isoforms. This database and a set of common contaminations provided by MaxQuant were used to generate peak lists from tandem MS spectra. The MaxQuant database search was performed with the following parameters: sample multiplicity two (Arg10 and Lys8 as heavy labels); tryptic specificity allowing up to three missed cleavages; acetylation of the N termini and oxidation of methionine as variable modifications; carbamidomethylation of cysteine as fixed modification; mass tolerances of 20 ppm for first search and 4.5 ppm for the main search of precursor ions. For identification, a false discovery rate of 1% was applied with a maximal peptide mass of 5600 Da, and a minimum of one unique peptide was required for protein identification. For quantification, the “re-quantify” and “match between runs” options were enabled, and a minimal ratio count of two unique peptides was selected. Perseus (version 1.5.2.6) was used for statistical analysis of MaxQuant data from six biological replicates. Entries identified only by site, reverse, or potential contaminants were removed. Data were further filtered for at least three valid values, normalized SILAC ratios from label-switch experiments were inverted, and log₂-transformed mean ratios were calculated. A one sample t test was applied and p values were calculated. Proteins with a p value < 0.05 and a mean log₂ ratio greater than 1, 2, or 4 were considered as moderately, significantly, or highly significantly enriched.

First, Twin-Strep-tag affinity purification (see above) or anti-CD22 immunoprecipitation with rabbit anti-CD22 (polyclonal; Pineda) as described before (60) was carried out. Proteins were separated via SDS-PAGE and transferred to nitrocellulose membrane using Trans-Blot Turbo Transfer System (Bio-Rad). In the case of ubiquitin analysis, the nitrocellulose membrane was denatured after protein transfer by incubation for 30 min at 4°C in denaturation buffer (6 M guanidine, 5 mM 2-ME, 1 mM PMSF, 20 mM Tris [pH 7.5]). For analysis, membranes were incubated with one of the following Abs: anti-CD22 (polyclonal; Pineda), anti–cullin 3 (polyclonal C-18 or monoclonal G-8; Santa Cruz Biotechnology), anti-ubiquitin (FK2; Enzo Life Sciences), and anti–Twin-Strep-tag (StrepMAB-Classic, Strep-tagII specific, IBA Lifesciences). If necessary, incubation with HRP-conjugated Abs was performed using anti-rabbit-IgG HRP (Cell Signaling Technology), anti-goat-IgG HRP (Cell Signaling Technology) or anti-mouse IgG HRP (Cell Signaling Technology). Proteins were detected with SuperSignal West Pico PLUS Chemiluminescent Substrate (Thermo Scientific).

CD22-TST-DT40 or CD22-TST-WEHI-231 cells were cultured in DT40 media or WEHI-231 media with 1 μM cullin inhibitor MLN4924 (Enzo Life Sciences) or solvent control for 12–14 h (5% CO₂, 37°C). In addition, 10 μM MG-132 and 20 mM NH₄Cl were added for 4 h to block proteasome and lysosome activity. A total of 2 × 10⁷ cells were harvested and lysed in ice-cold lysis buffer (138 mM NaCl, 50 mM Tris [pH 7.8], 10% glycerin, 1 mM sodium-orthovanadate, 0.5 mM EDTA pH 8, 1% Triton X-100; freshly supplemented with leupeptin, sodium-orthovanadate, PMSF, aprotinin, iodoacetamide, and N-ethylmaleimide) for 30 min on ice. Twin-Strep-tag purification of CD22 was preformed and samples were analyzed via immunoblotting (both as described above).

ELISA was used to determine total titer of different Ig isotypes in sera of naive mice. Polysorb plates (Nunc) were coated with 10 μg/ml isotypic-specific unlabeled Abs (anti-IgM, anti-IgG1, anti-IgG2b, anti-IgG2c, anti-IgG3, and anti-IgA; Southern Biotech) diluted in PBS and incubated overnight at 4°C. The next day, plates were blocked with 1% BSA in PBS for 2 h at 37°C. Afterwards, plates were washed three times with PBS. Starting dilution of 1:100–1:200 were chosen for serum samples, followed by a serial dilution of 1:3 in PBS containing 0.1% BSA. mIg Abs of a known concentration (Southern Biotech) served as standards. Serum samples were incubated overnight at 4°C on Polysorb plates. The next day, plates were washed three times with PBS and incubated with isotype specific alkaline phosohatase (AP) conjugated detection Abs (Southern Biotech) for 2 h at 37°C. By addition of the substrate for AP (4-nitrophenyl phosphate disodium salt hexahydrate tablet [Sigma-Aldrich]) dissolved in 20 ml 9.7% solution of 2,2’-iminodiethanol in H₂O and 0.5 mM MgCl₂ (pH 9.8), binding of AP-conjugated Abs was detected. The OD was measured by ELISA reader (Tecan) at 405 nm and data were analyzed using Soft Max Pro (Molecular Devices).

Murine splenic B cells were isolated via MACS using CD19 microbeads (Miltenyi Biotec) and incubated with anti-CD22-PE (Cy34.1; BD Biosciences) for 1 h at 4°C. After washing, endocytosis was induced by incubating the cells at 37°C for different periods of time. Samples were then washed with 0.2 M glycine/HCl (pH 2.5) to remove extracellular bound Abs and DAPI for dead cell exclusion. Cells were fixed using 2% PFA (Carl Roth) in PBS and analyzed on an LSR II (BD Biosciences). Only intracellular anti-CD22 Ab could be detected. At time point 0 min, an additional sample was prepared without washing off extracellular Abs and was then used as extracellular staining control. The mean fluorescence intensity (MFI) of this sample was set as 100%.

Statistical analyses were performed using GraphPad Prism software. Unpaired t test or Mann–Whitney U test was used to evaluate significance. Statistical data are presented as mean ± SD.

All raw data and original MaxQuant result files have been deposited to the ProteomeXchange Consortium (http://proteomecentral.proteomexchange.org) via the Proteomics Identification Database Partner Repository (61) with the dataset identifier PXD013801.

The immature chicken B cell line DT40 (62–64) was used to study the CD22 interactome. These cells were chosen because they do not constitutively express any endogenous Siglec isoforms and can be easily transfected using electroporation. Therefore, DT40 cells were transfected with murine CD22 that is C-terminally fused to a Twin-Strep-tag. Afterwards, the correct functioning of murine CD22 in the chicken B cell line was confirmed (Supplemental Fig. 1B–D). We could show that CD22 is able to inhibit BCR-induced Ca²⁺ signaling (Supplemental Fig. 1B), can bind to α2,6-linked sialic acids on the cell surface (Supplemental Fig. 1C), and is expressed at levels comparable to murine splenic B cells (Supplemental Fig. 1D). In our experimental setup, SILAC DT40 cells were used in their WT form (control) or as stably expressing CD22-Twin-Strep-tag cells. To induce BCR signaling, cells were stimulated with anti-chicken-IgM. Subsequently, the CD22 interactome was analyzed from unstimulated and anti-IgM–stimulated SILAC DT40 cells using quantitative affinity purification–mass spectrometry.

Based on the acquired data, mean log2 SILAC ratios of proteins were calculated and plotted against −log₁₀ p values. Proteins exhibiting a mean log₂ ratio ≥1 and a p value ˂ 0.05 (n = 6) were deemed significantly enriched with CD22. For unstimulated cells, we found 36 proteins to be enriched with CD22, whereas 64 proteins were enriched with CD22 upon BCR stimulation (Fig. 1A, Supplemental Table I). We were able to confirm already known CD22 interacting proteins, which associate with CD22 upon BCR stimulation. Thus SHP-1, parts of the BCR (including Ig-β, Ig–μ-chain, and Ig–λ-chain), and signaling proteins such as Syk, Grb2, and CD45 were highly enriched following anti-IgM stimulation. Two of the BCR stimulation–induced CD22 interaction partners (i.e., SHP-1 and Ig-β) were also identified in the resting state, suggesting already a weak interaction of CD22 with Ig-β and SHP-1 under unstimulated conditions (mean log₂ ratios: Ig-β = 1.69 and SHP-1 = 2.03). However, in BCR-stimulated cells, the association of Ig-β and SHP-1 with CD22 was considerably higher, as shown by the increase of their enrichment factors (mean log₂ ratios: Ig-β = 4.66 and SHP-1 = 3.53). Thus, our data confirm previous findings that CD22 is recruited to the BCR signaling complex after anti-IgM stimulation but is only weak associated to the BCR in the resting state (12, 65–67). In addition to known CD22-interacting proteins, various new potential CD22 interaction partners including cullin 3, clathrin, BCAP, SERCA2, and Dok3 were identified in this analysis (Fig. 1A). Here, Dok3 and SERCA2 seem to interact with CD22 independent of the cell’s activation status, as they were associated with CD22 regardless of the stimulation condition. However, the majority of so far unreported proteins associated with CD22 were only found after anti-IgM treatment. To further investigate CD22 signaling-dependent mechanisms in B lymphocytes, we decided to examine cullin 3 as novel potential CD22 interaction partner after BCR stimulation.

Cullin 3, an E3 ubiquitin ligase, was found to be associated with CD22 exclusively after anti-IgM stimulation (Fig. 1A). To further validate this interaction in CD22-TST-DT40 cells, CD22-Twin-Strep-tag affinity purifications were performed followed by anti–Twin-Strep-tag and anti–cullin 3 immunoblotting (Fig. 1B). We included a total cell lysate from WT DT40 cells as a positive control for cullin 3 detection. The specificity of the cullin 3 Ab was furthermore proven by analyzing B cell lysates from control mice (CUL3^fl/fl mb1^+/+) compared with lysates from CUL3^fl/fl mb1^CRE/+ mice, which have a B cell–specific cullin 3 deletion (Fig. 1C). Following BCR stimulation of CD22-TST-DT40 cells, Western blot analyses support the results from the quantitative affinity purification–mass spectrometry experiments. The results show two characteristic cullin 3 isoform bands at ∼80 and 89 kDa corresponding to nonneddylated and neddylated cullin 3 (68). In unstimulated cells, mainly the lower band (80 kDa) of cullin 3 was detected. This band represents the nonneddylated and therefore inactive form of cullin 3 (68). Upon BCR, the neddylated cullin 3 form (89 kDa) was additionally precipitated with CD22. This observation was further confirmed by coimmunoprecipitation experiments from primary splenic WT B cells. Here, we were able to detect both cullin 3 isoforms (nonneddylated and neddylated) associated with CD22 only after BCR stimulation (Fig. 1D).

We proceeded to characterize the relationship between cullin 3 and CD22 to understand the molecular connection between these two proteins. As cullin 3 is an E3 ubiquitin ligase and a scaffold protein for CRL3, ubiquitination analyses of CD22 were conducted. Therefore we performed CD22 immunoprecipitation from BCR-stimulated CD22-TST-DT40 cells, followed by anti-ubiquitin Western blot. The experiments showed that cells that have been treated with a cullin inhibitor (MLN4924) hardly revealed a signal of ubiquitinated CD22 after BCR stimulation (Fig. 1E). In contrast, ubiquitin-specific bands with the molecular mass of CD22 (140 kDa), as well as a weak higher molecular size band, were observed in CD22-TST-DT40 solvent control cells after 5 min of anti-IgM stimulation (Fig. 1E). This signal seems to be reduced after 30 min of stimulation. A similar experimental approach was conducted to investigate CD22 ubiquitination after BCR activation in the murine B cell line WEHI-231. The results of CD22-Twin-Strep-tag–transfected WEHI-231 cells show a similar pattern (Fig. 1F). Cells treated with 1 μM MLN4924 did not reveal a band for ubiquitinated CD22, whereas DMSO control cells showed a clear signal up to 120 min after BCR stimulation.

Ubiquitination of surface receptors leads to internalization (34) and CD22 is known to internalize and recycle after anti-CD22 Ab ligation (20, 21, 69). Thus, it was first analyzed if CD22 internalization and recycling takes also place after indirect CD22 activation by BCR stimulation. Therefore the CD22 internalization and recycling capacity was analyzed using flow cytometry. Isolated splenic B cells from WT mice were treated with BEL to inhibit early endosomes and prevent internalized surface receptors from recycling back to the surface. BEL antagonizes Ca²⁺ independent PLA2 activity, which is important for endosomal tubule formation and thereby inhibits trafficking from sorting to recycling endosomes (70). As shown in Supplemental Fig. 1E, addition of BEL before anti-IgM stimulation results in a reduced CD22 surface expression with time (starting at 60 min after stimulation), demonstrating CD22 internalization. In comparison, solvent control WT cells did not reveal a decrease of CD22 surface expression after BCR stimulation, indicating that CD22 is slowly internalized after 60 min of IgM stimulation and recycled back to the surface afterward. This experiment was further expanded to study CD22 internalization after BCR stimulation with respect to cullin activity. Hence, splenic WT B cells were treated with BEL and additionally either cullin inhibitor MLN4924 or with a solvent control. The unchanged CD22 surface expression revealed that cullin inhibition abrogated CD22 internalization after BCR stimulation, (Fig. 2A, 2B), whereas control cells showed a significant reduction of CD22 MFI starting 60 min after anti-IgM treatment. CD22 internalization was further investigated in splenic B cells from CD22-Y2,5,6F mice (12). These cells harbor point mutations within the three cytosolic ITIMs and therefore also in the clathrin-dependent internalization motifs (21). Our observations show that these B lymphocytes behaved like cullin inhibitor–treated cells showing unchanged CD22 surface MFI (Fig. 2B). The internalization of the BCR itself, however, was not affected by cullin inhibition and showed a very fast internalization kinetic, as detected by reduced levels of the Ig κ MFI after stimulation (Fig. 2C), in comparison with CD22 levels. In addition, B220 MFI was analyzed and no significance difference between MLN4924 and DMSO control cells was observed (Fig. 2D). Moreover, B220 MFI stayed constant after BCR stimulation. As mentioned before, fast CD22 endocytosis can be achieved by ligating CD22 with an anti-CD22 Ab. This more direct way of activating CD22 endocytosis can be measured by incubating splenic B cells with a fluorochrome-conjugated anti-CD22 Ab. At different time points, nonendocytosed Ab is washed off, and the fluorescent signal from internalized Ab was measured via flow cytometry. As depicted in Supplemental Fig. 1F, inhibition of cullin did not modify CD22 endocytosis rate in this direct activation model.

To further examine the role of cullin 3 in B lymphocytes, cullin 3^flox/flox mice (51) were obtained and mated to mb1-cre mice to generate a B cell–specific knockout of cullin 3. First, B cell development in the bone marrow (Fig. 3A) and B cell subsets in spleen (Fig. 3B), lymph node, blood, and peritoneal cavity (Supplemental Fig. 2A) were analyzed via flow cytometry in these mice (summarized in Table I). In the bone marrow, we observed a severe reduction of the total B cell count in CUL3^fl/fl mb1^CRE/+ mice, starting from the pre-B cell stage (Fig. 3A). Peripheral B cell numbers in different lymphatic organs were lowered, too. This affected all subpopulations of B cells, conventional B2 cells, marginal zone B cells, and B1 cells (Fig. 3B). The remaining B cells in bone marrow and spleen were further analyzed regarding their CD22 expression levels. We were able to show that CUL3^fl/fl mb1^CRE/+ B cells express significantly higher amounts of CD22 on the cell surface during developmental stages in the bone marrow and spleen but also on mature splenic B cells (Fig. 3C). Furthermore, we found generally decreased Ig serum level in naive young CUL3^fl/fl mb1^CRE/+ mice (Fig. 3D).

As KLHL6 was shown to be a substrate adaptor for cullin 3 (47, 71) and was identified in the quantitative CD22 interactome study upon BCR stimulation (Fig. 1A), B cell populations of KLHL6^−/− mice were investigated regarding their CD22 expression level. These, however, did not show CD22 overexpression (Supplemental Fig. 2B), as observed in B cell–specific cullin 3 deficient mice (Fig. 3C). Furthermore, the overexpression phenotype of CD22 in CUL3^fl/fl mb1^CRE/+ mice is not observed for other B cell surface receptors, like CD19, Siglec-G, IgM, or IgD (or IgM/IgD ratio), as their expression pattern is mostly unaltered on bone marrow and splenic B cells of CUL3^fl/fl mb1^CRE/+ mice (Supplemental Fig. 3A, 3B). However, some receptors seem to be influenced by a cullin 3 deficiency. It turned out that besides CD22, CD21 and CD23 expression levels are also enhanced on follicular B cells in the spleen (Supplemental Fig. 3B).

As inhibition of cullin leads to altered CD22 internalization after BCR stimulation in murine splenic B cells from WT mice (Fig. 2B), this experiment was also applied to the CUL3^fl/fl mb1^CRE/+ strain. Because of the low numbers of remaining splenic B cells available, only two time points, unstimulated and 120 min anti-IgM stimulation, were analyzed (Fig. 3E). The results of CUL3^fl/fl mb1^CRE/+ B cells resemble MLN4924-treated WT B cells, as no significant CD22 MFI reduction was observed. In contrast, control B cells (CUL3^WT/WT mb1^CRE/+) showed a moderate, but highly significant, decrease of CD22 on the cell surface (Fig. 3E).

Subsequent experiments were conducted to analyze the impaired B cell development in CUL3^fl/fl mb1^CRE/+ mice. First, flow-sorted pro-B cells were cultured for 6 d with or without IL-7 treatment (Fig. 4A left). The addition of 10 ng/ml IL-7 resulted in proliferation of CUL3^fl/fl mb1^+/+ pro-B cells. Signaling via IL-7– IL-7-Rα is important for early B cell development in the bone marrow to ensure correct Ig gene rearrangement and differentiation from pro- to pre-B cells (72, 73). In contrast to WT controls, CUL3^fl/fl mb1^CRE/+ pro-B cells did not show this expansion in cell numbers, instead cell numbers decreased with time. Overall, we detected a significant defect in the proliferation capacity of CUL3^fl/fl mb1^CRE/+ pro-B cells. The kinetics of these cells resemble more the untreated samples. The percentages of pro-B cells at day 4 within the cultures show no major difference between CUL3^fl/fl mb1^CRE/+ and CUL3^fl/fl mb1^+/+, indicating no escape from proliferation by further differentiation (Fig. 4A right). Furthermore, the expression level of IL-7-Rα, which is the receptor to IL-7, was examined (Fig. 4B). However, no differences were detected in the IL-7-Rα expression on bone marrow B cells between CUL3^fl/fl mb1^CRE/+ and CUL3^fl/fl mb1^+/+ mice. Others were able to show a regulatory role of cullin 3 in mitotic cell cycle entry and delayed mitotic entry in absence of cullin 3 (74). Such a possible defect in the pro-B cell subset could contribute to the proliferation defect, that we observed. Therefore we were interested in cell cycle phase distribution of pro-B cells from CUL3^fl/fl mb1^CRE/+ mice. Our results demonstrate significant enrichment of CUL3^fl/fl mb1^CRE/+ pro-B cells in the G2/M cell cycle phase and a tendency of reduced percentage in G1 phase (Fig. 4D), indicating a dysregulated cell cycle in pro-B cells.

Additional analyses focused on investigating indications for ongoing apoptosis in cullin 3-deficient B cells, which would also explain the observed phenotype. Therefore, intracellular staining of active caspase3, which is an indicator of ongoing apoptosis, was performed in bone marrow B cells and analyzed via flow cytometry (Fig. 4C). Early developmental B cell stages showed a clear tendency of more active caspase3 and mature recirculating CUL3^fl/fl mb1^CRE/+ B cells exhibited a significantly stronger caspase3 activation compared with control cells. Further experiments, regarding the expression level of BAFF-R and Fas-R, additional receptors critical for survival and apoptosis regulation, were performed to further analyze the mechanism of the developmental defect of B cells in CUL3^fl/fl mb1^CRE/+ mice. Thus, flow cytometry analysis of bone marrow and splenic B cell populations were conducted. Developing B cells in the bone marrow did not reveal a higher Fas-R expression; this only applied to the mature recirculating B cell pool (Fig. 4E, Supplemental Fig. 3C). However, splenic subpopulations, beginning at transitional T2 stage, had an enhanced Fas-R surface expression (Fig. 4E, Supplemental Fig. 3C). Furthermore, immature CUL3^fl/fl mb1^CRE/+ B cells in the bone marrow showed a reduced BAFF-R expression (Fig. 4F, Supplemental Fig. 3C). This observation was also confirmed in the splenic B cell populations (Fig. 4F, Supplemental Fig. 3C). In summary, pro-B cells display an proliferation defect with a tendency to more apoptosis induction and accumulation in the G2/M cell cycle phase cells. Higher Fas-R and reduced BAFF-R expression, however, seemed to mainly affect the more mature stages of B cell development in bone marrow and all B cell populations in the spleen, which could have direct consequences for B cell survival.

As CD22 is regulating B cell activation (5–9), it was important to study B cell activation status in cullin 3–deficient B lymphocytes, which among other phenotypes show an enhanced CD22 expression. To study basal and BCR-induced activation level, calcium mobilization assays of pre-B/immature B cells from the bone marrow were performed. After determining basal cellular calcium concentration for 50 s, anti-IgM stimulation was applied, and the changes in calcium concentration were measured for additional 150 s. The result showed reproducible higher basal calcium level (measured as indo-1 ratio of Ca²⁺ bound to Ca²⁺ unbound dye) before BCR stimulation in CUL3^fl/fl mb1^CRE/+ mice (Fig. 5A). Additionally it was observed that upon anti-IgM treatment in different concentrations, the CUL3^fl/fl mb1^CRE/+ B cells were not able to reach the cytosolic calcium level of control cells. This preactivation was then further analyzed on splenic B cells by staining of typical activation markers. The experiments revealed a higher expression level of the activation marker MHCII and CD86 (Fig. 5B, 5C). Additional experiments showed that proteins of the BCR signaling cascade, such as PLCγ2 and BTK, also have a higher basal activation, which increased even further after BCR stimulation (Fig. 5D, 5E).

In this study, we performed a global CD22 interactome analysis to identify CD22 interacting proteins in BCR unstimulated and stimulated cells. We identified various new CD22 interacting proteins that have not been described in the context of CD22 so far. Among those, we focused on investigating the connection between CD22 and cullin 3, an E3 ubiquitin ligase. E3 ubiquitin ligases are involved in regulating membrane receptor expression via ubiquitination followed by internalization (34, 48, 75). This process is mediated in either a clathrin-dependent or a clathrin-independent way (76–79). Our study revealed that after anti-IgM stimulation, CD22 is ubiquitinated in a cullin-dependent manner and later on internalized and recycled to the surface. We think that cullin 3, which was identified as CD22 interaction partner in our screening, is the cullin protein promoting CD22 ubiquitination. However, as we used a global cullin inhibitor for our functional ubiquitination assays, we cannot exclude the involvement of other cullin family members. In our analysis, we also observed some differences in CD22 ubiquitination kinetics between CD22-Twin-Strep-tag–transfected cell lines from chicken and mouse origin. Both systems, however, show a cullin-dependent CD22 ubiquitination in response to BCR activation, which is a new important finding in the field of Siglec biology.

Subsequent analyses focused on revealing the downstream effects of CD22 ubiquitination and led to our model (Fig. 6). We propose that upon BCR stimulation, CD22 is activated within the first minutes of B cell activation through tyrosine phosphorylation by kinases. Thereby a CD22 mediated signaling cascade is initiated. CD22 is then ubiquitinated presumably by CRL3, peaking at 120 min in the murine system (earlier in DT40 cells). However, CD22 stays on the surface for some time and is not endocytosed right away. This correlates with our observation of a delay in clathrin-dependent internalization of CD22. This process is initiated by binding of AP50 to CD22. AP50 is part of the clathrin-associated AP-2 adaptor complex and binds to CD22 via Y843 and Y863 (part of the internalization motifs YXXØ), which are located in the cytoplasmic tail of CD22. However, the tyrosines of CD22 become phosphorylated shortly after BCR activation, which blocks AP50 binding to CD22 (21, 80). Only after ∼30 min, when this phosphorylation diminishes, clathrin-dependent internalization is initiated (21, 80). Then CD22 molecules return to the cell surface by a mechanism called receptor recycling rather than being degraded in the lysosome (Fig. 6). Our experiments could show for the first time, to our knowledge, that CD22 internalization upon BCR stimulation is dependent on the ubiquitin machinery. The molecular reason of this rather slow internalization process can be explained with the need for the inhibitory signal of CD22 after BCR stimulation. By sustaining CD22 localization to the cell surface after B cell activation, instead of fast CD22 internalization, the mechanism ensures adequate regulation of BCR signaling. However, this inhibitory signal needs to be turned off eventually, and this newly determined CRL3-dependent internalization of CD22 might be the molecular way to guarantee this.

This newly identified role of cullin 3 in activated B cells raised the question about its function in steady-state B lymphocytes. To address this, we crossed CUL3^flox/flox mice to the B cell–specific cre-deleter strain mb1^cre/+ to obtain highly efficient B cell–specific cullin 3 deletion already in early developmental B cell stages (53). Our analyses demonstrate that a cullin 3 deficiency not only causes defects in CD22 internalization upon BCR stimulation but also leads to constitutive higher expression levels of CD22 in unstimulated B cells. This again points to an important regulatory role of cullin 3 with respect to CD22. However, we do not conclude that this additional regulatory mechanism of cullin 3 in the steady-state is necessarily mediated by the same cellular mechanism as the internalization of CD22 in activated B cells. Although cullin 3 is clearly involved in controlling CD22 surface expression in unstimulated and stimulated B cells, the mechanisms might be different. Besides CD22, CD21 and CD23 surface receptor levels also seem to be elevated, indicating an additional role for cullin 3 in CD21 and CD23 expression.

Cullin 3 is known to be already expressed during B cell development in the bone marrow. Hence, we were also interested in other B cell related functions of cullin 3. Despite previously published results that showed hardly any developmental defects in CUL3^fl/fl CD19^CRE/+ B cells (52), we observed a much stronger opposing phenotype. We would like to highlight that instead of using CUL3^fl/fl CD19^CRE/+ mice, which were used in the other study, we chose to mate CUL3^fl/fl with the mb1^CRE/+ deleter strain. A gene deletion under the mb1 promotor was shown to be more efficient, compared with the CD19^cre/+, and is therefore better suited to study B cell related functions (53). We chose the mb1^cre/+ system to guarantee an earlier and more efficient deletion of the loxP flanked DNA segment of cullin 3 and therefore avoid cells that escape this excision process (53). This more stringent approach revealed an important function of cullin 3 during B cell development, as B cell numbers were heavily impaired in CUL3^fl/fl mb1^CRE/+ mice. This phenotype was not observed by Mathew et al. (52), probably because of inefficient and incomplete deletion of CUL3 using the CD19^cre/+ system. So-called “deletion escapers” could fill up the B cell niche, as they might have a survival advantage compared with cullin 3–deficient cells. This idea is supported by our observations, as we only detected few B cells in the periphery of CUL3^fl/fl mb1^CRE/+ mice.

The surviving B cells were characterized by an apoptotic and at the same time preactivated phenotype. However, we do not think that cullin 3–deficient B cells are functionally unresponsive, as IgM and IgD levels were comparable to WT cells, and anergic B cells would be characterized by an IgM reduction and higher IgD expression (81–85). Furthermore, we do not conclude that CD22 overexpression causes the preactivated and apoptotic phenotype, because CD22-deficient B cells also show signs of preactivation and higher turnover (5). Therefore, this might rather be a direct or indirect effect of the cullin 3 deletion. However, whether the augmented CD22 expression is relevant for any of the B cell phenotypes in CUL3^fl/fl mb1^CRE/+ mice will be further analyzed using CD22^−/− × CUL3^fl/fl mb1^CRE/+ in the future. The crossing of CD22^−/− and CUL3^fl/fl mb1^CRE/+ is still ongoing, as CD22 and mb1 genes are both located on chromosome 7 with a distance of only 5.77 cM.

Developing B cells in the bone marrow showed a drastic reduction in the CUL3^fl/fl mb1^CRE/+ mice. This phenotype was observed starting at the pre-B cell stage. Hence, we addressed the question if the reduction of pre-B cells is caused by a proliferative or differentiational defect. Here we focused on IL-7 signaling, which is important for B cell development from pro- to pre-B cells in the bone marrow (72, 73). In fact, our data clearly demonstrate that CUL3^fl/fl mb1^CRE/+ pro-B cells fail to proliferate in presence of IL-7 but die instead. As IL-7R expression was unaltered, we suggest that a cell intrinsic problem causes this defect and propose that cullin 3 is important for pro-B cell proliferation. Moreover, cullin 3 was published to be involved in cell cycle control of hepatocytes (51), which might be one of its molecular function in B cells, too. Others have shown a regulatory role for cullin 3 in mitotic cell cycle entry and that cullin 3 knockdown cells undergo a delayed mitotic entry (74). In these experiments Cul3 small interfering RNA transfected HeLa cells showed an accumulation in G2/M phase (35.9% compared with 21.5% in control cells) (74). We observed this to a similar extent in CUL3^fl/fl mb1^CRE/+ pro-B cells. This accumulation of G2/M cell cycle phase might contribute to less pro-B cell proliferation.

Altogether, we showed that cullin 3 is a novel CD22-interacting protein that controls not only CD22 surface expression under steady-state conditions but also induces CD22 internalization by ubiquitination after BCR stimulation. Moreover, characterizing B cell–specific cullin 3 deficient mice (CUL3^fl/fl mb1^CRE/+), we revealed the importance of cullin 3 for normal B cell development. To our knowledge, this study provides a first insight into the role of cullin 3 on B cell development and additionally shows its importance for CD22 surface expression and receptor internalization after B cell activation.

We thank the Proteomics Identification Database team for data deposition to the ProteomeXchange Consortium.

The authors have no financial conflicts of interest.