Crlz-1 Controls Germinal Center Reaction by Relaying a Wnt Signal to the Bcl-6 Expression in Centroblasts during Humoral Immune Responses

The germinal center (GC) is a dynamic microanatomic site that is formed by the activated B cells within the follicles of secondary lymphoid organs during humoral immune responses (1, 2). Histologically, the GC is divided grossly into the dark zone (DZ) and light zones (LZ). The DZ is densely populated by the rapidly proliferating activated B cells, called centroblasts, whereas the LZ is sparsely populated by the resting B cells of next developmental stage, called centrocytes (3). Centroblasts undergo somatic hypermutations (SHM) within the V region of Ig gene, as well as its class switch recombination (CSR), catalyzed by activation-induced cytidine deaminase (AID) (4). Centrocytes are selected competitively, depending on the affinities of their hypermutated BCR for a limited amount of Ags trapped by Fc and/or complement receptors on the follicular dendritic cells (FDC) as well as the preferential association of their endocytically processed antigenic peptide-loaded MHC class II with TCRs of the follicular helper T cells (Tfh) and the consequential CD40L and cytokine signals (5). Depending on the nature and/or strength of signals from BCR, as well as CD40L and cytokines, the selected centrocytes could either shuttle back to the DZ to repeat a further centroblastic proliferation or proceed to differentiate into Ab-secreting plasma or memory cells (6–9) with a concomitant CSR (10). After the repeated shuttling of activated B cells between the DZ and LZ, the affinity of Ab is matured to evolve at the molecular level (11–13).

Bcl-6 is known to be the master regulator of centroblasts (10, 14, 15), regulating the expression of its various target genes directly or indirectly, especially cyclins D1–D3, p53, Blimp-1, and IRF-4 (16). Cyclins D1–D3 are G1 checkpoint regulators, suggesting their important roles during the centroblastic proliferation. p53, a DNA damage sensor, is repressed to tolerate DNA breaks during both SHM and CSR (10, 17, 18). Blimp-1 and IRF-4, which are the regulators of plasma cell differentiation, are repressed to block the differentiation of selected GC B cells into plasma cells and to continue their repeated shuttling between the DZ and LZ for Ab affinity maturation (10, 16).

Charged amino acid–rich leucine zipper-1 (Crlz-1) was found to be expressed especially in the proliferating cells such as pre-B cells (19, 20), spermatogonia, and Sertoli cells (21). This gene was targeted by Wnt/β-catenin signaling (20, 22), and its protein was shown to function by mobilizing cytoplasmic CBFβ into the nucleus to allow its heterodimerization with nuclear Runx (23). This knowledge, with many reports that canonical Wnt/β-catenin signaling is involved in the cellular proliferation (24, 25), prompted us to wonder whether Crlz-1 might also be expressed in the rapidly proliferating centroblast cells within GC during humoral immune response. It is well known that canonical Wnt signaling is transduced through the stabilization of β-catenin by recruiting Dvl to the Fzd/Lrp receptor complex and, subsequently, inactivating the β-catenin destruction complex (26–28). The stabilized β-catenin enters to the nucleus and binds to the promoter of the target gene with TCF/LEF for its transcriptional activation (29, 30). The Crlz-1 promoter contains this TCF/LEF binding sequence motif, which was shown to bind LEF-1 together with β-catenin to activate it (20, 22).

In this paper, we report that Crlz-1, which is induced by canonical Wnt/β-catenin signaling, controls GC reaction by regulating the expression of Bcl-6 in centroblasts. Mechanistically, Crlz-1 was shown to regulate the expression of Bcl-6 by mobilizing CBFβ into the nucleus to allow Runx/CBFβ heterodimerization and, thereby, its subsequent binding to the Bcl-6 promoter. Eventually, the expression of Bcl-6 was linked to the regulation of its various target genes, such as cyclins D1–D3, p53, Blimp-1, and IRF-4, correlating well with their roles in the rapidly proliferating centroblasts. In accord with this cascading regulatory axis, when Crlz-1 expression was inhibited in centroblasts by administration of Wnt inhibitors into the immunized mice, GC reaction was indeed found to be impaired or abolished because of the diminished proliferation of centroblasts and/or their earlier differentiation into plasma cells, with consequential decrements of Ab affinity maturation and CSR as well as smaller output of plasma cells.

Ramos (CRL-1596; American Type Culture Collection) and Raji (CCL-86; American Type Culture Collection) human cell lines were maintained as is usual (20). Normal primary cells from mice were maintained in IMDM GlutaMAX (31980030; Thermo Fisher Scientific) supplemented with 10% FBS and 1× Antibiotic-Antimycotic (15240112; Thermo Fisher Scientific).

Mice (BALB/c) were purchased from Koatech Technology (Pyeongtaek, Korea) and maintained in our specific pathogen–free facility. All the mice experiments were performed with the approval (KHUASP(SU)-18-D) of Kyung Hee University Institutional Animal Care and Use Committee and, thereby, according to its guidelines.

Five-week-old mice were immunized primarily by i.p. injection of 100 μg of NP28-CGG (N-5055C-5; Biosearch Technologies) dissolved in 100 μl of PBS together with 100 μl of Imject Alum (77161; Thermo Fisher Scientific).

Mesenteric lymph nodes were dissected out from the immunized mice after a week of primary immunization, unless noted otherwise, fixed with 4% paraformaldehyde (PFA) in PBS at 4°C for 12 h, and followed by treatment with 30% sucrose in PBS at 4°C for 24 h. The lymph nodes were rinsed with and immersed in OCT compound (4583; Tissue-Tek) and, finally, frozen at −70°C overnight. The lymph nodes were then cryo-sectioned with 6-μm widths onto the RNase-free slides, dried at room temperature (RT) for 2 h, and then stored at −20°C. For Figs. 1A and 1B, lymph nodes were dissected out after 3 d of secondary immunization with half the amount of the same immunogen plus adjuvant following a week of primary immunization.

Three hundred and eighty-nine–bp DNA fragments (+723 to +1111) of Crlz-1, as counted from its transcription start site, were cloned into pBluescript-II-KS(+) using 5′-XbaI and 3′-KpnI sites generated during its PCR amplification. For the sense RNA probe, KpnI-linearized plasmid was transcribed by T7 RNA polymerase, whereas for the antisense RNA probe, XbaI-linearized plasmid was transcribed by T3 RNA polymerase. Actually, 200 ng of each linearized plasmid was reacted by 40 U of T7 or T3 RNA polymerase (Promega) in a total volume of 20 μl with 2 μl of 10× DIG RNA Labeling Mix (11277073910; Roche) and 4 μl of 5× transcription buffer (Promega) at 37°C for 2 h. The reaction was treated with 2 U of RQ1 DNase (Promega) for 15 min and then its RNA was column-purified using Prober (iNtRON). The concentration of purified RNA was measured and verified by a dot-blot analysis.

For in situ hybridization (ISH), tissue sections were further fixed with 4% PFA in PBS at RT for 10 min followed by three PBS washes for 5 min each and then treated with proteinase K (50 ng/ml) at RT for 10 min followed by three PBS washes for 5 min each. The washed tissue sections were then prehybridized in 300 μl of hybridization solution (50% deionized formamide, 5× SSC, 5× Denhardt’s Solution, 250 μg/ml yeast tRNA, and 500 μg/ml sonicated salmon sperm DNA) at 50°C for 3 h and then hybridized using 2 ng/μl digoxigenin-labeled RNA probe in the same volume of hybridization solution as above at 50°C overnight. The hybridized tissue sections were consecutively washed twice with 2× SSC plus 50% deionized formamide (Amresco) at 37°C, twice with 2× SSC at 37°C, twice with 2× SSC at RT, and finally once with PBS plus Tween (PBST; 0.1% Tween 20), for 10 min each. The washed tissue sections were blocked using 5% sheep serum in PBST at 4°C for 2 h and treated with 0.75 U per slide of alkaline phosphatase (AP)–conjugated anti-digoxigenin Abs (11093274910; Roche) in 200 μl of blocking solution at 4°C overnight. The Ab-treated tissue sections were then washed thrice with PBST for 10 min each and once with 100 mM Tris–HCl buffer (pH 9.5) for 5 min. The hybridized probe with its bound AP-conjugated Ab was visualized by AP reaction using 200 μl per slide of BCIP/NBT Liquid Substrate System (Sigma-Aldrich) at RT for 20 min. After stopping the AP reaction, the slides of tissue sections were mounted using VectaMount AQ (Vector Laboratories).

For immunohistochemistry (IHC), the tissue sections were further fixed with 4% PFA in PBS at RT for 10 min followed by three PBS washes for 5 min each and then blocked using 5% BSA in PBST at RT for 30 min. For IHC of peanut agglutinin (PNA) or CD21/35, the tissue sections were treated with 1 μg per slide of biotin–PNA (L6135; Sigma-Aldrich) for 1 h or with 200 ng per slide of biotin–anti-CD21/35 (123405; BioLegend) for 2 h in 200 μl of the blocking solution at RT. Tissue sections were washed thrice with PBST for 10 min each and then incubated with 200 ng per slide of AP–streptavidin (S921; Invitrogen) in the blocking solution at RT for 1 h. After washing thrice with PBST for 10 min each and rinsing once with 100 mM Tris–HCl buffer (pH 9.5) for 5 min, the AP reaction was performed using the BCIP/NBT Liquid Substrate System (Sigma-Aldrich) at RT for 20 min to visualize the bound streptavidin. After stopping the AP reaction, the slides were mounted in the same way as above.

For IHC or immunofluorescence (IF) of Wnt3a and Bcl-6, tissue sections were initially processed in the same way as PNA, except blocking was performed with 5% rabbit serum (for Wnt3a IHC), donkey serum (for Wnt3a IF), or BSA (for Bcl-6 IHC). Tissue sections were then incubated overnight at 4°C with 1 μg of primary Abs (anti-Wnt3a: sc-26358, anti–Bcl-6: sc-858; Santa Cruz Biotechnology) in the blocking solution followed by three PBST washes for 10 min each. Tissue sections were then incubated with 1 μg of rabbit anti-goat IgG–AP (sc-2771; Santa Cruz Biotechnology) for Wnt3a IHC, goat anti-rabbit IgG–AP (ab97048; Abcam) for Bcl-6 IHC, or donkey anti-goat IgG–Alexa Fluor 488 (A-11055; Invitrogen) for Wnt3a IF in the blocking solution overnight at 4°C and washed with PBST at RT eight times for 5 min each (for Wnt3a and Bcl-6 IHC) or thrice for 10 min each (for Wnt3a IF, with care to minimize a light exposure). After rinsing them once with 100 mM Tris–HCl buffer (pH 9.5) for 5 min, AP reactions were performed in the same way as for the IHC of PNA, except for their reaction times (40 min for Wnt3a, 30 min for Bcl-6). The slides of AP reaction were mounted in the same way as above, whereas the slides of IF were mounted using Fluorescence Mounting Medium (Dako Products).

For the combined ISH and IF on the same tissue section, which would enable various images of fluorophore signals to be merged, tissue sections were further fixed with 4% PFA in PBS at RT for 10 min, followed by three PBS washes for 5 min each. After blocking with 5% rat serum in PBST at RT for 1 h, tissue sections were incubated with 200 ng per slide of biotin-conjugated anti-CD21/35 (123405; BioLegend) at RT for 2 h, followed by three PBS washes for 10 min each. Tissue sections were additionally fixed with 4% PFA in PBS at RT for 10 min, followed by three PBS washes for 5 min each. The section slides were then processed in the same way as in the ISH above, except performing (pre)hybridization at 37°C, followed by treating with 200 ng per slide for each of FITC-coupled anti-Ki67 (11-5698-80; eBioscience) and Alexa Fluor 350–coupled streptavidin (S11249; Invitrogen) and 2 μg per slide of rhodamine-coupled anti-digoxigenin (11207750910; Roche) in the blocking solution at 4°C overnight. The treated slides were then washed thrice with PBST at RT for 10 min each, taking care to avoid a light exposure. The slides were finally mounted using Fluorescent Mounting Medium (Dako Products).

For H&E staining, tissue sections were washed once with the deionized water for 2 min. The washed sections were then stained with Mayer’s hematoxylin (Dako Products) at RT for 2 min, followed by washing with softly flowing tap water for 5 min. The slides were then destained with 70% ethanol plus 5% HCl for 1 min, followed by washing with softly flowing tap water for 2 min. The tissue sections were next blued in Scott’s Solution (3.5 g of NaHCO₃, 20 g of MgSO₄ in 1 l of deionized water) for 1 min, followed by washing with softly flowing tap water for 5 min and then with 95% ethanol for 1 min. The sections were next stained in the eosin Y solution (HT1101128; Sigma-Aldrich) for 1.5 min, followed by two washes with 95% ethanol for 2 min each and one wash with deionized water for 5 min. The section slides were finally mounted using VectaMount AQ (Vector Laboratories).

The Neon Transfection System (MPK5000; Invitrogen) was used to transfect cells with small interfering RNA (siRNA) or DNA because of its high transfection efficiency without death of cells. The procedure was basically the same as described previously (20), except the use of a 10-μl Neon tip with 1 × 10⁵ cells to transfect 0.25 pmol of Bcl-6 expression plasmid (HG12083-UT; Sino Biological). The pulse condition for Ramos cells was once at 1330 V with a pulse width of 20 ms, whereas the one for primary centroblasts was once at 1300 V with a pulse width of 30 ms. The sequences of siRNA are as follows: scrambled, (sense) 5′-UCC UUC CUC UCU UUC UCU CCC UUG UGA-3′ and (antisense) 5′-UCA CAA GGG AGA GAA AGA GAG GAA GGA-3′; β-catenin, sc-29209, and sc-29210 from Santa Cruz Biotechnology; Crlz-1, (sense) 5′-CCU GAA GAC UAA GUA UAA UCU CUA C-3′ and (antisense) 5′-GUA GAG AUU AUA CUU AGU CUU CAG GUA-3′.

Stock solutions of 65 mM KYA1797K (S8327; Selleckchem, HY-101090; MedChemExpress) and 1 M axitinib (S1005; Sellekchem) were initially made and serially diluted to the working concentrations of 5–12.5 mM in PBS and 25–50 μM in DMSO, respectively. Two milliliter cultures of Ramos or primary centroblast cells at a concentration of 5 × 10⁴ cells/ml were treated with 2 μl of 5–12.5 mM solution of KYA1797K or 25–50 μM solution of axitinib, respectively. In the case of primary Tfh cells, 1 ml cultures of them at a concentration of 5 × 10⁴ cells/ml were treated with 1 μl of 12.5–25 mM solution of KYA1797K. Actually, the final concentrations of KYA1797K and axitinib in each set of parallel cultures corresponded to 5–12.5 μM (Ramos and primary centroblast cells) or 12.5–25 μM (primary Tfh cells) and 25–50 nM, respectively.

The working solutions of 10–40 mM of XAV939 (sc-296704; Santa Cruz Biotechnology), 50–100 μM of niclosamide (N3510; Sigma-Aldrich) and 10–20 mM of cyclosporine A (ab120114; Abcam) were similarly made in DMSO and added to the cultures.

A stock solution of 100 μg/ml Wnt3a (ab81484; Abcam) was made in water, and 4 μl of it was added to 2-ml culture of primary centroblast cells to get a final concentration of 200 ng/ml.

Mice were administered with KYA1797K or axitinib by a daily i.p. injection of the drug from the second day of their primary immunization for a period of 5 d. One hundred microliters each of its diluted solutions (7.25, 14.5, and 29 mg/ml in PBS for KYA1797K or 17.5 and 35 mg/ml in DMSO for axitinib) was used for the daily injection. On the seventh day from the primary immunization with or without administration of Wnt inhibitors, mesenteric lymph nodes, bone marrow cells, and blood were taken and processed for the various analyses, as described in the corresponding sections of experiments, including ISH, IHC, ELISA, ELISpot and flow cytometry.

These methods were performed essentially as described previously (20). The primer sequences with the annealing temperatures as well as the correct sizes of PCR products are given in Supplemental Table I. The Abs from Santa Cruz Biotechnology were anti–β-catenin (sc-7199), anti-Sas10 (sc-135392), anti–Bcl-6 (sc-365618, sc-7388), anti-NFATc2 (sc-514929), anti–β-actin (sc-130656), normal rabbit IgG (sc-2027), and HRP-conjugated mouse IgGκ binding protein (sc-516102). The Abs from Abcam were anti-CBFβ (ab33516), anti-UTP3 (ab60063), anti–β-actin (ab-8227), anti–Lamin-B1 (ab16048), and HRP-conjugated anti-rabbit IgG (ab6721, ab205718).

Immunoprecipitation was performed basically as described previously (23), except for the use of protein A/G plus agarose beads (1861760; Thermo Fisher Scientific). The Abs for immunoprecipitation were anti-PEBP2β (sc-56751; Sant Cruz Biotechnology), anti-UTP3 (ab60063; Abcam), and normal mouse IgG (sc-2025; Sant Cruz Biotechnology). The Abs for immunoblot were the same as the ones of Western blot above.

For the sorting of GC B cells, cells from the mesenteric lymph nodes of immunized mice were forced through 40-μm cell strainer (Falcon) with PBS containing 1× Antibiotic–Antimycotic (Thermo Fisher Scientific), transferred to 1× Red Blood Cell Lysis Solution (Miltenyi Biotec), and incubated at RT for 10 min. After centrifugation, the pelleted cells were resuspended in 50 μl of staining buffer (2% FBS and 0.01% NaN₃ in PBS) per 1 × 10⁶ cells and mixed with 1 μl (0.2–0.5 μg) each of FITC- or allophycocyanin–anti-CXCR4 (551967 or 558644; BD Biosciences) and PE–anti-CD83 (558205; BD Biosciences) Abs per 1 × 10⁶ cells for 20 min at 4°C. The labeled cells were washed twice with the sorting buffer (0.5% BSA and 2 mM EDTA in PBS) and sorted into two populations of centroblasts (CXCR4^highCD83^low) and centrocytes (CXCR4^lowCD83^high) using a FACSAria cell sorter (BD Biosciences). The analysis of population between centroblasts and centrocytes was performed using a Gallios flow cytometer (Beckman Coulter) and FlowJo software. For further analysis of their surface IgM and IgG levels, the same amounts of PerCP/Cy5.5 anti-mouse IgM (562034; BD Biosciences) and PE/Cy7 anti-mouse IgG1 (406613; BioLegend) as those of FITC–anti-CXCR4 and PE–anti-CD83 were added to the cells for four-color flow cytometry.

For the sorting of Tfh cells, the CD4⁺ cells were initially isolated by anti-CD4 microbeads (130-117-043; Miltenyi Biotec) in the MACS system (Miltenyi Biotec) following the manufacturer’s protocol and then sorted out by PE-conjugated anti-ICOS Ab (552146; BD Biosciences) to obtain CD4⁺ICOS⁺ Tfh cells in the FACS system (BD Biosciences).

Cells were labeled using an FITC BrdU Flow Kit (559619; BD Biosciences) following the manufacturer’s instructions. Briefly, the last-counted cells transfected with siRNA or treated with Wnt inhibitors, as described above, were incubated with 10 μl per 1 × 10⁶ cells of 1 mM BrdU for 1 h to label the newly synthesized DNA of proliferating cells and then their BrdU-incorporated DNA was detected by FITC–anti-BrdU Abs. Ten microliter per 1 × 10⁶ cells of 7-aminoactinomycin D (7-AAD) solution that binds to total DNA was also added to the cells when they were treated with FITC–anti-BrdU Abs. The cells labeled with two colors, as above, were run on a Gallios flow cytometer (Beckman Coulter) and analyzed by FlowJo software. As another way of cell cycle analysis, cells were stained with propidium iodide (PI) to quantify their DNA contents. Briefly, Ramos cells transfected with siRNA or treated with Wnt inhibitors, as described above, were fixed in 1% PFA on ice for 1 h and then permeabilized with 70% ethanol. Cells were then collected by centrifugation and treated with 200 μl of RNase A solution (200 μg/ml in PBS, 12091039; Thermo Fisher Scientific) at 37°C for 30 min and stained with 1 ml of 50 μg/ml PI (P4170; Sigma-Aldrich) in PBS at RT for 30 min. After two washes with PBS, cells with their DNA stained with PI were run on a Gallios flow cytometer (Beckman Coulter) and analyzed by Flowing Software, which was downloaded free from the University of Turku (Cell Imaging Core, Turku Center for Biotechnology, Turku, Finland).

Cells, which were transfected with TYE563-labled scrambled siRNA or Cy5-labeled Crlz-1 siRNA, as above, were mounted onto poly-l-lysine–coated coverslips and processed as described previously (20). This time, however, we used goat serum, anti-CBFβ (ab33516; Abcam) and Alexa Fluor 488–labeled goat anti-rabbit IgG (ab150077; Abcam).

For ELISA, 96-well plates (Corning) were coated with 5 μg/ml NP30– or NP8–BSA in PBS at 4°C overnight. The plates were washed thrice with PBST (0.05% Tween 20 in PBS) and blocked in a blocking buffer (0.1% Tween 20 and 1% BSA in PBS) at RT for 2 h, followed by three PBST washes. While blocking, the serially diluted sera were prepared from blood samples, which were collected from the orbital sinus of immunized mice with or without administration of Wnt inhibitors by clotting at RT for 2 h and then centrifuging at 6000 × g at 4°C for 20 min. The serially diluted sera were added to the plates as coated above and then incubated at RT for 2 h. The plates were washed five times with PBST and treated with a detection Ab (HRP-conjugated anti-mouse IgG or IgM) diluted 1:5000 in an assay buffer (0.05% Tween 20 and 0.05% BSA in PBS) at RT for 1 h, followed by five PBST washes. TMB Substrate Solution (N301; Thermo Fisher Scientific) was added to the plates. After an appropriate time of incubation, the reaction was stopped using a stop solution (N600; Thermo Fisher Scientific). The plates were finally read at 450 nm using SpectraMAX 190 Microplate Reader (Molecular Devices).

For ELISpot, Immobilon-P 96-well plates (Millipore) were treated with 70% ethanol at RT for 2 min followed by three washes with water and coated with 30 μg/ml NP30–BSA (100 μl/well) in PBS at 4°C overnight. After washing sequentially four times with PBS and two times with culture media (IMDM GlutaMAX containing 10% FBS and 1× Antibiotic-Antimycotic), the plates were incubated with 200 μl per well of the culture media at RT for 2 h. Cell suspensions of bone marrow isolated from the immunized mice with or without treatment of Wnt inhibitors were serially diluted and added to the plates and then incubated at 37°C for 24 h in an atmosphere of 5% CO₂ and water saturation. The plates were washed twice with PBS and thrice with PBST, followed by soaking for 3 min each. One hundred microliters of detection Ab (catalog no. 7076, HRP-conjugated anti-mouse IgG; Cell Signaling Technology) diluted 1:500 using 10% FBS in PBS was added to the plates and incubated at RT for 1 h. After washing the plates five times with PBST followed by soaking for 3 min each and, briefly, two times with PBS, 3-amino-9-ethyl-carbazole (AEC) substrate solution was added to the plates for the development of spots. The AEC substrate solution was prepared by adding 333.3 μl of 100 mg/ml AEC in N,N-dimethylformamide to 10 ml 0.1 M sodium acetate solution (pH 5) and then passing the mixture through 0.45-μm syringe filter and, finally, adding 5 μl of 30% H₂O₂. The development of spots was monitored until distinct spots would emerge, then stopped by washing extensively with water. After drying the plates in air completely, the spots were inspected and counted automatically using an ImmunoSpot S6 Versa Analyzer (Cellular Technology).

Error bars in the graphs represent SEM. When necessary, the paired two-tailed Student t test was employed to show statistical significance. A p value ≤0.05 was considered statistically significant.

ISH for Crlz-1 was performed to histologically localize its expression sites within the lymph nodes of immunized mice, as Crlz-1 had been found to be expressed in the lymphoid organs in our previous Northern blot (21). In these ISH experiments, Crlz-1 expression was anticipated within the DZ of GC, where the rapidly proliferating centroblasts are located, based on our previous reports that Crlz-1 is especially expressed in the proliferating cells such as pre-B cells, spermatogonia, and Sertoli cells (19–21). To our anticipation, Crlz-1 expression was indeed detected within the DZ of GC, initially by AP reaction (Fig. 1A) and then by more sensitive IF (Fig. 1B). Experimentally, ISH and IHC were performed separately on the adjacent tissue sections for AP reactions, whereas they were performed together on the same tissue section for IF. ISH signals for Crlz-1 were shown to be overlapped with IHC signals of PNA by the AP method (Fig. 1A) as well as with the ones of Ki67 DZ but not CD21/35 LZ markers by the IF method (Fig. 1B). PNA is a GC marker for centroblasts and centrocytes, and thereby more detectable within the DZ of densely populating centroblasts. Ki67 is a marker of proliferating centroblasts in the DZ, and CD21/35 is a marker of FDC in the LZ. When the red image of Crlz-1 was merged variously with the green one of Ki67 and/or the blue one of CD21/35, the red signals of Crlz-1 turned out to be colocalized with the green ones of Ki-67 but not the blue ones of CD21/35 (Fig. 1B), thereby suggesting that Crlz-1 should be expressed in the centroblasts within the DZ of GC.

Wnt3a was also found to be expressed in GC, as compared with PNA, when IHC was performed by AP reaction using the adjacent tissue sections (Fig. 1C), and mainly localized within the DZ, as demonstrated by merging the red IF image of Wnt3a with the green one of Ki67 (Fig. 1D). As our results would demonstrate repeatedly in the following sections, Wnt3a expressed within the DZ of GC might have autocrine and/or paracrine effects on centroblasts to cause their rapid proliferation by regulating the expression of Crlz-1 and, thereby, Bcl-6 and its various target genes.

To confirm the expression of Crlz-1 in centroblasts within the DZ of GC, as observed histologically in Fig. 1, and to study conveniently its functional roles at both cellular and molecular levels, we had sought a centroblast-like cell line. We came across the reports that Ramos and Raji cells from Burkitt lymphomas, which were originated from GC B lymphoblasts, are centroblast and centrocyte like, respectively (31, 32). To verify these reports and use those cell lines as such cells, we sorted out mouse normal primary centroblasts and centrocytes and compared them with Ramos and Raji cells. Experimentally, based on the expression levels of CXCR4 and CD83 as their surface markers (3, 33–35), the cells isolated from the mesenteric lymph nodes of immunized mice were sorted out into two populations. One of them was CXCR4^highCD83^low within quadrant (Q)4 and defined to be centroblasts, whereas the other one was CXCR4^lowCD83^high within Q1 and defined to be centrocytes (Fig. 2A). Initially, the expression of CXCR4 and CD83 was checked by RT-PCR in the primary centroblasts and centrocytes as well as Ramos and Raji cells to confirm their identities. As anticipated, CXCR4 was high but CD83 was low in the primary centroblasts and Ramos cells, whereas the opposite was the case in the primary centrocytes and Raji cells (Fig. 2B), indicating that Ramos and Raji cells were indeed centroblast and centrocyte like, respectively. Surprisingly enough, when we checked the expression of Crlz-1 and its related genes, including Wnt3a, β-catenin, LEF-1, CBFβ, and Runx1, as well as Bcl-6 and its target genes, such as AID and cyclins D1–D3, all the tested genes were found to be highly expressed in the primary centroblasts as compared with the primary centrocytes (Fig. 2B, left gels), suggesting that these genes might be potentially involved in the proliferation and function of centroblasts during GC reaction. Furthermore, this differential expression pattern of checked genes between the primary centroblasts and centrocytes was also observed similarly between Ramos and Raji cells (Fig. 2B, right gels), strongly supporting our confident usage of these cells as centroblast- and centrocyte-like cells, respectively. Notably, Crlz-1 was definitely expressed highly in the primary centroblast and Ramos cells but scantly in the primary centrocyte and Raji cells (Fig. 2B), assuring our histological results in Fig. 1. Therefore, the expression and function of Crlz-1 in centroblasts could be conveniently studied at both molecular and cellular levels by using Ramos cells.

Crlz-1 was first studied in Ramos cells with regard to its effects on their proliferation and gene expression by transfecting them with siRNA for Crlz-1 as well as β-catenin because we suspected that Crlz-1 expression in centroblasts might be related to their proliferation through the regulation of its target genes. Indeed, the proliferation of Ramos cells was decreased in these siRNA transfections (Fig. 3A). Significantly, we noticed that Bcl-6 was decreased by these siRNA transfections (Fig. 3B–D). This finding was very significant because Bcl-6 is the master regulator of GC reaction (10, 14, 16). Certainly, β-catenin siRNA was shown to diminish both Crlz-1 and Bcl-6 expression, whereas Crlz-1 siRNA was shown to diminish Bcl-6 but not β-catenin expression, as demonstrated at the levels of not only mRNA (Fig. 3B, 3C) but also proteins (Fig. 3D), suggesting that Crlz-1 might relay the upstream Wnt/β-catenin signal to the downstream expression of Bcl-6. To mechanistically explain the regulation of Bcl-6 by Crlz-1, the possibility that Bcl-6 could be regulated by Runx/CBFβ was envisioned, based on our previous reports that CBFβ was mobilized into the nucleus by Crlz-1 to allow Runx/CBFβ heterodimerization when their cDNAs were cotransfected (23), whereas the nuclear CBFβ in Crlz-1–positive pre-B cells disappeared along with its cytoplasmic reappearance when Crlz-1 siRNA was transfected (20). Interestingly, when the human Bcl-6 promoter was scrutinized to identify the Runx/CBFβ–binding TGTGGT consensus, a far distal one at −3501 and two proximal ones at −436 and −516 were identified from its transcription start site. Indeed, the chromatin region including the two proximal ones turned out to be bound by Runx/CBFβ in Ramos cells, as this chromatin was captured by anti-CBFβ in our chromatin immunoprecipitation (ChIP) assay (Fig. 3J). Furthermore, this ChIP band disappeared when Crlz-1 or β-catenin was knocked down by its siRNA (Fig. 3J), while nuclear CBFβ was diminished by the same siRNA transfections (Fig. 3I), corroborating the regulatory axis from Wnt/β-catenin to Bcl-6 through Crlz-1 and Runx/CBFβ. Certainly, the binding interaction between Crlz-1 and CBFβ as the fundamental basis of Crlz-1 function was again demonstrated in Ramos cells by performing coimmunoprecipitation experiments, in which Crlz-1 was surely shown to be detected in the immunoprecipitate by anti-CBFβ Ab and vice versa (Fig. 3K).

The decreased expression of Bcl-6 by the knockdown of Crlz-1 as well as β-catenin led, eventually, to the decreased or increased expressions of its target genes such as cyclins D1–D3, p53, Blimp-1, and IRF-4 (Fig. 3B, 3C). High expression of cyclins D1–D3, but low expression of p53 are required for the rapid centroblastic proliferation without stopping of cell cycle to repair DNA breaks during SHM as well as CSR at the stage of centroblasts of GC reaction (10). This fact explained the decreased proliferation of Ramos cells when Crlz-1 or β-catenin was knocked down (Fig. 3A), leading to the decreased expression of Bcl-6 and, thereby, the decreased expression of cyclins D1–D3 and the increased expression of p53 (Fig. 3B, 3C). In these contexts, cells were blocked at the G1 checkpoint, resulting in the population increase in the G1 phase, as shown by flow cytometric cell cycle analyses after staining them doubly with FITC–BrdU and 7-AAD (Fig. 3E, 3F) as well as singly with PI (Fig. 3G, 3H). Interestingly, Ramos cells were found mostly in the S phase and very scantly in G1 and G2/M phases in these flow cytometric cell cycle analyses of cells stained doubly with FITC-BrdU and 7-AAD (Fig. 3E), explaining the unusual shape of the flow cytometric diagram of cells stained singly with PI, which showed a big S phase peak with its embedded small shoulder peaks of G1 and G2/M phases (Fig. 3G). The population increases in G1 phase after knockdown of Crlz-1 or β-catenin were so dramatic that they were easily recognizable in the flow cytometric diagrams (Fig. 3E, 3G). The population increase in G1 phase had already been expected by the fact that cyclins D1–D3 were decreased by the knockdown of Crlz-1 or β-catenin (Fig. 3B).

Blimp-1 and IRF-4, which are repressed by Bcl-6, should be released from it for centroblasts to stop their proliferation and differentiate into plasma cells (10, 36), which express IgJ as their differentiation marker (19). Blimp-1 and IRF-4 were indeed shown to be derepressed through the decreased expression of Bcl-6 when Crlz-1 as well as β-catenin were knocked down (Fig. 3B), indicating that Crlz-1 might maintain the proliferation of centroblasts, in part, by repressing Blimp-1 and IRF-4 through the regulation of Bcl-6 and thereby blocking their differentiation into plasma cells. The decreased proliferation of Ramos cells in these siRNA transfections could not be explained by the death of cells because any significant cell death was not observed.

Because Crlz-1 turned out to be targeted by the Wnt signal in Ramos cells (Fig. 3) as in pre-B cells (20, 22), we used several Wnt inhibitors, including KYA1797K, which is water-soluble and reported to be a powerful Wnt inhibitor (37), axitinib (38), XAV939 (39), and niclosamide (40) for the functional studies of Crlz-1 in these cells (see Supplemental Fig. 1 for axitinib, XAV939 and niclosamide).

As shown in Fig. 4B, KYA1797K decreased the expression of Crlz-1 as well as β-catenin. Furthermore, the decreased Crlz-1 expression led to the decreased expression of Bcl-6 and thereby decreased or increased expression of its target genes, such as cyclins D1–D3, p53, Blimp-1, and IRF-4 (Fig. 4B, 4C), causing the decreased proliferation of Ramos cells (Fig. 4A). The decreased expression of β-catenin, Crlz-1, and Bcl-6 by KYA1797K was also demonstrated at the protein level (Fig. 4D), confirming the regulatory axis from Wnt to Bcl-6 through Crlz-1. These results, obtained by treating Ramos cells with Wnt inhibitors, were basically the same as those obtained when Crlz-1 or β-catenin was knocked down with its siRNA, indicating, again, that Crlz-1 links Wnt/β-catenin signal to the regulation of Bcl-6 and thereby its target genes to maintain the proliferation of centroblasts while preventing their differentiation to plasma cells.

Confirmatively, just as in Crlz-1 or β-catenin knockdown, KYA1797K treatment also resulted in G1 block (Fig. 4E, 4F), the decreased nuclear mobilization of CBFβ (Fig. 4G), and, thereby, disappearance of Runx/CBFβ binding to the Bcl-6 promoter (Fig. 4H). Again, as in siRNA transfections, the decreased proliferation by KYA1797K could not be explained by the death of cells because any significant cell death was not observed.

To verify the results obtained using Ramos cells in normal primary centroblasts, the centroblasts of mesenteric lymph nodes were sorted out by labeling them with allophycocyanin–CXCR4 and PE–CD83 Abs (Fig. 5A). When the sorted primary centroblasts were transfected with siRNA for Crlz-1 as well as β-catenin or treated with KYA1797K, their proliferation was surely decreased (Fig. 5B, 5C), while their expression of Crlz-1 and/or β-catenin and, thereby, Bcl-6 was decreased (Fig. 5D–G), just as in Ramos cells. The decreased expression of Bcl-6 led eventually to the decreased or increased expression of its target genes: the expression of cyclins D1–D3 was decreased, whereas those of p53, Blimp-1, and IRF-4 were increased (Fig. 5D–G). Furthermore, the nuclear CBFβ in primary centroblasts was found to disappear when Crlz-1 was knocked down by its siRNA transfection (Fig. 5H), indicating that CBFβ could not be mobilized into the nucleus and/or might be shuttled back to the cytoplasm in the absence of Crlz-1. As observed in the confocal images, CBFβ was localized in both the cytoplasm and the nucleus in the scrambled siRNA–transfected cells (Fig. 5H, left images), whereas it was localized only in the cytoplasm in the Crlz-1 siRNA-transfected cells (Fig. 5H, right images). Therefore, all those results obtained using Ramos cells were verified in the normal primary centroblasts, consolidating the critical role of Crlz-1 at the stage of rapidly proliferating centroblasts during GC reaction. Niclosamide and XAV939 were also tried in the primary centroblasts to obtain basically the same results (Supplemental Fig. 1). Again, any significant cell death of primary cells was not observed and could not be the cause of decreased proliferation.

The decreased proliferation after Crlz-1 or β-catenin siRNA transfection and KYA1797K treatment was rescued by Bcl-6 cDNA transfection (Fig. 6A) and Wnt3a addition (Fig. 6C) in Ramos and primary centroblast cells, respectively. Experimentally, both Bcl-6 cDNA transfection and Wnt3a addition were executed after 48 h of siRNA transfection and KYA1797K treatment, respectively, and then chased to observe their rescue of proliferation for an additional 48 h. The last-counted cells in these rescue experiments were checked for the expressions of β-catenin, Crlz-1, and Bcl-6 by RT-PCR and/or Western blot (Fig. 6B, 6D). Consistent with the rescue of proliferation, they were well correlated with both the cascading gene regulatory axis and the experimental processes, as illustrated in the corresponding panels (Fig. 6A–D).

Cyclosporine A, which is an inhibitor of calcineurin in the noncanonical Wnt5a pathway through NFAT and NF-κB, as reported by Kim et al. (41), was shown not to affect the proliferation of the primary centroblasts (Fig. 7A) and Ramos cells (Fig. 7B). Consistently, cyclosporine A, which indeed inhibited the nuclear mobilization of NFAT in Ramos cells, did not affect the nuclear levels of β-catenin, Crlz-1, and Bcl-6 proteins in these cells (Fig. 7C, see 26Discussion), indicating that the canonical Wnt3a pathway should regulate the rapid proliferation of centroblasts truly via β-catenin, Crlz-1, and Bcl-6.

Based on our results above, we anticipated that GC reactions would be impaired when Wnt inhibitors were administered into the immunized mice. This hypothesis was initially tested by performing IHC of PNA, using several spaced tissue sections prepared from the lymph nodes of immunized mice with or without KYA1797K administration. Much more than anticipated, to our surprise, GC in these mice was found to be impaired by the administration of KYA1797K by about one half at a low dose (0.725 mg per mouse) and abolished almost or completely at a medium dose (1.45 mg per mouse) or a high dose (2.9 mg per mouse) as compared with the control mice injected with PBS only (Fig. 8A). This impairment of GC, as observed by the IHC of PNA, should certainly reflect the lack of GC B cells consisting of centroblasts and centrocytes within it, as PNA is their surface marker. However, other cells, such as Tfh and FDC, which are also known to participate in the GC reaction, were demonstrated not to be affected by KYA1797K. To check whether Tfh cells were affected by KYA1797K, these cells were sorted out from the immunized mice (see Supplemental Fig. 3B for their sorting diagram), treated with the drug, and examined for its effects on their proliferation and gene expression. Tfh cells were shown not to be affected by KYA1797K with regards to their proliferation (Supplemental Fig. 2A) as well as their expression of β-catenin, Crlz-1, and Bcl-6 genes (Supplemental Fig. 2B). Furthermore, it should be noted that the expression levels of these genes in Tfh cells are much lower than those in centroblast cells (Supplemental Fig. 2C). The effects of KYA1797K on FDC were also checked by IHC with CD21/35, together with a control IHC with PNA, using adjacent tissue sections. As shown in Supplemental Fig. 2D, FDC was found to be maintained despite of the impaired or abolished GC by the administration of KYA1797K, indicating that the drug did not affect FDC, at least with regard to its presence.

We repeated the same experiments using axitinib, which inhibits Wnt signaling by inducing β-catenin degradation (38). Although axitinib is water-insoluble and probably not as good as water-soluble KYA1797K (37) in terms of bioavailability, this drug also impaired GC formation (Supplemental Fig. 1D) but less prominently as compared with KYA1797K.

Adjacent tissue sections showing the impaired GC at the low dose of KYA1797K were further assessed to see whether the expression of Wnt3a and Crlz-1 as well as Bcl-6 was actually inhibited among the remnant GC. Confirmatively, the expression of Wnt3a and Crlz-1 and thereby Bcl-6 was surely decreased (Fig. 8B), supporting that the impairment or abolishment of GC was due to the decreased expression of these genes, and thereby the diminished proliferation of centroblasts as shown in the flow cytometric diagrams (Fig. 8C) and their statistical analyses (Fig. 8D). Notably, the decreased expression of Wnt3a by KYA1797K reminded us of its autoregulatory mechanism as in the case of β-catenin (42). Of course, the decrease or absence of these expressions should be the natural consequence of impaired or abolished GC, and vice versa, after administration of KYA1797K. Thus, histologic experiments to check the expression of those genes at medium or high drug doses were not necessary, as GC at such doses was already abolished almost or completely (Fig. 8A).

The diminished population of centroblasts in impaired GC was obviously demonstrated by the flow cytometric analyses of cells isolated from the mesenteric lymph nodes of mice administered with KYA1797K after their immunization (Fig. 8C, 8D). The population of centroblasts dropped faster than the one of centrocytes by KYA1797K, became similar with it at the low dose, and dropped even below it, with their population ratio sometimes reversed, at the medium dose (Fig. 8D). Convincingly, in these centroblasts sorted out from the remnant GC at the low dose, the expression of all the genes in the regulatory axis, including β-catenin and Crlz-1 as well as Bcl-6 and its target genes, such as cyclins D1–D3, p53, Blimp-1, and IRF-4, was found to be affected similarly (Fig. 8E, 8F) as in the experiments using Ramos (Fig. 4B) and primary centroblast cells (Fig. 5F).

The impaired or abolished GC reaction after administration of KYA1797K led to the consequential decrements of Ab affinity maturation and CSR as well as a smaller output of plasma cells. Certainly, the output of anti-NP IgG-secreting plasma cells dropped by KYA1797K (Fig. 9E), as measured by ELISpot using the cells from bone marrow, where the long-lived and high-affinity Ab-secreting GC-derived plasma cells are retained (43, 44). A smaller output of plasma cells might be due to their earlier differentiation without a further centroblastic proliferation in the presence of KYA1797K. The earlier differentiation of centroblasts into plasma cells could be speculated, based on our results that the plasma cell differentiation-related genes, such as Blimp-1 and IRF-4 as well as IgJ as its marker gene, were increased by KYA1797K treatment in Ramos (Fig. 4B) and primary centroblast cells (Fig. 5F) and in such cells isolated from the KYA1797K-administered mice (Fig. 8E). Consistently, the ratio of anti-NP8 versus anti-NP30 IgG Abs as a degree of high-affinity Ab also dropped by KYA1797K, as measured by ELISA using serially diluted sera (Fig. 9B–D).

CSR of Ig μ to γ genes during GC reaction was also blocked by the administration of KYA1797K and stayed mostly in their original expression state of surface IgM (IgM^+/highIgG1^−/low; Q1), as demonstrated by four-color flow cytometry for centroblasts (Fig. 9F, 9G) and centrocytes (Fig. 9H, 9I), which were initially gated from the lymph node cells using their surface markers (Supplemental Fig. 3A) and subsequently analyzed for their surface levels of IgM and IgG. Without KYA1797K administration (0 mg per mouse), two major populations of centroblasts with surface expression profiles of IgM^−/lowIgG1^−/low (Q4) and IgM^−/lowIgG1^+/high (Q3; class-switched population) showed up, whereas only one major population of centrocytes with a profile of IgM^−/lowIgG1^−/low (Q4) showed up (see 26Discussion). However, anti-NP30 IgG in blood was moderately diminished by KYA1797K (Fig. 9B), whereas anti-NP30 IgM in blood was not much affected by it (Fig. 9A), as assayed by ELISA using serially diluted sera, indicating that most IgM Abs might be derived from extrafollicular foci and/or other histological loci than GC, and CSR to IgG might also occur there during primary immune responses.

Crlz-1 was originally cloned because of the ability of its protein to bind CBFβ by the yeast two-hybrid method and named as such, indicating the presence of many charged amino acids and leucine zipper–like motifs (45). Based on this information, we speculated that Crlz-1 might bind to CBFβ using these characteristic amino acid sequences and motifs in the cytoplasm and then carry CBFβ into the nucleus using its nuclear localization signal, which was identified to be KRAI of amino acid residues from 398 to 401 (S.Y. Choi, J.H. Pi, S.-K. Park, and C.J. Kang, unpublished results) in the N-terminal end of SAS10 homology domain (45). KRAI was well conserved among all the species that we have examined. However, a CBFβ-binding motif could not be mapped with a single linear stretch of amino acids so far, indicating that the motif might be a conformational one with several distanced amino acid sequences. Furthermore, as Crlz-1 was rapidly degraded by a proteasome-independent mechanism in pre-B cells (20), the regulatory role of Crlz-1 in centroblasts might also be enabled by its rapid turnover and, so, its dynamic regulation of Runx/CBFβ heterodimerization with nuclear mobilization of CBFβ in its presence and/or cytoplasmic shuttling back of CBFβ in its absence.

There have been at least three known different Wnt pathways: the canonical pathway involving β-catenin, the planar cell polarity pathway, and the Ca⁺² pathway. The latter two are called as the noncanonical pathways. In our hands, the rapid proliferation of centroblasts turned out to be dependent on the canonical pathway but not on the noncanonical Ca²⁺ pathway. Actually, cyclosporine A as an inhibitor of calcineurin, which was proposed to participate in the noncanonical Wnt5a/Ca²⁺/NFAT/NF-κB/Bcl-6 pathway to protect GC B cells from apoptosis (41), could not affect the expression of β-catenin, Crlz-1, and Bcl-6, and thereby the centroblastic proliferation, which, in this paper, was definitely shown to be dependent on the canonical Wnt3a/β-catenin pathway through the Crlz-1/CBFβ/Bcl-6 regulatory axis. In the report by Kim et al. (41), we noticed that Wnt5a had an inconsistent effect on the expression of Bcl-6, despite its protective antiapoptotic effects on GC B cells, depending on the experimental time points and the sources of GC B cells.

CSR of Ig genes during GC reaction was also blocked by KYA1797K, as BCR consisting of original μ chains was mostly detected in both centroblasts and centrocytes in the drug-administered mice. However, when those of control mice without the administration of KYA1797K were checked, two populations of centroblasts and one population of centrocytes were detected in terms of surface levels of IgM and IgG. In the case of centroblasts, one population was IgM^−/lowIgG^+/high, indicating that the cells have experienced CSR of Ig genes from μ to γ, and the other one was IgM^−/lowIgG^−/low. Interestingly, in the case of centrocytes, the only population was IgM^−/lowIgG^−/low. One explanation for the population of negative or low IgM and IgG in both centroblasts and centrocytes could be that cells might be in a state of negative or low surface Ig expression during their rapid proliferation in the case of centroblasts and temporarily just after their transition from DZ to LZ in the case of centrocytes. Another explanation could be that a significant portion of those cells experiencing SHM as well as CSR of Ig genes might fail to express them because of their changes of reading frames, stops of translation, and/or errors of recombination, etc. (35).

We also noticed that centrocytes with the class-switched IgG on their membranes were almost not detected in the control mice. There might be several explanations for this phenomenon. One possibility could be that such centrocytes might have already differentiated into the plasma cells secreting IgG. Another possibility could be that their BCR of class-switched IgG might be internalized with Ag by endocytosis to display the antigenic peptide on MHC class II and present it to Tfh cells.

We thank Dr. J. H. Lim at Columbia University for comments on the manuscript.

The authors have no financial conflicts of interest.