Stem cell characteristics promote aggressiveness of diffuse large B-cell lymphoma

Diffuse large B-cell lymphoma (DLBCL) may present initially in bone marrow, liver and spleen without any lymphadenopathy (referred to as BLS-type DLBCL), which is aggressive and frequently associated with hemophagocytic syndrome. Its tumorigenesis and molecular mechanisms warrant clarification. By gene microarray profiling with bioinformatics analysis, we found higher expression of the stem cell markers HOXA9 and NANOG, as well as BMP8B, CCR6 and S100A8 in BLS-type than conventional DLBCL. We further validated expression of these markers in a large cohort of DLBCL including BLS-type cases and found that expression of HOXA9 and NANOG correlated with inferior outcome and poor prognostic parameters. Functional studies with gene-overexpressed and gene-silenced DLBCL cell lines showed that expression of NANOG and HOXA9 promoted cell viability and inhibited apoptosis through suppression of G2 arrest in vitro and enhanced tumor formation and hepatosplenic infiltration in a tail-vein-injected mouse model. Additionally, HOXA9-transfected tumor cells showed significantly increased soft-agar clonogenic ability and tumor sphere formation. Interestingly, B cells with higher CCR6 expression revealed a higher chemotactic migration for CCL20. Taken together, our findings support the concept that tumor or precursor cells of BLS-type DLBCL are attracted by chemotaxis and home to the bone marrow, where the microenvironment promotes the expression of stem cell characteristics and aggressiveness of tumor cells.


Materials and methods
DLBCL cases. We enrolled 110 cases of DLBCL including 9 cases of BLS-type and 101 conventional DLBCL-NOS cases from the archives of the National Cheng Kung University Hospital from 1995 to 2007. The diagnosis of DLBCL was based on the World Health Organization (WHO) classification scheme 1 . The BLS-type DLBCL cases have been reported previously 11 . The inclusion criteria for BLS-type DLBCL were cases initially only involving BM with or without involvement of the liver or spleen (termed "BLS" type). No patients presented with lymphadenopathy at diagnosis, and none had a previous diagnosis of lymphoma. Other recognized LBCL variants such as intravascular lymphoma (IVL), T-cell/histiocyte-rich (THR) LBCL were excluded as was localized primary bone lymphoma. These cases were included with full pathogenetic characterization and adequate clinical staging. Imaging studies-computed tomography (CT) of whole body, bone scan, magnetic resonance imaging (MRI), and positron emission tomography (PET)-demonstrated that no other organs were involved 11 . Clinical data, including serum level of lactate dehydrogenase (LDH), Ann Arbor stage, International Prognostic Index (IPI) score, and overall survival (in months) were obtained by chart review. All tumor specimens were fixed in 10% neutral formalin solution and paraffin-embedded.
Most patients were treated with the R-CHOP (rituximab, cyclophosphamide, doxorubicin, vincristine, and prednisone) regimen as a first-line treatment with curative intent according to the NCCN guidelines 12 . All the patients were followed for an interval ranging from 0.1 months to 224.4 months with a mean duration of 46.6 months.
There were three cases of BLS-type DLBCL with frozen stored marrow tissue available. CD20-positive tumor cells were first isolated from cryopreserved bone marrow cells by using magnetic microbeads coated with anti-CD20 antibody (MACS Column, Miltenyi Biotec Inc., Auburn, CA, USA). The purification of the lymphoma cells was more than 90% ( Supplementary Fig. S1). For comparison, two cases of conventional DLBCL with tumor cells in effusions were isolated using the same approach 13 . These studies were approved by the institutional review board (National Cheng Kung University Hospital NCKUH-B-ER-105-394) and were in accord with the Helsinki Declaration of 1975, as revised in 2013. Informed consent was taken from the patients. Immunohistochemical staining. Immunohistochemical analysis was performed on formalin-fixed, deparaffinized tissue sections following heat-induced epitope retrieval 15 . The staining was graded as positive when ≥ 10% of tumor cells were reactive, as described in a previous study 16 . Appropriate tissues were used as positive and negative controls, respectively. The five candidate markers yielded from microarray analysis were used to validate their expression in clinical samples: BMP8B (N-19, 1:25, sc-6900, Santa Cruz Biotechnology, Dallas, TX, USA), NANOG (polyclonal, 1:25, ab21624, Abcam, Cambridge, MA, USA), HOXA9 (polyclonal, 1:20, ab92565, Abcam), S100A8/MRP8 (7C12/4, 1:25, ab20220, Abcam), and CCR6 (polyclonal, 1:75, ab78429, Abcam).

RNA in situ hybridization.
To detect NANOG and HOXA9 mRNA transcripts, the RNAscope assay (Advanced Cell Diagnostics, Hayward, CA, USA) was performed on BLS-type DLBCL cases (n = 9) 17 . Briefly, 5 µm thick tissue sections were pretreated with heat and protease and hybridized with target RNA-specific oligonucleotide probes (Advanced Cell Diagnostics, Hayward, CA, USA). Hybridization with target probes, preamplification, amplification, and chromogenic detection using diaminobenzidine (DAB) were carried out as per manufacturer's instructions. All steps were performed manually in the presence of appropriate controls. Punctate brown signals present in the tumor cells were considered to be positive.

DLBCL cell lines and EBV-transformed lymphoblastoid cell lines (LCLs). Five DLBCL cell lines
were used for in vitro and in vivo experiments: HT, SU-DHL-5, U-2932, U-2940 and HBL-2 (Supplementary Table S1). The sources were DSMZ (Braunschweig, Germany) and ATCC (Manassas, VA, USA). The cells were cultured at 37 °C and 7% CO 2 in RPMI 1640 medium (Gibco/BRL, Grand Island, NY, USA) supplemented with 10% heat-inactivated fetal bovine serum (FBS), 4 mM of glutamine, 75 units/ml of streptomycin, and 100 units/ ml of penicillin. Cell viability was determined using the trypan blue exclusion test or MTT (3-[4,5-dimethylthiazol-2-yl]-2,5-diphenyltetrazolium bromide) assay 18 . Transformed B-lymphoblastoid cell lines (LCLs) immortalized by Epstein-Barr virus (EBV) infection were used for comparisons 18 . Reverse transcriptase-polymerase chain reaction (RT-PCR) and quantitative real-time PCR. RT-PCR analysis for gene expression and the quantitative real-time PCR (qPCR) were performed by an intercalator-based method (Roche Applied Science, Mannheim, Germany), as described previously 15,19 . The glyceraldehyde 3-phosphate dehydrogenase (GAPDH) gene was used as an internal control. Experiments were Scientific Reports | (2020) 10:21342 | https://doi.org/10.1038/s41598-020-78508-7 www.nature.com/scientificreports/ performed in duplicate and the results were analyzed by software (LightCycler Software Version 4.0, Roche Applied Science). Expression was detected using the relevant primers for the five selected genes detected by gene microarrays: h-GAPDH forward: 5′-AGG TCA TCC CTG AGC TGA ACG G-3′, reverse: 5′-CGC CTG CTT  CAC CAC CTT CTT G-3′; h-BMP8B forward: 5′-CTT TCG TGG TCA CTT TCT TC-3′, reverse: 5′-TGG ACG  TCA TCA AAG ATC C-3′; h-CCR6 forward: 5′-GGG AAT CAA TGA ATT TCA GC-3′, reverse: 5′-CAA TCG  GTA CAA ATA GCC TG-3′; h-HOXA9 forward: 5′-ATT GGA GGA AAT GAA TGC TG-3′, reverse: 5′-GAA  ACC CCA GAT TCA TCA AG-3′; h-NANOG forward: 5′-CCA GAA CCA GAG AAT GAA ATC-3′, reverse:  5′-TGG TGG TAG GAA GAG TAA AG-3′; h-S100A8 forward: 5′-GTA TAT CAG GAA AAA GGG TGC-3′,  reverse: 5′-TAC TCT TTG TGG CTT TCT TC-3' . Immunofluorescent staining. Immunofluorescent staining was performed as described in a previous study 18 . Briefly, DLBCL (2 × 10 6 ) cells were cultured in 6-well plates. After cytospinning at 350 rpm for 15 min, cells were transferred onto poly-L-lysine-coated glass slides for immunofluorescence staining. Formaldehyde (100 μl, 4%) was added to fix cells at room temperature for 15 min followed by 0.1% triton for 15 min. After washing with PBS, 3 drops of background-reducing reagent (Dako, S3022, Carpinteria, CA, USA) were added. The cells were incubated with 50-100 μl primary antibodies at 4 °C overnight and then incubated at room temperature in the dark for 1 h with FITC-conjugated anti-rabbit secondary antibody ( . Polyacrylamide gel electrophoresis and immunodetection of five candidate gene products were performed as described previously 15,18 . Transfecting NANOG-or HOXA9-specific shRNA into DLBCL cell lines. Short hairpin RNAs (shRNAs) were designed against the target sequences of the NANOG and HOXA9 genes. The shRNAs were controlled for sequence specificity using a BLAST search and did not show any homology to other known human genes. Plasmids expressing gene-specific shRNA were constructed using synthetic oligonucleotides cloned into the BglII/HindIII cloning sites of the pSUPER-EGFP vector (pSUPER RNAi System; OligoEngine, Seattle, WA, USA). In all RNA interference experiments, a negative control vector containing scrambled shRNA was included (pLKO.1-TRC1 and pLKO-TRC2, Supplementary Fig. S4, S5). The target sequences of the short hairpin RNAs Cell death and cell cycle assays by flow cytometry. Flow cytometric analysis was performed (Becton Dickinson, Mountain View, CA, USA) as described previously 18,20 . Cell viability was determined by the trypan blue exclusion test. Apoptosis and other forms of cell death were evaluated by measuring the DNA content using annexin V and propidium iodide (PI) affinity as previously described 21 . Briefly, each sample of 2.4 × 10 6 cells was transfected with candidate gene-specific shRNA or control vector, and then cultured in 6 ml of medium. Each sample of 1.5 ml was collected after 48-72 h. The sample was then centrifuged, and the pellet was incubated with staining solution (PI [50 μg/ml]; 0.1% sodium citrate; 0.1% triton) overnight at 4 °C in the dark. Core DNA content was measured using a logarithmic amplification in the FL2 (for annexin V) and FL3 (for PI) channels of the flow cytometer (FACSCalibur with CellQuest Pro 4.0.2; Becton Dickinson) 18,20 . Cell-cycle analysis was also measured using flow cytometry. The distribution of the DNA content of individual cells was stained with PI and measured by using a linear amplification in the FL3 channel.

Transfection of DLBCL cell lines with NANOG and HOXA9 genes.
Scientific Reports | (2020) 10:21342 | https://doi.org/10.1038/s41598-020-78508-7 www.nature.com/scientificreports/ Murine xenograft model for functional assay of NANOG and HOXA9 genes. Female NOD/SCID (non-obese diabetes/severe combined immunodeficiency) mice, 6-8 weeks of age, were injected via the tail veins with 1 × 10 7 DLBCL cell lines with differential expression of NANOG or HOXA9 genes. Each cell line was inoculated into a group of mice (n = 4). Tumor volume was measured by calipers every other day, and the formula (width 2 x length × 0.52) was applied to approximate the volume of a spheroid for a maximum of 120 days 22 .
Tumor-bearing mice were sacrificed by CO2 inhalation, and solid tumors were studied by flow cytometry and immunohistochemistry. Human tumor xenografts were confirmed by evaluation of a human DLBCL cell phenotype, CD19, CD20 and a high Ki-67 index. Tumor numbers and sizes were measured, expressed as a mean for each, and correlated with the expression levels of stem cell markers in each cell line. Mice without visible tumor xenografts were sacrificed within 120 days. In these grossly negative mice, necropsy was performed to further investigate the presence of DLBCL cells. For each mouse, liver and spleen were dissected, serially sectioned and made into formalin-fixed, paraffin-embedded tissue blocks to detect microscopically the presence of human DLBCL cells, expressed as percentage of hepatosplenic infiltration. All procedures involving animals were performed in accord with institutional policies and animal ethics and were approved by the Institutional Animal Care and Use Committee (National Cheng Kung University; IACUC approval number: 106091).
Clonogenic (anchorage-independent growth) assay. The clonogenic potential of human lymphoma cells was assessed via the human colony-forming cell assay using methylcellulose-based media (Complete Meth-oCult, Stemcell Technology, Tukwila, WA, USA) as described in a previous study 23 . Briefly, each 2 × 10 3 HT or SU-DHL-5 cells (0.05 ml) transfected with HOXA9 were added to 0.5 ml medium per well in 24-well plates. The cultures were incubated for 10-14 days and then colonies were enumerated manually with an inverted microscope.  Statistical analysis. Appropriate statistical tests were used to examine the relationships and correlations between variables, including χ 2 -test, paired and unpaired t-tests, and Kendall's tau correlation. Overall survival was measured from initial diagnosis to death from any cause, with follow-up data of surviving patients assessed at the last contact date. Estimates of overall survival distribution were calculated using the method of Kaplan and Meier 24 . Time-to-event distribution was compared using the log-rank test 25 . A Cox proportional-hazard model was used to test the simultaneous influence on survival of all covariates found to be significant (p < 0.01) in the univariate analysis 26 . The analyses were carried out using statistical software (SPSS, Inc., Chicago, IL, USA). All cell line experiments were performed at least in triplicate, and the corresponding data were shown. The results were presented as mean ± standard deviation (SD). Statistical significance was defined as p < 0.05 (*p < 0.05; **p < 0.01; ***p < 0.001).

Chemotaxis and cell migration assays.
Ethics approval and consent to participate. The studies were approved by the institutional review board (National Cheng Kung University Hospital NCKUH-B-ER-105-394).

Results
Gene microarray analysis yielded up-regulated genes in BLS-type DLBCL. The microarray study yielded 2,501 genes that were upregulated in BLS type DLBCL as compared with conventional DLBCL (NCBI, Gene Expression Omnibus, accession number GSE136545). After analysis by the DAVID program, 14 important genes with a > fourfold change became a focus for further studies (Supplementary Table S2). For hypothesis testing, we focused particularly on 5 relevant genes involved in stem cell signature (HOXA9, NANOG), chemotaxis (CCR6) and the microenvironment (BMP8B and S100A8/MRP8). Bioinformatics analysis using Ingenuity Pathway Analysis (IPA) showed cross-linked pathways between those genes ( Supplementary Fig. S6). The pathways involved may include cell-to-cell signaling and interaction, cellular migration, development and proliferation ( Supplementary Fig. S7).

Validation of mRNA expression of five selected genes in DLBCL cell lines.
To validate the expression of five selected genes (BMP8B, CCR6, HOXA9, NANOG and S100A8) detected by microarray data, qRT-PCR was performed on five DLBCL cell lines and the lymphoblastoid cell line, LCL. All 5 DLBCL cell lines showed enhanced mRNA expression of HOXA9 and NANOG, but lower expression of CCR6 and S100A8 compared with LCL ( Fig. 2A). The mRNA expression levels of BMP8B were variable among these cell lines.
Differential protein expression of five selected genes in DLBCL cell lines and LCL. We also used these DLBCL and the LCL cell lines to test for protein expression of the 5 candidate genes. The relative protein expression levels are shown in Fig. 2B. The U2940 and U2932 cell lines had higher expression levels of HOXA9 and NANOG than the HBL-2, HT and SU-DHL-5 cell lines. The expression pattern of mRNA and protein was www.nature.com/scientificreports/ discordant for HOXA9, NANOG and S100A8 in these cell lines, a known common event 27 . We therefore performed subsequent functional studies based on the protein expression levels and focused on the role of the stem cell proteins, HOXA9 and NANOG in DLBCL. www.nature.com/scientificreports/ Expression of stem cell markers NANOG and HOXA9 affected DLBCL G2/M cell cycle arrest and apoptosis. To test the function of NANOG and HOXA9, we transfected both genes into HT and SU-DHL-5 cells and knocked down both genes in U2932 and U2940 cells. We did these experiments to observe the effects on cell apoptosis and cell cycle transition, because the HT and SU-DHL-5 cells had lower expression whereas the U2932 and U2940 cells had higher expression of both proteins (Fig. 2B). As shown in Fig. 3A-F and Supplementary Fig. S10, we found that transfection of HOXA9 and NANOG decreased cell apoptosis and decreased G2/M phase cell cycle arrest in DLBCL cells. Conversely, attenuated expression of NANOG and HOXA9 in U2932 and U2940 cells enhanced G2/M cell cycle arrest and increased cell apoptosis (Fig. 3G-L). These results support the interpretation that expression of the stem cell markers NANOG and HOXA9 promotes survival of DLBCL cells.
DLBCL with higher expression of stem cell genes formed more and larger tumor nodules as well as more frequent hepatosplenic infiltration in xenograft mice. NOD/SCID mice were used to test the lymphomagenetic ability of B-cell lines that express HOXA9 and NANOG. To mimic lymphoma cell infiltration, mice were inoculated with cell lines via tail vein injection and observed for the number and size of tumor nodules, as well as foci of hepatosplenic infiltration. As shown in Fig. 4 and Supplementary Table S3, LCL and U2940 cells formed a larger number of tumor nodules (n = 4/mouse and n = 3.75/mouse, Fig. 4A, B and E) and showed more frequent hepatosplenic infiltration (100% and 88%, Fig. 4C, D and G) in xenograft mice. The mean tumor size showed borderline significance ( Fig. 4A and F).
Expression of stem cell protein HOXA9 enhanced colony formation and tumor sphere formation. We tested the survival effect of the stem cell marker HOXA9 in DLBCL cells by in vitro clonogenic assay.
In contrast with control MYC-transfected parental HT and SU-DHL-5 cells, HOXA9 transfection of these cell lines showed significantly enhanced clonogenic ability (SU, p = 0.0046; HT, p = 0.0272, Fig. 5). This result suggests that expression of HOXA9 promotes DLBCL survival and proliferation. We further used the tumor sphere formation assay to test the stemness function. In comparison with MYC control, SU cells with stable expression of HOXA9 showed a trend for more sphere formation (p = 0.098, t-test, Supplementary Fig. S11). This assay further provides proof that DLBCL cells with overexpression of HOXA9 bears stemness potential.

B cells with higher CCR6 expression showed higher chemotaxis ability. To test whether CCR6
expression plays a role in the chemotactic attraction of activated B cells, a Boyden chamber assay was used to measure chemotactic ability. As shown in Fig. 6, under the effect of CCL20, a cognate ligand for CCR6, both LCL and DLBCL cells with higher expression levels of CCR6 (Fig. 6A) showed higher chemotactic migration ability than cell lines with lower CCR6 expression (Fig. 6B-H).

Discussion
BLS-type DLBCL, including cases of primary bone marrow DLBCL [28][29][30][31] , is a distinct type of DLBCL with very aggressive behavior. In this study, we found that expression of the stem cell markers HOXA9 and NANOG in DLBCL correlates with poorer prognosis. Notably, these stem cell markers have not been found in lymph nodebased DLBCL 4,5,32 , indicating a novel aspect of BLS-type DLBCL tumorigenesis, which may be associated with the specific bone marrow niche. The stem cell-like feature or "stemness" of BLS-type DLBCL may account for this clinically aggressive behavior. NANOG, a homeodomain transcription factor, is critical for the propagation of human embryonic stem cells and induction of pluripotency 33,34 . NANOG is upregulated in DLBCL cells that survive treatment with doxorubicin and phenylbutyrate, through overexpression of FOXO4 35 . Phenylbutyrate, a histone deacetylase inhibitor, induces stemness in human pluripotent stem cells 36 . Development of stemness in DLBCL is thus proposed as being responsible for resistance to both drugs 35 . Accordingly, together with our findings, it appears that BLS-type DLBCL acquire a stem cell phenotype through overexpression of NANOG and HOXA9. As a transcription factor, the usual subcellular localization of active NANOG should be in the nucleus. However, growing evidence shows that cytoplasmic expression of NANOG occurs in a subset of tumors and is usually associated with a poorer prognosis in patients with various types of cancers [37][38][39][40][41] . The detailed mechanisms of cytoplasmic localization of NANOG remain under investigation. In 2009, NANOG protein was shown to have a nuclear export signal, suggesting nuclear-cytoplasmic shuttling of NANOG 42 . Recently, it was shown that SRSF3 binds to NANOG mRNA to facilitate its nuclear-cytoplasmic export independent of splicing. In the absence of SRSF3 binding, NANOG mRNA is sequestered in the nucleus resulting in severe downregulation of protein levels 43 . Most importantly, the function of cytoplasmic NANOG is not compromised 34 . Thus, NANOG is not expressed exclusively in undifferentiated cells and both nuclear and cytoplasmic NANOG can function as transcription factors in a cell type-specific manner 37 .
Homeobox gene HOXA9, an evolutionarily conserved transcription factor, plays an important role in hematopoiesis, leukemogenesis and lymphopoiesis at a very early stage 44 . Acting as an oncoprotein, HOXA9 overexpression is associated with an adverse prognosis in adults with acute leukemia 45 . HOXA9 is also important in promoting proliferation and the infiltrative abilities of MLL-rearranged acute leukemia 46 . The finding of HOXA9 expression in BLS-type DLBCL has not been reported previously. Interestingly, HoxA9 activation in precursor B cells increases proliferation independent of stromal cell support; activation of HoxA9 enhances expression of c-Myb and IGF-1R 47 . In parallel, we have previously found that nuclear expression of c-MYB in about 10% of DLBCL cases correlated with a poorer prognosis and unfavorable clinical factors 18 . Because c-Myb is important in the pro-B to pre-B transition 48 , expression of c-MYB in DLBCL also may be a phenomenon related to a stem cell feature. This link may suggest that overexpression of HOXA9 or c-MYB is present in a subset of DLBCL, predicting aggressive behavior probably through the expression of a stemness phenotype by lymphoma cells. Notably, www.nature.com/scientificreports/ our functional assays showed that overexpression of NANOG and HOXA9 significantly promoted survival of DLBCL cells, in vivo tumorigenesis, and anchorage-independent clonogenic ability. Together with the association of NANOG overexpression with chemotherapeutic resistance 35 , our results indicate that overexpression of HOXA9 and NANOG leads to colony formation in addition to growth promotion and tumor cell infiltration. Two hypotheses have been raised to address the possible mechanisms of lymphomagenesis: one is the stem cell (lymphoma-initiating cells) theory; the other is normal counterpart derivation theory. The stem cell theory denotes that all the lymphomas are of stem cell origin and then differentiate into different stages likened to normal lymphocytic development 49 . This theory may underlie the development of composite lymphomas, in which two histologically distinct lymphomas occur within the same organ and share the same immunoglobulin gene rearrangement 50 . Also reflected is the report that a patient with B-cell lymphoma subsequently developed acute myeloid leukemia (AML); both the lymphoma and the AML harbored the same biallelic TET2 mutations 51 . The normal counterpart theory suggests that each lymphoma originates within a lymphocyte at the differentiation stage that corresponds most closely to the lymphoma phenotype 52 . The detection of t(14;18) circulating cells in healthy individual 53 , and the belief of multistep lymphomagenesis both argue for this hypothesis. Because most BLS-type DLBCL cases show an activated B-cell immunophenotype, we suggest that the tumor progenitor cells are activated B cells that colonize bone marrow before neoplastic transformation, a model akin to the normal counterpart theory. Importantly, bone marrow homing and the expression of stem cell markers could represent novel, potential therapeutic targets 54 .
The tumor microenvironment also plays a role in tumor growth and progression 55 . Bone morphogenetic protein (BMP) signaling is a key pathway controlling stem cells and their niche 56 . Expression levels of BMP8B are significantly increased in the bone marrow of gastric cancer patients with metastatic disease, consistent with a role of a secreted factor in cancer progression 57 . Interestingly, BMP8B is highly expressed in human bone marrow, liver and spleen 58 . Together with our finding of BMP8B expression in a subset of DLBCL tumors, it suggests that BMP8B signaling promotes DLBCL progression in a paracrine or autocrine way. Likewise, S100A8, a calcium-binding protein, is involved in inflammatory processes 59 . An inflammatory microenvironment, which mainly includes S100A8/9 (calprotectin), promotes cancer metastasis 59 . However, the role of S100A8 in the bone marrow microenvironment is largely unknown. Zambetti et al. found that bone marrow mesenchymal nicheinduced genotoxic stress in hematopoietic stem cells causes leukemic evolution through p53-S100A8/9-TLR4 inflammatory signaling 60 . Furthermore, by upregulating S100A8/A9 expression in tumor cells, tumor-infiltrating macrophages can promote tumor invasion and migration 61 . Consistent with these observations, our finding of S100A8 expression in BLS-type DLBCL suggests that the bone marrow niche may promote DLBCL growth and progression, probably through lymphoma-microenvironment interplay mediated by BMP8B and S100A8/9.
In this study, we identified CCR6 expression in ~ 20% of DLBCL cases and this finding was associated with poorer patient outcome. Interestingly, stromal cells isolated from bone marrow can secrete CCL20, the cognate ligand of CCR6 upon serum-free culture 62 . CCL20 is known to be highly upregulated by hypoxia in bone marrowderived cells 63 . These data suggest that the CCR6-CCL20 axis plays a pivotal role in the homing of BLS-type   11 , the cause of this preferential homing to the bone marrow, liver and spleen depends likely on the expression of adhesion molecules that preferentially bind to resident bone marrow mesenchymal stromal cells. On the other hand, overexpression of CCR6 in  www.nature.com/scientificreports/ lymphoma cells might possibly drive tumor growth through CCL20-mediated recruitment of tumor-associated macrophages 64,65 . Finally, we also show the high tumorigenic ability of the EBV-immortalized LCL cell line, as reported previously 66 , in which CD226 also may play a role in addition to EBV-encoded oncoproteins 67,68 . The limitation of this study is a relative small number of DLBCL cases for gene microarray analyses, which is due to the uncommon occurrence of BLS-type DLBCL. However, we have validated the findings of stem cell characteristics by quantitative real-time PCR, RNA in situ hybridization, protein expression in clinical samples, and both in vitro and in vivo studies. Furthermore, our additional cohort (n = 79) with nine cases of BLS-type DLBCL also showed that DLBCL cases with HOXA9 overexpression carried a poorer prognosis (p = 0.028, Log Rank test, Supplementary Fig. S12). We hope our findings can lead to the therapeutic development of BLS-type www.nature.com/scientificreports/ DLBCL and warrant further studies. Regarding the therapeutic agents for the rare entity, we have previously found that rigosertib (ON 01910.Na) can inhibit growth of DLBCL by cytoplasmic sequestration of sumoylated C-MYB/TRAF6 complex and specific knockdown of c-Myb and TRAF6 induced tumor cell apoptosis and cell cycle arrest 18 . Based on this finding, rigosertib (ON 01910.Na) or inhibitors of c-Myb and TRAF6 might be a candidate agent beneficial for patients with BLS-type DLBCL.
In conclusion, our findings suggest that tumor precursor cells of BLS-type DLBCL are attracted by chemotaxis and home to the bone marrow from the periphery at the activated B-cell stage. When these precursor cells arrive in the bone marrow, these cells acquire genetic changes and express stem cell signatures under the effects of the bone marrow microenvironment, thereby promoting aggressiveness of DLBCL. Our findings shed some light on the study of lymphomagenesis, may be useful in patient stratification for therapy, and may provide some clues for developing novel therapeutic strategies.