Patients’ leftover cells fixed in Carnoy’s solution were used for FISH analysis under IRB approval (#605). Patients diagnosed with lymphoma were asked to participate in the study. The fixed samples were stored at -20°C until use. To perform DNA-FISH, cells were spotted onto glass slides and let dry. After pepsin treatment (2% pepsin in 0.01N HCl, 1 h at room temperature), cells were washed in PBS and dehydrated through an ethanol gradient (70%, 90%, and 100% for 10 min each). Samples were heated at 80°C for 3 min to separate DNA strands, then 1.25–12.5 ng/μL of probes in hybridization buffer (4xSSC, 0.5 mM EDTA, 10% dextran sulfate, 25% deionized-formamide in deionized H₂O) were applied and incubated at room temperature for 2 h before imaging. Locus Specific Identifier (LSI) probe for TFL was prepared from clone RP1-281H8 on chromosome 6q25.1-25.3 (BACPAC Resources Center) with QIAGEN Large-Construct Kit (Qiagen, Gaithersburg, MD) and green fluorescein-labeled Vysis CEP 6 (D6Z1) (Abbott, Lake Bluff, IL) was used as centromere probe (CEP). LSI probe was labeled with orange dUTP with nick translational kit (Abbott). At least 100 cells were evaluated, and when 5% or more of the evaluated cells were CEP>LSI, these samples were defined as loss of TFL locus based on the background in the normal samples. On the contrary, when 5% or more of evaluated cells were CEP<LSI, these samples were defined as a gain of TFL locus. Otherwise, samples were determined as having no TFL deletion (CEP=LSI).

Bone marrow, lymph node, and tissue samples of patients diagnosed with hematological malignancies were collected for the purpose of diagnosis. The leftover samples were used for FISH and RNA blot analysis with written informed consent or an opt-out approach under institutional review board approval at Kobe University (approval #605).

RNA blots were hybridized with ³²P-labeled human TFL cDNA probes. A 963-bp human TFL probe, which includes the complete open reading frame of TFL2 (10), was amplified using a forward primer 5’-ATGGAGCACCCCAGCAAGATGGAATTC-3’, and a reverse primer, 5’-CTACCCACCATAAGGACAATGCTGC-3’. The image was measured with the ImageJ software (ImageJ, RRID: SCR_003070). Ethidium bromide (EtBr) was used as a loading control. Each region of interest for RNA blot and EtBr image was gated, and the pixel density was captured. The relative quantification value is determined by the ratio of the net band to net loading control.

Mice were under the husbandry care of the Institute for Experimental Animals, Kobe University Graduate School of Medicine. Tfl ^-/- mice generated by gene targeting (Minagawa et al.) were back-crossed for more than twelve generations into a C57BL/6 background. VavP-Bcl2-Tg mice were kindly provided by Dr. Cory and Dr. Bouillet (The Walter and Eliza Hall Institute of Medical Research, Department of Medical Biology, University of Melbourne). Wild-type C57BL/6 mice were purchased from CLEA Japan (Chiba, Japan). Animals were maintained under specific pathogen-free conditions and on 12 h light/12 h darkness cycle. Both female and male mice were used in these studies otherwise specified. The Animal Care and Use Committee of Kobe University approved all animal studies.

Paraffin-embedded sections were used for the immunohistochemical staining of Peanut Agglutinin (PNA), Ki-67, and NFκB1. After inhibiting endogenous peroxidase activity with methanol-containing 0.3% hydrogen peroxidase for 30 min, antigen retrieval was done with citrate buffer (pH 6.0). Sections were blocked with 5% goat serum (Jackson Immunoreseach Labs, West Grove, PA) and 5% FBS (Thermo Fisher Scientific, Waltham, MA)/PBS for 1 h. Primary biotinylated PNA (Vector Laboratories, Burlingame, CA), or anti-Ki-67 or NFκB1 antibody (Abcam, Cambridge, MA) was used at optimized dilutions and detected by the avidin-biotin-horseradish peroxidase complex (ABC) system (Vector Laboratories) using horseradish peroxidase (HRP), or HRP polymer for mouse tissue, Histostar™ (MBL, Nagoya, Japan). Sections were counterstained with hematoxylin. Proportion of positive cells were calculated with the ImageJ software.

The reagents for flow cytometry were from BD Biosciences (San Jose, CA). Cells were suspended in PBS/0.5% BSA/2 mM EDTA. Cell analyses were performed on BD FACScan and Accuri flow cytometer with BD CellQuest and FlowJo software (FlowJo, RRID: SCR_008520, BD Biosciences). Live-dead cell and singlet gating was performed during the analysis and cell sorting. For sorting of bone marrow cells (B220⁺, B220^-IgM^-, or B220^-IgM⁺), bone marrow cells were harvested by flushing the femur aseptically in RPMI 1640 (Sigma-Aldrich, St. Louis, MO) and single-cell suspension was obtained by gentle aspiration several times. Cells were washed with PBS/0.1% BSA/2 mM EDTA and stained with FITC-anti-IgM (BD Pharmingen, San Diego, CA) and PE-anti-B220 (BD Pharmingen) antibodies. Cell analysis and sorting were performed on a MoFlo XDP flow cytometer with Summit software (Beckman Coulter, Brea, CA).

Bone marrow cells were obtained as described above. Splenocytes were harvested by gentle aspiration with RPMI 1640. Soon after bone marrow cells and Splenocytes were harvested, and fractionated bone marrow cells (B220⁺, B220^-IgM^-, or B220^-IgM⁺ cells) were sorted, they were cultured at a concentration of 10⁶ cells/well in 500μL RPMI 1640 supplemented with 10% FBS in a non-coated 24-well dish at 37°C, 5% CO². The cell culture supernatant was harvested for Cxcl13 ELISA at 96 h.

We compared RNAs of bone marrow B220^-IgM⁺ cells from VavP-Bcl2Tg and Vavp-Bcl2Tg/Tfl ^-/- in the two independent experiments. We used a cDNA array chip with SurePrint G3 Mouse GE microarray kit 8x60k (G4852A, database: mm9:NCBI37:Jul2007, Agilent Technologies, Santa Clara, CA). The chip was hybridized with labeled cDNA probes prepared by reverse transcription from 100 ng polyA mRNA using the respective protocol from Agilent Technologies. Probes used for hybridization were labeled with Cy3 dye. Overnight incubation was followed by stringent washing as recommended by the manufacturer. The hybridization signals were scanned and evaluated with a scanner (G2539A) and Scan Control software (version A.8.5.1) and digitalized with Feature Extraction software (Agilent Technologies, version 10.10.1.1). The obtained data were analyzed with GeneSpring GX (version 12.6.1, Agilent Technologies). These procedures were outsourced to Oncomics (Lui-dong, Korea).

Samples were mixed with 1.5 ml TRIzol (Life Technologies, Carlsbad, CA) and stored at -80 ˚C. RNA extraction and quantitative RT-PCR (qPCR) were performed as described previously (14). Briefly, qPCR was performed with gene-specific primers with SYBR Green dye (ThermoFisher Scientific, Waltham, MA). The sequence of each primer is available upon request. Each gene expression was normalized with the value of beta-actin. For the cDNA array, RNAs were purified with an RNeasy kit (Qiagen). The primers used are available upon request.

Quantification for Cxcl13 in bone marrow extra fluid, plasma, and supernatant of culture medium was done using a mouse Cxcl13 ELISA Kit (R&D systems, Minneapolis, MN).

The luciferase assay was performed as described previously (14). Briefly, HeLa cells (CLS Cat# 300194/p772_HeLa, RRID: CVCL_0030) were transfected with psiCheck2 plasmids (Promega, Madison, WI) with or without 3’UTR of CXCL13 or IL-2. Cells were co-transfected either with the control or TFL expression plasmid. For suspension cells, Amaxa nucleofector (Lonza, Basel, Swiss) was used for the transfection of those vectors according to the manufacturer’s developed protocols. Luciferase activity was determined using the Dual-Luciferase reporter assay system (Promega, Madison, WI). The expression value of Renilla Luciferase (vector construct containing 3’UTR) was normalized by Firefly Luciferase (internal control). Experiments were triplicated, and the median value was chosen for the data. Obtained data with the TFL vector were further normalized by those with the control vector (no TFL expression). We confirmed >10,000 relative light units (RLU) in Renilla Luciferase and >1,000 RLU in Firefly Luciferase activity to ensure the transfection efficacy in each cell line. Each experiment was performed at least 3 times to show standard error and perform statistical analysis.

All data are presented as the means ± SEM. Kaplan-Meier with log-rank test was used to estimate survival curves and for statistical comparison of the two groups. The student’s t-test was applied when comparing two independent groups. All the statistical analyses were performed using GraphPad Prism software (GraphPad Prism, RRID: SCR_002798, GraphPad, San Diego, CA).

First, we sought clinical relevance between the loss of TFL and the development of lymphoid neoplasms. We made a fluorescence in situ hybridization (FISH) probe for TFL locus detection, covering 110 kbp around TFL located lesion, 6q25 ( Figures 1A, B ), and tested FISH analysis for bone marrow and lymph node or tissue samples from lymphoid malignancy patients (IRB#605). In 164 bone marrow samples, 124 samples were diagnosed as mature B cell neoplasm, of which 12.9% of patients showed TFL locus deletion (16 in 124 patients; Figure 1C , Supplemental Table 1 ). For the lymph node or tissue samples, 13.6% of patients diagnosed with mature B cell neoplasm were detected as TFL deletion (12 in 88 patients; Figure 1D , Supplemental Table 2 ). TFL locus deletion was also seen in other types of lymphoma, such as mature T and NK cell neoplasm (3 in 15 patients) and Hodgkin lymphoma (2 in 15 patients). In FL, 2 patients with TFL loss in the bone marrow samples were Grade 2, and 6 patients with TFL loss in tissue samples were 5 with Grade 2 and 1 with Grade 3, while no TFL deletion was found in those with Grade 1. To confirm the loss of TFL expression, we performed an RNA blot analysis for several lymphoma/leukemia patients. This experiment showed several patients with mature B lymphoid malignancy decreased TFL expression (Patients #3-6 and 13-16; Figures 1E, F ). These data indicate the loss of TFL expression has occurred in mature B cell neoplasms frequently. For FL and DLBCL, TFL expression increased in higher grades of FL and transformed FL (FL/DLBCL) (Patients #10-12), while decreased TFL was seen more in DLBCL (Patients # 3-6).

Tfl ^-/- mouse was developed in our lab previously (12). Their median survival was 119 week (range 82-144 weeks), comparable to wild-type C57BL/6 mice (the median survival: 108-116 weeks). To determine the effect of tumorigenesis caused by loss of TFL in matured B cell neoplasms, we utilized vavP-Bcl2 transgenic (Bcl2-Tg) mice as FL-bearing mice (15). For this aim, we developed VavP-Bcl2-Tg/TFL deficit (Bcl2-Tg/Tfl ^-/-) mice and observed their survival periods. Bcl2-Tg mice developed lymphadenopathy and splenomegaly and died around 60 weeks (median survival of 64 weeks, Figure 2A ), which is significantly shorter than wild-type mice suggesting follicular lymphoma development as reported. Surprisingly, Bcl2-Tg/Tfl ^-/- mice died around 20 weeks earlier than Bcl2-Tg mice (median survival of 38 weeks, p<0.0001). These phenomena occurred in Bcl2-Tg/Tfl ^+/- mice and both female and male mice ( Figure 2A and Supplemental Figure 1 ). While no sex biases were seen in Bcl2-Tg and Bcl2-Tg/Tfl ^-/- mice, female Bcl2-Tg/Tfl ^+/- mice died earlier than male mice. Bcl2-Tg/Tfl ^-/- mice showed similar lymphadenopathy, and the size of splenomegaly was not different from Bcl2-Tg (data not shown). We performed several necropsies, but Bcl2-Tg/Tfl ^-/- mice showed no other malignant tumor except for lymphadenopathy (data not shown). Morphological and histological findings in the spleen, lymph nodes, and bone marrow between those two strains showed no significant difference ( Figure 2B ). No significant morphological transformation of lymphocytes was seen in Bcl2-Tg/Tfl ^-/- mice. Lymphocyte infiltration patterns in the kidney, liver, and lung were also similar between the two groups ( Supplemental Figure 2 ). On the other hand, increased Ki-67 positive cells were noted in Bcl2-Tg/TFL ^-/- mice in the spleen, lymph nodes, and bone marrow compared with Bcl2-Tg mice ( Figure 2C ).

Next, we performed peanut agglutinin (PNA) staining of spleen sections to visualize germinal center B cell proliferation. Bcl2-Tg/Tfl ^-/- mice showed reduced germinal center B cell accumulation in the follicles compared with Bcl2-Tg ( Figure 3A ). MCPIP-1, a TFL family member, has been shown to demonstrate deubiquitination affecting the NFκB pathway (16), and TFL is also suggested to regulate the NFκB pathway (17). NFκB1 in the spleen of Bcl2-Tg and Bcl2-Tg/Tfl ^-/- mice expressed mainly interfollicular area ( Figure 3B ), indicating highly activated intrafollicular B cells in this area (18). Interestingly, increased NFκB1 positive cells were observed in Bcl2-Tg/Tfl ^-/- mice spleen compared with Bcl2-Tg mice. Complete blood counts in peripheral blood showed no significant differences between both groups except platelet counts (p<0.05, Supplemental Figure 3A ). When stratified the results into blood counts at 20-30 weeks and 35-45 weeks, MCV and MCH were significantly decreased (p<0.05), and platelet count tended to decrease (p=0.052) in Bcl2-Tg/Tfl ^-/- mice at 35-45 weeks ( Supplemental Figure 3B ). Interestingly, the body weight of Bcl2-Tg/Tfl ^-/- mice began to decrease at week 25 and nadired at week 35 compared with Bcl2-Tg ( Figure 3C ), suggesting that this body weight loss may contribute to shorter survival.

To clarify the causes of earlier death in Bcl2-Tg/TFL ^-/- mice, we carefully examined their bone marrow and splenocyte phenotypes. We found a prominent B220^-IgM⁺ population in the bone marrow of Bcl2-Tg mice, which was rare in Tfl ^-/- and wild-type mice ( Figure 4A : Wild type 2.0 ± 1.9%, Tfl ^-/- 0.9 ± 0.4%, Bcl-2-Tg 20.0% ± 5.8%, Bcl2-Tg/Tfl ^-/- 14.9% ± 4.1%). While this B220^-IgM⁺ proportion of Bcl2-Tg/Tfl ^-/- was decreased compared with Bcl2-Tg (p<0.01), total cell counts of this population were increased (mean cell count of the wild type was 2.1 x 10⁵, Tfl ^-/- 2.1 x 10⁵, Bcl-2-Tg 27.6 x 10⁵, and Bcl2-Tg/Tfl ^-/- 31.7 x 10⁵, per femur respectively, N=2). This population was positive for CD25, CD11b, and Igκ and partially positive for CD4, Gr-1, and CXCR4 ( Supplemental Figure 4 ). We also confirmed F4/80 expression (94.7%) in Bcl2-Tg mice (data not shown). This population contains a distinctive small fraction of T-cells (CD3+), erythroid (Ter119+), and myeloid progenitor cells (Gr-1+). No significant phenotypic difference was seen in the B220^-IgM⁺ fraction between Bcl-2-Tg, and Bcl2-Tg/Tfl ^-/- bone marrow. Although a B220^-IgM⁺ population was found in both wild type and Bcl2-Tg spleen, it was less prominent (Wild type 8.4%, Tfl ^-/- 3.3%, Bcl-2-Tg 7.9% ± 4.4%, Bcl2-Tg/Tfl ^-/- 11.8% ± 2.3%). Therefore, we thought this unique population could affect lymphomagenesis in Bcl2-Tg mice. We speculated that the function of this unusual cell population could be altered by TFL deficiency, resulting in shorter survival in Bcl2-Tg/Tfl^-/- mice. To elucidate the functional changes in these populations, we performed a cDNA expression array using B220^-IgM⁺ sorted bone marrow cells of both Bcl2-Tg and Bcl2-Tg/Tfl ^-/- mice ( Supplemental Files 1, 2 ). In these cells in Bcl2-Tg/Tfl ^-/- mice, 182 genes were upregulated at least two-fold relative to Bcl2-Tg mice ( Supplemental File 3 ). Among them, we selected Il-21, Cxcl13, Cdk6, Cdk1, and Cxcr7, all cytokine, cell cycle, or cancer-related genes. Their mRNA expression was confirmed by real-time quantitative PCR. Among these upregulated genes, Cxcl13 and Cxcr7 mRNA expression in Bcl2-Tg/Tfl ^-/- mice was significantly higher than Bcl2-Tg mice, although the expressions varied in each Bcl2-Tg/Tfl ^-/- mice ( Figure 4B ). To further confirm Cxcl13 overexpression, we measured Cxcl13 concentration in bone marrow extracellular fluid (BMEF) and plasma samples in both strains at week 40-44. We found a significant increase in Cxcl13 concentration in Bcl2-Tg/Tfl ^-/- mice compared to Bcl2-Tg mice in BMEF ( Figure 4C ). Plasma Cxcl13 levels in Bcl2-Tg/Tfl ^-/- mice were extremely high, with a median of 20 ng/ml ( Figure 4D ). It is noteworthy that the increase in Cxcl13 begins around week 20, coinciding with weight loss in Bcl2-Tg/Tfl ^-/- mice, followed by an early death ( Figures 3C , 4E ).

Next, we asked whether B220^-IgM⁺ cells are the main producers of Cxcl13. We first tested the culture assay of bone marrow and solenocytes of wild-type or Bcl2-Tg mice to address this question. The supernatant showed a significantly increased concentration of Cxcl13 from bone marrow cells and splenocytes in the Bcl2-Tg mice group ( Figure 5A ). Next, we examined Cxcl13 secretion from bone marrow and splenocytes of either Tfl^-/- or Bcl2-Tg/Tfl^-/- mice. In Bcl2-Tg/Tfl^-/- mice, the bone marrow was the main producer of Cxcl13 secretion ( Figure 5B ). Subsequently, we sorted bone marrow cells according to B220 and/or IgM expression (B220⁺, IgM^-, or IgM⁺; Figure 5C ) and cultured each fractionated cells to measure the Cxcl13 secretion. The supernatant of each cell population was analyzed and tended to increase Cxcl13 secretion from B220^-IgM⁺ cells in both Bcl2-Tg and Bcl2-Tg/Tfl^-/- mice with 2 independent experiments (IgM⁺; Figures 5D, E ). Thus, B220^-IgM⁺ cells in the bone marrow are considered to be the main producers of Cxcl13.

Finally, to determine whether TFL regulates CXCL13 mRNA through the degradation of mRNA 3’UTR, we performed a reporter assay with a luciferase vector containing 3’ UTR of CXCL13 mRNA. Co-transfection with the TFL expression vector showed decreased luciferase activity compared to the controls in all B lineage lines studied (Daudi, Nalm6, Molt-13, and Namawala) ( Figure 6A ), suggesting that TFL regulates CXCL13 via 3’UTR mRNA degradation in B lineage cells. These phenomena were not, however, seen in some other cell lines of other lineages, including THP-1, Jurkat, or CCRF-CEM. TFL did regulate IL-2 in all tested cell lines ( Figure 6B ). CXCL13 regulation seems to be restricted mainly in B cell lineage cells, while IL-2 regulation occurs promiscuously.