Access to Follicular Dendritic Cells Is a Pivotal Step in Murine Chronic Lymphocytic Leukemia B-cell Activation and Proliferation

In patients with B-cell chronic lymphocytic leukemia (B-CLL), tumor cell survival and progression are linked to interactions between leukemia cells and nontumor cells in microenvironments of the spleen, peripheral blood, and bone marrow (1–3). Little is known about how CLL cells get access to a putative proliferation and survival niche within lymphoid organs, and a spatial and kinetic resolution of this process is desirable for two reasons. First, the microanatomy of B-cell activation might illuminate the cellular encounters leading to apoptosis resistance versus proliferation induction in CLL cells. Second, signaling events could alter trafficking behavior of leukemia B cells, which is likely relevant for B-cell stimulation.

B-CLL is the most frequent leukemia subtype affecting adults in Western countries (4). It develops as an expansion of clonal mature CD5⁺ B cells (5, 6), and gene expression profiles of CLL cells show that these malignant B cells resemble antigen-experienced memory B cells (7, 8). The precise cell-of-origin is debated. Depending on differentiation markers, or a putative antigen experience in conjunction with B-cell receptor (BCR) selection, either a marginal zone (MZ) B-2 cell, a B-1 lymphocyte, or a hematopoietic stem cell origin have been proposed (9, 10).

Several genomic aberrations involving karyotypic alterations or mutations in TP53, NOTCH1, and other markers are frequently reported in CLL cases (3, 11). High T-cell leukemia/lymphoma 1 (TCL1) oncogene expression is linked to an aggressive B-CLL phenotype, with leukemic cells displaying a nonmutated immunoglobulin (Ig) receptor and profound ZAP70 expression (3, 12). A mouse B-CLL model has been developed by introducing a transgene encoding human Tcl1 under the control of the immunoglobulin heavy chain Eμ-enhancer, resulting in Tcl1 overexpression at the mostly mature stages of B-cell development (13). Tcl1 transgenic mice develop B-CLL–like tumors characterized by an expansion of B220^loIgM⁺CD5⁺CD11b⁺ cells, initially localized in the peritoneum and, subsequently, in the spleen, bone marrow, and peripheral blood. This B-CLL model has proven suitable to phenocopy several aspects of the human disease, including low proliferative progression, tissue distribution, and B-cell differentiation status.

As in the migration of their normal cellular counterparts, trafficking and homing of CLL cells to and within the tissue microenvironment are regulated by adhesion molecules in concert with the chemokine–chemokine receptor system (14). The homeostatic chemokine receptor CXCR4 has emerged as a key receptor in the recruitment of and cross-talk between malignant B cells and their protective bone marrow microenvironment (15). Expression of secondary lymphoid organ (SLO)–homing receptors, including CCR7 and CXCR5, on human B-cell non-Hodgkin and Hodgkin lymphoma correlates with their dissemination and tissue localization to and within specific SLO compartments (16, 17).

In B-CLL pathogenesis, a predominant role of BCR signaling in CLL ontogeny has emerged (1). Furthermore, it remains still elusive how the microenvironment assumes a central role in B-cell activation and survival. Here, we aimed to link the mechanisms of leukemia cell activation and the microenvironment dependence during CLL pathogenesis in Eμ-Tcl1 mice. We intravitally visualized the CXCR5-mediated guidance of leukemic B cells toward splenic B-cell follicles via a route normally taken by MZ B cells. In germinal center (GC) light zones, tumor B cells productively interacted with follicular dendritic cells (FDC), which accelerated proliferation and clinical progression. Leukemia cells initiated LTαβ-dependent reciprocal cross-talk with resident follicular stroma cells, leading to CXCL13 release. Selective targeting of stroma-associated CXCL13 and LTβR signaling abrogated this paracrine feedback loop and retarded leukemia growth.

To assess a putative role of CXCR5-governed leukemia cell migration and expansion in a CLL model, we crossed Eμ-Tcl1 transgenic and Cxcr5^−/− mice and followed their spontaneous tumor development compared with Eμ-Tcl1 mice (Fig. 1A). We detected leukemia B cells (CD19⁺CD5⁺; Supplementary Fig. S1A–S1C and Fig. 1A) in the spleen and peripheral blood of Eμ-Tcl1 mice after 2 months. In Cxcr5^−/−Eμ-Tcl1 mice, disease onset was substantially delayed. At 4 to 6 months of age, tumor cells in Eμ-Tcl1 mice accounted for 30% to 50% of all lymphocytes, whereas tumor cells were barely detectable in Cxcr5^−/−Eμ-Tcl1 mice. After >6 months, Eμ-Tcl1 mice showed a further increased tumor load (>50%) in spleen and peripheral blood, whereas that in Cxcr5^−/−Eμ-Tcl1 mice was still much lower. At 5 to 6 months of age, we also detected more leukemia B cells in the bone marrow of Eμ-Tcl1 compared with Cxcr5^−/−Eμ-Tcl1 mice (Fig. 1A). Eμ-Tcl1 leukemia cells in the spleen (Fig. 1B) expressed high levels of functional CCR7, CXCR4, and CXCR5 chemokine receptors, as shown in vitro by their migration in response to the respective chemokines CCL21, CXCL12, and CXCL13 (Fig. 1C). High expression levels of the homeostatic chemokine receptors were not restricted to the spleen, because lymph node (LN)–, peripheral blood–, and bone marrow–derived leukemia cells also abundantly expressed CCR7, CXCR4, and CXCR5 (Supplementary Fig. S1D). Integrins β1, β2, β7, α4, and αLβ2 (LFA-1) were readily detected on all Eμ-Tcl1 leukemia cells, and quantitative real-time PCR (qRT-PCR) revealed substantial expression for LTα, LTβ, and TNFα in Eμ-Tcl1 and Cxcr5^−/−Eμ-Tcl1 cells (Supplementary Fig. S2A and S2B). Hence, Eμ-Tcl1 leukemia B cells are equipped with functional homeostatic chemokine receptors and integrins that enable them for access to SLOs.

We confirmed abundant high CXCR5 expression on human CLL tissue (Supplementary Fig. S2C) and on human peripheral CLL blood samples (Supplementary Fig. S2D). Much weaker expression was observed on CLL cells localized within proliferation centers, which are thought to harbor a proliferative fraction of CLL leukemia B cells (18).

Next, malignant B cells derived from diseased Eμ-Tcl1 or Cxcr5^−/−Eμ-Tcl1 mice were adoptively transferred into wild-type (WT) congenic recipients. After 72 hours, leukemia cells were detected in the spleen, peripheral blood, and bone marrow. Eμ-Tcl1 cells were predominantly localized within splenic B-cell follicles (Fig. 1D). In contrast, Cxcr5^−/−Eμ-Tcl1 cells were completely absent from B-cell follicles and were mostly found in the MZ outside the metallophilic macrophage ring (MOMA-1⁺). To further assess whether this follicular localization had any consequences on leukemia progression, we applied a CXCL13 antibody to inhibit CXCR5 signaling. Expansion of adoptively transferred tumor B cells was substantially diminished in spleens of CXCL13-treated mice (Fig. 1E).

Collectively, these data indicate that CXCR5 contributes to accelerated leukemia development, potentially by conferring a survival advantage to Eμ-Tcl1 tumor cells within the B-cell follicle.

The slower kinetics of tumor growth in Cxcr5^−/−Eμ-Tcl1 transgenic mice could relate to a reduced proliferation rate. Using genome-wide expression arrays, we examined relative changes in gene expression of splenic leukemia cells derived from diseased Eμ-Tcl1 versus Cxcr5^−/−Eμ-Tcl1 mice. Gene set enrichment analysis (GSEA) identified seven different proliferation signatures that discriminated Cxcr5^+/+ from Cxcr5^−/− leukemias (Fig. 2A and B; Supplementary Fig. S3A). For example, the gene-expression signature of the neighborhood of the cell division cycle 2 (CDC2; G₁–S and G₂–M) in the GNF2 expression compendium (19) predicted a proliferative advantage in CXCR5-expressing Eμ-Tcl1 tumor cells and included genes involved in cell-cycle regulation, DNA synthesis, and protein translation. Conversely, CXCR5^−/− versus CXCR5^+/+ leukemia did not differentially express gene sets of apoptosis-mediating pathways like the Kyoto Encyclopedia of Genes and Genomes (KEGG) apoptosis signature (Fig. 2C; ref. 20).

Eμ-Tcl1 leukemia B cells represent the unmutated and aggressive type of B-CLL (21). A bias toward the restricted use of V_H gene segments was noted, indicating that the expanded CD5⁺ B-cell population in Eμ-Tcl1 mice could be derived from B-1 lineage cells. We performed a transcriptome analysis of leukemia cells, murine B-1 B cells, and MZ B cells (Fig. 2D and E). Two gene expression signatures comprised of the top 100 downregulated and upregulated genes in B-1 cells compared with MZ B cells were defined. When gene expression of these signatures was compared in leukemia cells and in follicle cells derived from WT mice, a strikingly higher similarity of Eμ-Tcl1 leukemia cells to B-1 B cells as compared with MZ B cells was revealed.

Next, we adoptively transferred leukemia cells to synchronize pathogenic events at the onset of the disease (22). Leukemia cells were retrieved and we used qRT-PCR to assess their gene expression pattern, focusing on genes reportedly involved in CLL survival (23). WNT6 and WNT10A were modestly elevated in Eμ-Tcl1 tumor cells, whereas Cxcr5^−/−Eμ-Tcl1 leukemia cells showed substantially higher gene expression for the transcription factor RUNX2 (Supplementary Fig. S3B). The cell-cycle inhibitor p21, a transcriptional target of RUNX2 (24), was also upregulated in Cxcr5^−/−Eμ-Tcl1 leukemia (Supplementary Fig. S3C), indicating that CXCR5 confers Eμ-Tcl1 tumor cells with a proliferative advantage.

When leukemia cells were cocultured with the FDC/HK cell line, the activities of serine/threonine kinases p42/p44, p38, and AKT were indiscriminable in both genotypes (Supplementary Fig. S3D). In addition, a comparable proliferation rate of WT and CXCR5-deficient tumor cells was detectable when leukemia cells were cocultured with the stroma cell line M2-10B4 (Fig. 2F). This indicates that when tumor cells get access to stromal cell support in vitro, their proliferation is supported. In vivo, CXCR5 controls this access to a tumor growth-promoting stromal niche.

Lymphocyte transition from the blood into the spleen occurs at the MZ. Follicular B and T cells then migrate along the fibroblastic reticular cell (FRC)–rich MZ bridging channels (MZBC) toward the splenic white pulp (WP) and eventually enter the periarteriolar lymphoid sheaths (PALS). This movement is largely CCR7-dependent (25). In contrast, MZ B cells undergo CXCR5- and S1P1-controlled constitutive shuttling between the MZ and the follicle (26).

It is unknown how CLL cells access the B-cell follicle. Applying a sequential static imaging approach, SNARF-1–labeled tumor or B cells were adoptively transferred into B6 recipients. One to five hours later, spleens were stained with anti-B220 to locate B-cell follicles and anti–MOMA-1 to highlight the border between WP and MZ (Fig. 3A and B). We counted the numbers of labeled Eμ-Tcl1 leukemia cells or normal B cells in four regions of the spleen: the red pulp (RP; MOMA-1⁻B220⁻), MZ (MOMA-1⁺B220⁺), B-cell follicle (MOMA-1⁻B220⁺), and T/B zone interface (Fig. 3A and B). One hour after transfer, normal B cells were found predominantly in the RP but also in the MZ (Fig. 3B and C), whereas >50% of tumor cells were localized either within the MZ or in B-cell follicles (Fig. 3A and C). Two hours after transfer, B cells were still predominantly located in the RP, with a smaller proportion in the MZ, and single B cells appeared at the T/B–cell border. Tumor cells were equally distributed between MZ and B-cell follicle with a smaller proportion in the RP. Three to five hours after transfer, the proportion of leukemia cells in B-cell follicles further increased, whereas tumor cells remained absent from the T/B–cell border (Fig. 3A–D). In contrast, B lymphocytes transiently accumulated at the T/B zone border (between 2 and 3 hours) before entering the B-cell follicle (3–5 hours; Fig. 3B–D).

The positioning of B cells at the T/B–cell border is regulated by CCR7 and EBI2 chemotactic activity (27). After adoptive transfer, both Ccr7^−/− and Ccr7^+/+ leukemia cells showed the same direct migratory route from the MZ into B-cell follicles, with an accumulation at the FDCs (Supplementary Fig. S4A). Likewise, Ebi2^−/−Eμ-Tcl1 cells exhibited the same localization behavior (Supplementary Fig. S4B and S4C). Clinically, spontaneous leukemia development in Ebi2^−/−Eμ-Tcl1 double-transgenic animals compared with Eμ-Tcl1 mice revealed no alteration in tumor growth (Supplementary Fig. S4D).

In summary, Eμ-Tcl1 leukemia B cells directly cross the MZ sinus, reaching the B-cell follicle faster than follicular B cells, and this process is tightly regulated by the CXCL13–CXCR5 signaling axis.

FDCs reside within primary follicles and in the GC light zone of secondary B-cell follicles. They express CXCL13, which enables CXCR5-expressing B cells to migrate into lymphoid follicles. Because a strong correlation has been proposed between antigen-stimulated BCR signaling and the clinical course of CLL, we identified follicular stromal networks that locally interact with Eμ-Tcl1 leukemia cells. Eight hours after transferring leukemia cells, tumor cells tightly intermingled with FDCs (Fig. 4A). These cells stained strongly for the proliferation marker Ki67 (CD45.2⁺Ki67⁺), indicating a proliferation-driving interaction. Of all Ki67⁺ leukemia cells, 77.5% ± 5.8% cells were tightly associated with the FDC networks, whereas only 22.1% ± 3.6% of follicular Ki67⁺ leukemia cells were located outside of the FDC networks (Fig. 4B).

Next, we simultaneously injected differentially labeled follicular B cells (B220⁺CD21^intCD23^hi) with Eμ-Tcl1 leukemia cells and quantified the proportion of transferred cells within the FDC-rich zone (Fig. 4C). Less than 50% of all transferred follicular B cells were found at the FDC networks. In contrast, more than 70% of leukemia cells strongly colocalized with FDCs (Fig. 4D).

Consistent with the animal model, FDCs could be detected in human CLL specimens at variable rates and shapes, ranging from distinct reticular forms to more complex networks (Supplementary Fig. S5).

Leukemic cells of patients with progressive CLL express the protein tyrosine kinase ZAP-70, which is functionally associated with increased BCR signaling (28). Another tyrosine kinase that is uniformly overexpressed and constitutively active in CLL is Bruton's tyrosine kinase (BTK; refs. 29–31). Here, using an antibody that cross-reacts with phosphorylated (p) ZAP-70 and p-Syk (32), we showed substantially enhanced expression of p–ZAP-70/Syk in Eμ-Tcl1 leukemia cells compared with B lymphocytes. This finding was corroborated with a p-BTK antibody (Fig. 4E and F), indicating that Eμ-Tcl1 leukemia cells exhibit increased BCR activity.

BCR engagement and growth factor supply could depend on the environmental context. Here, tissue-resident leukemia cells from the spleen showed an enhanced proportion of cells in S-phase (BrdUrd⁺), suggesting a stronger stimulus in SLOs compared with cells derived from peripheral blood (Fig. 4G). Thus, Eμ-Tcl1 leukemia cell contacts to FDC networks could be a prerequisite for tumor cell proliferation that is enhanced by BCR signaling and local cytokine provision.

To further dissect these mechanisms, we cocultured seven different Eμ-Tcl1 leukemia cell clones in the presence of FDC/HK cells (33). When HK cells were prestimulated with LTα1β2, an almost 2-fold higher survival rate of leukemia cells as compared with HK cells alone was seen. This indicated that engagement of the LTβR could induce provision of B-cell growth factors (Fig. 4H). Next, we mimicked cytokine and growth factor conditions that were suggested to support either GC B-cell proliferation or follicular B-cell lymphoma expansion (34, 35). When supplemented, a substantial increase in cell proliferation was obtained for BAFF, CXCL12, Sonic hedgehog (SHH), hepatocyte growth factor (HGF), CXCL13, and IL15 (Fig. 4H). The use of the γ-secretase inhibitor DAPT, and thus NOTCH pathway blockade, served as a negative control. To discriminate between an antiapoptotic effect and a proliferation induction, we repeated this assay for selected cytokines and measured instead BrdUrd uptake. We observed enhanced proliferation rates of Eμ-Tcl1 leukemia cells (BrdUrd⁺) after stimulation with LTα1β2 and an even stronger proliferation in the presence of IL15 (Fig. 4I). Treatment of cocultures with the CXCR4 antagonist AMD3100 inhibited BrdUrd uptake (Fig. 4I), which supports the notion that the CXCR4 signaling directly stimulates proliferation of CLL cells. Inclusion of the Gαi/o-protein inhibitor pertussis toxin (PTX) considerably diminished leukemia B-cell viability in HK cocultures (Supplementary Fig. S6A). Hence, BrdUrd uptake could not be reliably determined because of the predominant fraction of apoptotic leukemia cells.

In accordance, when human CLL cells were cocultured with FDC/HK cells, enhanced proliferation of CLL cells was also observed and could be further enhanced by adding IL15 (Supplementary Fig. S6B). Prestimulation with LTα1β2 had a modest effect, which might be explained by higher endogenous expression of LTα and LTβ in human CLL cells (Fig. 7E).

If antigen recognition is important for Eμ-Tcl1 leukemia cell growth, then restricting the antigen specificity of leukemia cells against an irrelevant antigen might reduce their fitness in vivo. We generated Eμ-Tcl1 × BCR^HEL double-transgenic mice and analyzed them after repetitive immunization with the cognate antigen HEL or without (Supplementary Tables S1 and S2) for the development of a clonal B-CLL–like disease. Leukemia cell development equipped with the cognate BCR^HEL could not be observed in nonimmunized or HEL-immunized mice, indicating that sole stimulation of the BCR^HEL does not confer sufficient signaling to promote Eμ-Tcl1 leukemia development.

The localization of Eμ-Tcl1 tumor cells near FDCs suggested that this network might provide the combined chemoattractant, nutritive, and antigenic stimuli for BCR engagement. To visualize the dynamic behavior of Eμ-Tcl1 leukemia cells in relation to FDCs, we injected Alexa Fluor 568–conjugated anti-CD21/CD35 Fab fragments for intravital FDC staining (Supplementary Fig. S7), together with CFDA-labeled Eμ-Tcl1 leukemia or B cells. During a 30-minute imaging period, we tracked individual motile Eμ-Tcl1 leukemia cells or B cells that were localized within the B-cell follicle or close to the FDCs for as long as the cells remained within the z stacks (height, 40–50 μm) of optical sections (Supplementary Movies S1 and S2). Tracks of motile leukemia or B cells were also projected onto an image of the entire z stack, representing a midpoint in the imaging time period (Fig. 4J). We could distinguish between a more stationary FDC-associated and a more motile follicular-located leukemia cell population based on total displacement from their origin (Fig. 4K), track velocity (Fig. 4L), and total track length (Fig. 4M).

All three parameters were substantially decreased when leukemia cells were close to the FDC networks, as opposed to those located within the outer B-cell follicle (Fig. 4K–M; Fig. 4J, top; Supplementary Movie S1). This association was also obtained in follicular B cells (Fig. 4K–M and Supplementary Movie S2), but fewer B cells were associated with FDCs (Fig. 4D and J, bottom). We conclude that the decreased displacement rate and increased interaction time of leukemia B cells with FDCs could result from the tumor cells' cognate nature and their ability to encounter antigen-laden FDCs.

To assess follicular Eμ-Tcl1 leukemia positioning following the organization of GCs into dark and light zones, we immunized B6 mice with sheep red blood cells. GC B-cell positioning in the dark zone is crucially dependent on high CXCR4 expression (36). SNARF-1–labeled tumor cells were i.v. transferred; 8 hours later, splenic sections were stained for PNA⁺ GCs (centroblasts; dark zone), CD21⁺CD35⁺ FDC networks (light zone), and IgD⁺ B-cell follicles (Fig. 5A). Tumor cells were also pretreated with the CXCR4 antagonist AMD3100 (Fig. 5B–D). Eμ-Tcl1 leukemia cells localized exclusively in the CD21⁺CD35⁺ GC light zone, independently of CXCR4-signaling inhibition (Fig. 5B). Inhibition of CXCR4 signaling after AMD3100 treatment was confirmed in vitro in a chemotaxis assay (Fig. 5D).

The sphingosine-1-phosphate receptor S1PR2 helps to confine B cells to the GC (37). GC B cells showed the highest gene expression of S1pr2. In contrast, Eμ-Tcl1 leukemia cells exhibited low S1pr2 expression comparable to the amounts of S1pr2 in MZ B cells (Fig. 5E). Collectively, Eμ-Tcl1 leukemia cells are predominantly recruited to the CD21⁺CD35⁺ GC light zone, but they do not access the PNA⁺ dark zone.

Clinically, an unresolved problem in the treatment of indolent lymphoma is its propensitiy to relapse and to progressively develop chemoresistance. The close association of Eμ-Tcl1 leukemia cells with FDCs raised the question of whether tumor cells might reciprocally support or initiate stromal cell network differentiation. To analyze the mechanisms of a putatively stroma-regulated treatment failure, we first assessed the stroma-inducing capacity of tumor cells in Rag2^−/− mice that are devoid of mature lymphocytes and lack FRCs and FDCs. These mice still have MAdCAM-1⁺ marginal reticular cells (MRC), which produce CXCL13 and express the LTβR. Over a 21-day observation period, Eμ-Tcl1 leukemia cells induced formation of follicular-like CXCL13-expressing FDC (CD21⁺CD35⁺) networks in Rag2^−/− mice (Supplementary Fig. S8A). Leukemia-loaded spleen also upregulated CXCL13 und LTβ gene expression (Supplementary Fig. S8B).

Radiotherapy eliminates hematopoietic cells, whereas mesenchymal stromal elements survive. To assess whether residual leukemia cells could exploit stromal contacts and engage with FDCs and MRCs, we transferred Eμ-Tcl1 cells into irradiated congenic recipients. Although essentially all benign B220⁺ B cells were depleted after 48 hours, FDCs persisted and the MRC ring expanded (Fig. 6A, top left). Transferred CLL cells accumulated around the FDC networks in an even stronger manner than in untreated mice (Fig. 6A, bottom). Functionally, transfer into irradiated hosts conferred tumor cells with an enhanced proliferative capacity (Fig. 6B and C), and the rate of tumor cells in close proximity to FDC networks that stained strongly for Ki67 was substantially higher compared with untreated controls (Fig. 6B). Irradiation of mice substantially enhanced splenic Baff gene expression about 4-fold, whereas APRIL was weaker upregulated (Fig. 6D). We confirmed expression of the corresponding cytokine receptor BAFFR on leukemia cells (Fig. 6E).

To demonstrate that irradiation enhances proliferation in a BAFF-dependent manner, we transferred tumor cells together with a blocking anti-BAFF antibody into irradiated mice. Treatment with the anti-BAFF antibody completely abrogated irradiation-induced proliferation in Eμ-Tcl1 tumor cells (Fig. 6F).

Taken together, CLL cells are recruited to mesenchymal stromal cells, which in turn promote chemokine-mediated tumor cell attraction, growth acceleration through BCR stimulation, and paracrine provision of cytokines.

FDC stimulation with cell-bound LTαβ or soluble LTα/TNFR1 agonist can induce the expression of CXCL13 and other cytokines (38). Because the leukemia cell itself was also a source of lymphotoxin (Supplementary Fig. S2B), we hypothesized that leukemia cell–associated LTαβ conferred the ligand for LTβR activation on FDCs and MRCs. To assess this regulatory feedback loop, we inhibited LTαβ–LTβR signaling in Eμ-Tcl1 mice using an LTβR–Ig fusion protein. Spontaneously diseased Eμ-Tcl1 mice with a 5% to 15% tumor load in peripheral blood were treated with the decoy receptor protein LTβR–Ig. Eμ-Tcl1 mice treated with a control Ig showed a substantially increased tumor load in peripheral blood from days 0 to 38 of treatment, whereas no such increment was observed in LTβR–Ig–treated mice (Fig. 7A). Tumor load in the spleen of isotype-treated Eμ-Tcl1 mice was also higher than in LTβR–Ig-treated animals (Fig. 7B). Immunohistology showed substantial FDC disappearance (Fig. 7C) and a severe disturbance of MAdCAM-1⁺ and BP-3⁺ B-cell follicle stromal structures (Supplementary Fig. S9A) in LTβR–Ig–treated animals. Splenic Ccl21 and Cxcl13 gene expressions were lower in LTβR–Ig–treated Eμ-Tcl1 mice compared with controls (Fig. 7D).

Supporting a pathogenic role of LTβR signaling, disease onset and progression were substantially delayed in Ltα^−/−Eμ-Tcl1 double-transgenic mice compared with Eμ-Tcl1 animals (Supplementary Fig. S9B). Hence, tumor cell production of LTα accelerates tumor growth. Conversely, interference with LTβR signaling leads to growth reduction in a murine CLL model.

In line with the data from Eμ-Tcl1 mice, lymphotoxin and TNFα transcripts could be detected in sorted human CLL cells, normal B cells, and in the human B-CLL line MEC-1 (Fig. 7E). Adoptive transfer of MEC-1 cells into NOD/SCID/c-γ-chain^−/− mice elicited a follicular-like FDC network (Fig. 7F, left), a process which could be completely abrogated by LTβR–Ig treatment (Fig. 7F, right). Thus, LTβR signaling is an essential part of the reciprocal relationship between leukemia B cells and mesenchymal stromal cells.

In this study, we tracked the trafficking routes of murine CLL cells into protective microenvironmental niches in SLOs, and identified FDCs as a crucial resident stromal cell population that supports consecutive steps of leukemia pathogenesis. Using the Eμ-Tcl1 transgenic mouse (13) as a CLL model, we also obtained functional evidence that the chemokine receptor CXCR5 has a dominant role in leukemia cell microanatomic localization.

Eμ-Tcl1 B cells showed a hierarchy among expressed homeostatic chemokine receptors. CXCR4 has been identified as a survival factor in CLL (39) and as an important homing receptor of neoplastic B cells to the bone marrow (40), whereas CXCR5 expression is predominantly associated with CLL positioning in SLOs (41). Although CXCL12–CXCR4 engagement accelerates human B-CLL and murine Eμ-Tcl1 leukemia cell proliferation in vitro, in vivo this activity requires access to a CXCL12-providing niche. In our murine CLL model, CXCR5 crucially controls homing of leukemia cells into the B-cell follicles of SLOs. CXCR4 could not compensate for CXCR5 deficiency regarding their migratory functions: CXCR5-deficient leukemia cells were restricted to the MZ zone but not attracted to the B-cell follicle, and tumor cells remained absent from the dark zone of the GC, the attraction to which is CXCL12–CXCR4 governed (36).

In the absence of CXCR5, spontaneous onset of disease in Eμ-Tcl1 mice was severely delayed. Gene expression profiling uncovered a proliferative advantage in CXCR5-expressing Eμ-Tcl1 tumor cells, whereas no obvious differences were found in apoptosis-mediating pathways. In conjunction with a compartment-specific higher proliferation rate, this observation is in line with the view that B-CLL is not a static or accumulative disease that simply results from long-lived lymphocytes defective in apoptosis (42, 43).

Migration of normal B cells toward follicles is mediated by the CXCL13–CXCR5 signaling axis and a stromal cell network (25, 44). CXCL13, which is produced by the FDC network and MRCs, guides circulating naïve B cells in the proximity of FDCs, a prerequisite for the formation of B-cell follicles (36). Moreover, CXCR5 also plays a unique role in trafficking and homing of B-1 B cells (45). In patients with CLL, leukemia cells express high levels of functional CXCR5, and significantly higher CXCL13 serum levels were found compared with healthy controls (41).

With regard to these correlative studies in patients with CLL, we aimed to dissect the CXCR5-dependent spatial positioning of Eμ-Tcl1 leukemia cells in vivo. We found that upon adoptive transfer, Eμ-Tcl1 leukemia cells lodged within the splenic B-cell follicles in close proximity to FDCs. In sharp contrast, Cxcr5^−/− leukemia cells accumulated in the MZ of the spleen. Bajénoff and colleagues (46) showed that T and B lymphocytes are guided by FRCs and enter the WP across MZBCs. Here, we observed that Eμ-Tcl1 leukemia cells directly crossed the MZ sinus and reached the B-cell follicle faster than follicular B cells. Hence, leukemic cells exhibited remarkable functional similarity to MZ B-2 lymphocytes, which also shuttle directly from the MZ to the B-cell follicle (26). Gene expression profiling revealed a higher similarity of Eμ-Tcl1 leukemia cells with B-1 B cells, as compared with MZ B cells. This result is consistent with the initial classification of these tumors (47), and would also fit with several properties of the human unmutated variant of B-CLL. Although physiologic differences between murine and human B-cell populations exist, prohibiting definitive conclusions (48), similarities are also appreciated that affect structural restriction of the BCRs, the polyreactivity toward autoantigens, expression of CD5, and the presentation as activated antigen-experienced B cells (3, 9, 49). B-1 cells can induce and associate with FDCs in mice; however, trafficking routes of these cells have not been thoroughly analyzed (50).

CXCL13–CXCR5 signaling is crucial for B-cell migration under steady-state conditions and enhances antigen encounter and BCR-triggered B-cell activation (51). A strong correlation has been proposed between antigen-stimulated BCR signaling and the clinical course of CLL (3). Involvement of a microbial-derived or autoantigen component in the heavy chain Ig repertoire selection has been suggested in many CLL cases, leading to the occurrence of stereotyped BCRs (1, 52). Alternatively, a BCR-mediated antigen-independent, cell-autonomous signaling mechanism might account for CLL signaling activity. However, this in vitro model does not rule out involvement of in vivo extrinsic antigens as additional enhancers (53).

How could Eμ-Tcl1 leukemia B cells receive proliferation stimuli in vivo? Using two-photon microscopy, we found that Eμ-Tcl1 leukemia cells showed decreased displacement rates and increased interaction times with follicular FDCs compared with leukemia cells outside this area. This could result from their cognate nature and ability to encounter antigen-laden FDCs. The preferred tumor cell localization in the FDC-rich GC light zone may also be due to the paracrine provision of growth-promoting factors (34, 35). BCR engagement, as evidenced by activated BTK, and the receipt of growth factors could act differentially during tumor ontogeny and in a complementary manner. In this view, a strong antigen dependence exists in early stages, followed by competition for natural ligand availability, then acquisition of additional oncogenic lesions, and, finally, selection for mutants that can co-opt autonomous signaling from either the BCR or their downstream effectors (54).

The differentiation of FDCs is essential for SLO organogenesis (55) and final FDC maturation crucially requires lymphoid tissue-inducer (LTi) cells and B cells expressing LTαβ (56). Consistent with previous reports that B-2 and B-1 B cells can induce the formation of mature FDC networks (50, 57), a human leukemia cell line efficiently induced FDC networks in NOD/SCID/c-γ-chain^−/− mice. This implies that B leukemia cells themselves provide the crucial factors for FDC differentiation in vivo. The cross-talk between leukemic B cells and rudimentary stromal cells could be of relevance for a posttherapeutic minimal residual disease state. Irradiation of WT mice elicited an accumulation of leukemia B cells around radio-resistant FDCs and MRCs. This spatial proximity accelerated the growth rate of the tumor cells, most likely through uncompeted access to FDCs.

Notably, much of the cellular proliferation in human CLL arises in pseudofollicular proliferation centers in SLO, which are unique to CLL among all other B-cell malignancies, but they are observed in inflamed tissues of patients with systemic autoimmune disorders as well (1). Functional similarities between inflammation- and CLL-associated proliferation centers may include their capacity to initiate or maintain autoantigen stimulation, indicating that autoimmune diseases and CLL could share part of their pathogenic trait. The occurrence of FDCs in proliferation centers has been reported (58), thus supporting the view that leukemia B cell–FDC encounter is a major source of cellular proliferation.

We infer from our data that leukemia cell–associated LTαβ confers the required ligand such that FDC networks can be induced de novo. In this paracrine network, further FDC stimulation with cell-bound LTαβ can induce the expression of CXCL13 and other proinflammatory cytokines (59). To break this feedback loop, we treated mice with an LTβR–Ig fusion protein and found that targeting the LTβR signaling pathway profoundly retarded murine CLL growth. FDC loss and CXCL13 reduction indicated that inhibition of stroma–lymphoma cell cross-talk can translate into an ameliorated clinical course. Targeting the stromal environment in SLOs is an appealing treatment option for indolent lymphoma because these benign compartments are subject to low selective pressure for mutations and epigenetic changes (60).

We propose that our stroma induction model can mimic an early pathogenic condition in which, upon arrival of transformed clonal CLL cells in SLO, leukemia cells encounter a microenvironment that enhances their proliferation. Reciprocally, tumor cells impose stroma remodeling. Increased proliferative activity in proximity to FDC networks could enhance genetic vulnerability, resulting in mutation acquisitions that render tumor cells increasingly independent from antigenic triggering (1).

When applying an LTβR-decoy receptor, selective targeting of the stroma interface alone led to tumor growth retardation; this strategy is applicable to a minimal residual disease state in which tumor cells might be protected in LT-dependent niches. Furthermore, it may be possible to inhibit leukemia cell access to antigen-presenting cells, either through selective interference with CXCR5 activity as shown here, or by targeting the FDC function in growth factor supply. Both immunologic treatments are rational strategies to complement traditional cytotoxic therapies for the cure of leukemia and lymphoma.

More detailed methods can be found in Supplementary Methods.

The microarray data presented in this publication have been deposited in NCBI's Gene Expression Omnibus (61) and are accessible through GEO series accession number GSE60925.

Assays were performed in 5-μm-pore Transwell plates (Corning) for 4 hours at 37°C, exactly as described previously (16). CCL21, CXCL12, and CXCL13 were used at a concentration of 100 nmol/L, 25 nmol/L, and 300 nmol/L, respectively (R&D Research Diagnostics).

Peripheral blood samples from treatment-naïve CLL patients were purified over a Ficoll gradient. Tumor cells (CD19⁺CD5⁺) were FACS-sorted and RNA was immediately extracted. The study was conducted according to the Declaration of Helsinki and in accordance with local ethical guidelines; written informed consent of all patients was obtained.

Eμ-Tcl1 transgenic mice were monitored for signs of disease by quantification of the frequency of tumor cells in peripheral blood by flow cytometry (CD5⁺CD19⁺). Spleen-derived leukemia cell suspensions were prepared from diseased mice (>8 months old and tumor cell load >50% in the periphery) by tissue homogenization, depletion of red blood cells, and FACS sorting.

Murine bone marrow stromal cells (M2-10B4; ATCC-CRL-1972) were obtained from the ATCC in July 2010. The cells were passaged two to three times over a period of 3 weeks and aliquots were frozen in liquid nitrogen. All experiments were performed with these aliquots. The human FDC line FDC/HK was obtained from Y.S. Choi (Ochner Clinic foundation, New Orleans, LA) in 2011 and cultured as described above. No further authentication was done for both cell lines.

Blocking LtβR–Ig [100 μg; murine receptor fused to mouse mAb IgG1; Biogen Idec or generated as described previously (62)] or isotype control antibody (MOPC 21) were injected repeatedly i.p. into tumor challenged and Eμ-Tcl1 transgenic mice, respectively.

Results are expressed as arithmetic means ± SEM if not otherwise indicated. Values of P ≤ 0.05 were considered statistically significant, as determined by the unpaired Mann–Whitney test, the two-tailed unpaired Student t test, or the Wilcoxon signed rank test where appropriate.

No potential conflicts of interest were disclosed.

Conception and design: A.E. Hauser, G. Lenz, A. Rehm, U.E. Höpken

Development of methodology: K. Heinig, V. Stache, A.E. Hauser, R. Brink, U.E. Höpken

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): K. Heinig, M. Gätjen, I. Anagnostopoulos, K. Gerlach, R.A. Niesner, Z. Cseresnyes, A.E. Hauser, T. Hehlgans, J. Westermann, A. Rehm

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Gätjen, M. Grau, V. Stache, I. Anagnostopoulos, R.A. Niesner, Z. Cseresnyes, A.E. Hauser, P. Lenz, J. Westermann, G. Lenz, A. Rehm, U.E. Höpken

Writing, review, and/or revision of the manuscript: K. Heinig, M. Grau, R.A. Niesner, A.E. Hauser, T. Hehlgans, J. Westermann, B. Dörken, M. Lipp, G. Lenz, A. Rehm, U.E. Höpken

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): A.E. Hauser, R. Brink, A. Rehm

Study supervision: B. Dörken, M. Lipp, A. Rehm, U.E. Höpken

The authors thank Carlo Croce (Columbus University, Columbus, OH) and Jeffrey L. Browning (Biogen Idec, Cambridge, MA) for providing essential reagents and Heike Schwede and Kerstin Krüger for excellent technical assistance.

This work was funded by grants from the Deutsche Krebshilfe (grant number 107749), German Research Foundation (DFG), and the Berliner Krebsgesellschaft to U.E. Höpken and A. Rehm; grants from the DFG, the Deutsche Krebshilfe, and the Else Kröner-Fresenius-Stiftung to G. Lenz; the DFG-funded project “JIMI-a network for intravital microscopy” to R.A. Niesner, Z. Cseresnyes, and A.E. Hauser; and by a Doctoral Scholarship from the Philipps-Universität Marburg to M. Grau.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.