Lymphoma Angiogenesis Is Orchestrated by Noncanonical Signaling Pathways

The concept of reciprocal cross-talk between malignant B cells and their microenvironment in secondary lymphoid organs (SLO) is supported by genetic signatures found in high-grade B-cell non-Hodgkin lymphoma (B-NHL). Gene expression profiling (GEP) of primary diffuse large B-cell lymphoma (DLBCL) samples showed that differences in the microenvironment affect patient survival after chemotherapy treatment. The stromal-1 signature, related to extracellular matrix (ECM) deposition and histiocytic infiltration, is associated with favorable outcome, whereas the angiogenesis-related signature (stromal-2 signature) is prognostically unfavorable (1). The microanatomic correlate of this angiogenesis-related signature is an increased microvessel density (MVD; refs. 2, 3).

An essential structural and functional component of the lymphoma stroma consists of the blood and lymphatic vasculature, which besides servicing the local cells, shapes the tumor survival niche. Antiangiogenic strategies are based on the activities of VEGF family molecules in cancer-associated angiogenesis (4, 5). Lymphoma cells themselves exhibit proangiogenic activity as they release VEGFs, but attempts to combine multimodal chemo-/immunotherapies with VEGF inhibitors have not been beneficial in B-cell lymphoma (6, 7). Therefore, it is plausible that other non-VEGF angiogenic pathways prevail and cause enhanced vascular assembly. Lymph nodes (LN) are equipped with specialized endothelial cell (EC) subsets that line the blood and lymphatic vasculature. A comparison of transcriptomes of fibroblastic reticular cells (FRC), blood ECs (BEC), and lymphatic ECs (LEC) taken from naïve LNs and LNs exposed to an inflammatory milieu showed marked immune cell recruitment in the latter (8). However, it remains unclear whether exposure to lymphoma has the same impact on stromal cells as inflammation.

Earlier studies analyzing lymphoma-engendered alterations in ECs relied on syngeneic or xenogeneic lymphoma cells transplanted subcutaneously. In these models, marked adaptations of the lymphatic vasculature occurred, which correlated with the increased availability of VEGFC from tumor-associated macrophages (9). However, introducing cells subcutaneously or into xenogeneic hosts does not necessarily reflect the dissemination behavior of aggressive B-cell lymphoma (10). Intravenously administered Myc-driven B-cell lymphoma cells enter distinctive microanatomic compartments of LNs and spleen via CCR7- and CD62L-regulated mechanisms, indicating that tumor cell immigration is mediated via high endothelial venules (HEVs) (11). HEVs are segments of postcapillary venules and are characterized by specialized BECs that express chemokines and adhesion molecules required for extravasation of lymphocytes into LN parenchyma (12). Within the niche, lymphoma cells stimulate reciprocal cross-talk with CD45⁻/gp38⁺ FRCs in a lymphotoxin β receptor (LTβR)–dependent manner (13). During an immune response, activated lymphotoxin α (LT)−expressing B cells directly induce differentiation of ECs into a HEV phenotype (14, 15). The precise contribution of morphogenic cells and molecular factors to angiogenesis is likely to vary depending on the stimuli, and it remains unclear whether lymphoma B-cell exposure engenders a condition that phenocopies high MVD. Here, we explored lymphoma-induced LN vascular reprogramming in an Eμ-Myc lymphoma-transplanted mouse model. GEP of FRCs, LECs, and BECs and comparison of angiogenic factors expressed by stroma and lymphoma cells revealed that mechanisms different from conventional solid tumor angiogenesis are active in B-cell lymphoma–exposed LNs.

The full list of mice used is given in Supplementary Materials and Methods. All experiments were conducted in compliance with the institutional guidelines of the Max Delbrück Center for Molecular Medicine (Berlin, Germany) and approved by the Landesamt für Gesundheit und Soziales Berlin, Germany (G0104/16; G0052/12; G0373/13; G 0058/19).

Formalin-fixed paraffin-embedded biopsy specimens from DLBCL of not otherwise specified (NOS) subtype with either a BCL2 or BCL6 rearrangement (single hit), high-grade B-cell lymphoma with MYC and BCL2 or BCL6 rearrangements (double hit), from Burkitt lymphoma (BL) as well as from nonneoplastic palatine tonsils, and nonneoplastic lymph nodes (LN), were retrieved from the archives of the Institute of Pathology, Charité-University Medicine (Berlin, Germany).

Multiple tissue arrays (MTA) were obtained from US Biomax and contained various DLBCL specimens that were not further diagnosed according to cytogenetic rearrangements. IHC was performed as described in Supplementary Materials and Methods. The study involving primary human tissues was conducted according to the declaration of Helsinki and in accordance with local ethical guidelines.

The human B-NHL cell lines Su-DHL-4 and OCI-Ly7 (DLBCL), JeKo-1 [mantle cell lymphoma (MCL)] were obtained from DSMZ; Raji cells (Burkitt lymphoma) were from ATCC. Upon receipt, all cell lines were expanded over 1 to 2 weeks, and aliquots were immediately frozen in liquid nitrogen. Gene expression analysis was performed 2 to 3 days after thawing. Mycoplasma testing was not performed. Human umbilical vein cells (HUVEC) were purchased from Promocell and cultured in endothelial cell medium over 2 to 3 passages before use.

Patient-derived DLBCL and MCL xenograft samples (PDX) were obtained from Dana-Farber Cancer Institute (PRoXe depository, Boston, MA) and used after a single NOD.Cg-Prkdcscid Il2rg tm1 Wjl/SzJ (NSG) mouse passage. All primary cell lines were directly obtained from public depositories or from a commercial supplier and not additionally authenticated before use. Primary B cells (CD19⁺ CD45⁺ CD69⁻ CD4⁻ CD8⁻ CD14⁻ 7AAD⁻) from healthy donors were purified from PBMCs via a Ficoll gradient and further sorted by flow cytometry.

HUVECs were starved overnight in serum and growth factor–free endothelial basal medium (Promocell) at a cell density of 1 × 10⁴/cm² in a 24-well plate. Cell cultures were supplemented with VEGFA₁₆₅ and VEGFC, in addition LTα1β2 was added (all 10 ng/mL) for 24 hours. For VEGFC, a 13,5 kDa non-disulfide linked homodimeric protein consisting of two 116 amino acid polypeptide chains was chosen.

In a spheroid-based angiogenesis assay, HUVECs (at passage 3 for all experiments) were cultured for two days, trypsinized, washed in PBS, and resuspended in basal medium mixed with methocel stock solution (0.012% in basal medium) in a 4:1 ratio. Twenty-five microliters of cell suspension was pipetted dropwise on nonadherent culture dishes, turned upside-down to form hanging drop cultures. Collagen mix was prepared on ice using Collagen type I diluted in 10× PBS in a 4:1 ratio, pH 7. Spheroids were obtained after 24 hours, mixed in methocel–collagen medium, plated in cell culture dishes, and polymerized for 30 minutes at 37°C. Spheroids were stimulated with VEGFA, VEGFC, LTα1/β2 (all 25 ng/mL), and anti–VEGFR2 antibody (50 ng/mL, clone: 89106 R&D Systems) for 24 hours. Spheroids were fixed in 4% PFA, images were recorded with transmission microscopy, and sprouts were analyzed with ImageJ.

Single-cell suspensions were prepared and transferred intravenously exactly as described previously (13); at least 2 to 6 independent lymphoma clones derived from different Eμ-Myc mice were tested. Tumor load in recipient mice was determined by spleen weight, or by flow cytometric analysis. Transplantation of MCA313 primary fibrosarcoma cells was performed by subcutaneous injection of 1 × 10⁶ cells in PBS.

LTβR-blocking immunoglobulin (LTβR-Ig; 100 μg; Biogen Idec) or IgG1 isotype control antibody MOPC21 was injected twice intraperitoneally (100 μg), exactly as described previously (16).

Mice were injected intraperitoneally on days 1 and 6 after Eμ-Myc lymphoma cell administration either with 0.8 mg rat anti–VEGFR2 antibody (clone DC101), or with isotype control IgG1 (both from Bio X Cell). After fibrosarcoma transplantation, once small subcutaneous nodules became palpable, antibodies were injected on two occasions at 5-day intervals.

The VEGFR3 kinase inhibitor SAR131675 (SelleckChem) was dissolved in DMSO and further diluted in 0.6% methylcellulose/0.25% Tween 80. Animals were treated for 6 consecutive days with 100 mg/kg b.w. of the drug per os, starting at day 2 after Eμ-Myc lymphoma transplantation, or when fibrosarcoma became palpable.

LN stromal cells were sorted from Wt controls and from animals transplanted with Eμ-Myc lymphoma cells. The average lymphoma load in LNs was 4% to 10% of all lymphocytes. Stromal cell purity after sorting was >95%.

For each sample, 80 ng RNA was pooled from 4 to 6 independent sorting experiments and reverse-transcribed using the Illumina Total Prep RNA Amplification Kit. The biotin-labeled transcripts were hybridized to Illumina Mouse WG-6 v2.0 Expression BeadChips and processed for detection. Differential expression was evaluated as log₂-fold change (FC). A detailed description of microarray data generation, qRT-PCR, and bioinformatic processing is given in Supplementary Methods and were performed essentially as described previously (16). Data are deposited under GEO repository accession number GSE126033.

Splenocytes were labeled with SNARF-1 fluorescent dye (Thermo Fisher Scientific), and then, 3 × 10⁷ cells per recipient mouse were injected intravenously. Four hours after cell transfer, mice were sacrificed, inguinal LNs were dissected, and cells were analyzed by flow cytometry.

Detection of bromodeoxyuridine (BrdU) incorporation into proliferating ECs in mice was done essentially as described previously (16).

FITC-coupled dextran polymers (10, 40, and 150 kDa; all Sigma-Aldrich) diluted in PBS were injected intravenously (0.5 mg) into control, Eμ-Myc cell-transferred, or subcutaneously fibrosarcoma-bearing mice. Vessel perfusion was examined by intravenous injection of fluorescence-coupled Isolectin GS-IB4 (Thermo Fisher Scientific).

Inguinal LNs or fibrosarcoma tumors were fixed in 4% paraformaldehyde/PBS for immunohistology.

Statistical data were evaluated using GraphPad Prism (Version 6) software. The confidence level was 95%, with a significance level of 5% (α = 0.05). Results are expressed as the arithmetic means ± SEM. Data comparison with P values of ≤0.05 was considered statistically significant. P values were calculated by Wilcoxon signed-rank test, Mann–Whitney U test for nonnormally distributed data, a two-tailed unpaired Student t test for normally distributed data, or a paired Student t test, as indicated.

In humans, areas of high MVD were seen in aggressive B-NHL samples (Fig. 1A and B). Compared with nonneoplastic tonsils or LNs, the number of large vessels in the DLBCL samples was not affected, but the number of small vessels was increased 1.5- to 3-fold (Fig. 1C). These ratios were higher in high-grade B-NHL (MYC and BCL2 or BCL6 double translocated) and Burkitt lymphoma compared with single-hit DLBCL or DLBCL with unknown rearrangements. Small vessels were rarely PNAd positive (2/30 DLBCL cases), indicating that they represented a capillary phenotype, but not HEVs (Supplementary Fig. S1). Functionally, this lower vessel differentiation might result in impaired immigration of immunoprotective effector T cells (17).

To address the mechanisms of lymphoma-induced angiogenesis, we used transgenic Eμ-Myc mice that spontaneously develop lymphoma that mimic important aspects of human high-grade B-cell lymphomas (13, 18). Eμ-Myc lymphoma cells were transferred by intravenous injection into Wt mice. Between 9 and 12 days after lymphoma cell transfer (mean 5%–15% of all CD45⁺ cells), LNs rapidly developed a 3-fold increase in small vessels compared with an untreated cohort (Fig. 1D and E). This process was already visible by days 5 to 8, indicating that angiogenesis might be an early prerequisite for lymphoma progression. Anti-CD31 staining of LNs from spontaneously diseased Eμ-Myc transgenic mice with a high tumor burden revealed the occurrence of a majority of small vessels, which had substantial morphologic similarities both to LNs from Eμ-Myc lymphoma-transplanted mice and to human DLBCL (Fig. 1F).

The murine Eμ-Myc model therefore mimics microanatomic aspects of lymphoma-induced vascular alterations that occur in human high-grade B-cell lymphoma.

Stromal populations from inguinal LNs derived from mice were identified by differential expression of gp38, the adhesion molecule CD31, and the LEC marker Lyve1. The blood vasculature was clearly distinguishable from lymphatics by the absence of Lyve1 staining and, additionally, by a higher intensity staining of CD31 (Fig. 2A).

FRCs, LECs, and BECs were sorted from stroma-enriched fractions (CD45⁻) of pooled LNs from untreated and Eμ-Myc lymphoma–transplanted animals (Fig. 2B). Mice were sacrificed when tumor cells typically made up less than 10% of leukocytes in the LNs (days 8–12). FRCs exhibited the highest gene expression of the chemokines Ccl19, Ccl21, and of Il7. BECs expressed a high level of ESAM, and LECs expressed the marker gene Lyve1 (Fig. 2C and D). Using whole transcriptome GEP relative changes in gene expression from untreated stromal cells were compared with those from lymphoma-challenged mice. Subset-specific expression of the expected gene markers was confirmed (Supplementary Fig. S2A). Gene expression of lymphoma B-cell–specific genes, such as Ptprc (CD45R/B220), Ms4a1 (CD20), and Cxcr5 within datasets from stromal subsets was rigorously excluded (Supplementary Fig. S2B).

In all three subsets, there were significant overall differences in expression between the treated and untreated mice (Fig. 2E). A total of 710 genes were selectively upregulated in lymphoma-exposed cells. An overlap of 114 genes that were upregulated in all three subpopulations occurred (Fig. 2F and G). Tumor induction and progression has been linked to a chronic type of inflammatory milieu. Therefore, we explored whether lymphoma-dependent stromal cells showed expression of genes characteristic of a LN-specific inflammation signature (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE15907; ref. 8). Unexpectedly, no significant upregulation of inflammation signature genes was observed (Supplementary Fig. S2C and S2D).

On the other hand, similar to an inflammatory condition, gene ontology (GO) analysis (Supplementary Table S1A) as well as GSEA-confirmed enriched signatures related to proliferation and cell cycle in all three stromal cell subsets (Supplementary Fig. S3A–S3C; Supplementary Table S1B). Phenotypically, the proportion of BECs, LECs and FRCs in S-phase (EdU⁺) was substantially higher than in those from untreated LNs (Supplementary Fig. S3D). Collectively, the angiogenic switch in lymphoma-bearing LNs is characterized by proliferation of BECs. Gene expression and proliferation data confirmed the role of lymphangiogenesis during lymphoma progression. Lymphangiogenesis has been established as a structural pathogenetic factor supporting lymph flow and lymphoma dissemination (9, 19–21).

As an indication of structural alterations, in FRCs, the KEGG gene sets correlated with “Collagen_Formation” and “Extracellular_Matrix_Organization” were significantly enriched with downregulated genes (Supplementary Table S1B). Both signatures contain genes (e.g., Collagen genes; Timp1, 2) that determine the formation and stability of the ECM. Among other functions that sustain tissue homeostasis, one important function of the ECM includes VEGFA gradient shaping and integrin signaling, which are both involved in regulating the balance between stalk cell proliferation and tip cell migration in the formation of the vascular branches (22). Thus, this signature may indicate a perturbance of ECM function regarding guidance of angiogenesis.

BECs assemble into microvessels associated with a negative clinical prognosis; thus, we focused on blood vessels in greater detail.

Inguinal LNs were removed when tumor load was low (5%–10% of all CD45⁺ cells) or medium (10%–15%), RNA was extracted and analyzed using angiogenesis gene-specific qRT-PCR arrays (Fig. 3A). With a medium tumor load, among the strongest upregulation of mRNA compared with nontumor bearing LNs was noted for Vegfc (>4-fold); VEGFC is a ligand for VEGFR2 and VEGFR3 and functions as a growth factor for lymphangiogenesis and for BECs (23).

GEP on nontransformed follicular B cells derived from control mice and Eμ-Myc lymphoma cells from transplanted mice was performed. The latter exhibited gene sets characteristically upregulated in human DLBCL as well (FC > 1, P < 0.05; Supplementary Fig. S4A and S4B). Applying an angiogenesis-related GSEA signature, several strong immunomodulators and proangiogenic factors were found to be substantially upregulated in lymphoma cells. This list contained genes encoding secreted proteins such as Lgals1 (Galectin-1) and Lgals9 (Galectin-9) involved in suppressing T lymphocytes and niche formation (Fig. 3B; ref. 24).

Vascular remodeling at the disease onset was analyzed by confocal microscopy on thick tissue sections (100 μm). Because transplanted Eμ-Myc lymphoma cells home preferentially to the paracortical T-cell zone of LNs in a CCR7-dependent manner, we costained the T-cell area (CD3⁺) and quantified the CD31⁺ 3D surface area within it. LECs (Lyve1⁺) were largely absent from the paracortical T-cell zone; however, they were also found to be expanded (Supplementary Fig. S4C; Fig. 3D). Lymphoma induced an expansion of HEVs in the LNs of both the Eμ-Myc transgenic and transplanted animals, and they had many more small vessels and capillaries extending into the T-cell zone than control LNs (Fig. 3C and D).

During sprouting angiogenesis, growing capillaries are spearheaded by specialized ECs termed “tip cells” (25, 26). We analyzed the BEC-specific gene expression from lymphoma-challenged LNs against a refined signature of tip cell genes; however, no strong enrichment with respect to the BEC-specific ranking of genes was found (Fig. 3E; Supplementary Fig. S4D). Likewise, the KEGG “VEGF Signaling Pathway,” which relates to VEGFA/VEGFR2–induced signaling, was not enriched (Fig. 3F; Supplementary Fig. S4E). In contrast, a gene signature of distinct transcriptional responses to VEGFC/VEGFR3 stimulation was modestly enriched among the top 10 genes up- or downregulated in BECs (Fig. 3G; ref. 27).

GSEA indicated a significant enrichment of genes downregulated by the hypoxia-related HIF1α gene among the genes upregulated in BECs (Fig. 3H), suggesting a lack of activation of this signaling route. Tissue staining using the hypoxia marker pimonidazole was negative in LNs from lymphoma-transplanted mice and scant in LNs from Eμ-Myc transgenic mice, which was in striking contrast to the marked hypoxia demonstrated in subcutaneously grown fibrosarcoma tissue (Supplementary Fig. S4F). A Notch target gene signature was downregulated in BECs (Fig. 3I), indicating the lack of a Notch-dependent feedback loop from tip to stalk cells.

A stem cell transcriptional pattern was enriched among the BEC-specific genes. A “Stemness signature” (28) of 100 genes was selected, 35 of which were upregulated (FC > 1; P ≤ 0.001) among the BEC-specific genes (Fig. 3J). Genes related to spindle assembly and kinetochore formation, for example, Bub1, Mad2l1, and Nusap1 were significantly different. The forkhead TF Foxm1, which binds to the majority of kinetochore and cell division gene promoters (29), was upregulated (FC = 2.6), indicating that it has a coordinating function in lymphoma-induced BEC reprogramming.

We visualized ECs in LNs from tamoxifen-inducible Cdh5-GFP reporter mice. The vascular surface area increased 2-fold upon lymphoma induction. Correspondingly, vessel length increased about 2-fold (mean control: 16.687 μm; mean Eμ-Myc: 32.767 μm; P = 0.05). A complex vascular network intermingled with dense loops with smaller diameters, and concomitantly, a 3-fold increased frequency of branching points was seen (Fig. 4A–C). However, upon normalization of vessel length, distances between branches and loops were not significantly changed, indicating that the leading morphologic effect was an increase in vascular density. The isolectin GS-IB4 was infused systemically. Perfused vessels were quantified using the ratio of all double-stained IB4⁺/GFP⁺ vessels relative to all GFP⁺ vessels. In total, about 80% of the vasculature was found to be open for intraluminal flow (Fig. 4D).

Blood vessels of lymphoma-challenged and untreated LNs were mainly impervious to fluorescent 10, 40, and 150 kDa dextrans, whereas vessels in fibrosarcoma tissue leaked 10 and 40 kDa forms (Supplementary Fig. S5A–S5C). In addition, mice were inoculated with Eμ-Myc B cells, and then transplanted with SNARF-1–labeled splenocytes. Total numbers of CD4⁺ and CD8⁺ T cells, monocytes (CD11b⁺), and B lymphocytes (B220⁺) were about 2-fold higher in lymphoma-bearing LNs (Supplementary Fig. S5D).

In Cdh5-GFP mice, lymphoma-exposed LNs exhibited numerous slender filopodial bursts that emanated from the lateral sides of vessels, but only rarely extended from the leading edge where tip cells are usually located (Fig. 4E). Several cellular protrusions reminiscent of blebs normally associated with apoptosis, cell migration, and division in ontogeny (30, 31) were observed. Next, to explore tip cell occurrence further anti-Esm1 immunostainings were performed (Fig. 4F). Esm1 expression was found to be EC (Cdh5-GFP⁺) restricted, but rather distributed throughout the blood vasculature. This pattern is consistent with the observation that Esm1 expression is restricted to tip cells in the retina, but widely distributed throughout carcinoma vasculature with a disturbance of the VEGFA gradient (32). The Notch ligand Dll4 was variably expressed in ECs with filopodia (Supplementary Fig. S5E). Collectively, disseminated localization of Esm1⁺ sprouting ECs together with variable Dll4 expression are distinguishable from highly ordered morphologies and functions of tip cells in developing organs.

To determine the role of LTβR–LTαβ signaling, we transplanted B cells from Eμ-Myc and Eμ-Myc × Ltα^−/− mice. Using a cell-cycle gene-specific qRT-PCR array, we found 34 of 78 genes upregulated in Eμ-Myc lymphoma-exposed BECs (Fig. 5A). In contrast, this proliferation signature was not obtained when Ltα-deficient Eμ-Myc lymphoma cells were transplanted (4/78 genes upregulated; Fig. 5B).

When Eμ-Myc lymphoma-bearing mice received the inhibitory decoy receptor protein LTβR-Ig, a significant delay in the growth of lymphoma was observed. In addition, the vascular surface area (CD31⁺) and the proliferation rate (BrdU⁺) of BECs were almost normalized to those from control mice (Fig. 5C–F). In lymphoma-bearing Cdh5-GFP mice, the dense and hyperplastic GFP⁺ vasculature was not visible when animals were LTβR-Ig-treated (Fig. 5G); in particular, the number of branching points was about half that in controls (Fig. 5H and I); loss of cuboid ECs with a concomitant decrease in luminal PNAd staining was confirmed (Supplementary Fig. S6A).

To exclude that the growth reduction was caused by indirect effects on other LTβR-expressing cell populations (33), Cdh5-cre/ERT2 × LtβR^fl/fl mice were generated. Reduction in LTβR expression in BECs from tamoxifen-treated mice was confirmed by flow cytometry (Supplementary Fig. S6B and S6C). After administration of lymphoma cells, a substantial decrease in CD31⁺ vascular surface area, tumor load and ICAM1 expression was observed compared with Cdh5-cre/ERT2 controls (Fig. 5J and K; Supplementary Fig. S6D).

We examined the growth kinetics of Eμ-Myc lymphoma in mice treated with a blocking anti–VEGFR2 antibody. Compared with controls, similar lymphoma growth rates, identical vascular surface areas (CD31⁺), as well as branching points were obtained in lymphoma-exposed LNs (Fig. 6A–C). In contrast, anti–VEGFR2 antibody treatment caused a substantial tumor and vascular growth retardation in subcutaneously grown fibrosarcoma (Supplementary Fig. S7A–S7C).

Next, we separated LN stromal cell subsets, Eμ-Myc lymphoma cells, and nontransformed leukocytes. Upregulation of Vegfc was seen in FRCs, but the strongest gene expression was seen in the lymphoma B cells (Fig. 6D). Vegfd gene expression was reduced (Supplementary Fig. S7D). Correspondingly, VEGFC serum levels were much higher in Eμ-Myc mice (Fig. 6E). In accordance, several human lymphoma cell lines and the patient-derived lymphoma samples (PDX) expressed much higher VEGFA and VEGFC mRNAs compared with normal B cells (Supplementary Fig. S7E).

Stimulation of spheroid-grown HUVECs by VEGFC–induced sprouts at lower rate than treatment with VEGFA (Fig. 6F). Although LTα1β2 alone had no stimulatory effect on sprouting, the combination with VEGFC revealed a synergistic growth effect (mean 25%). This effect was not seen in the combination of LTα1β2 with VEGFA (Fig. 6F). Stimulation of HUVECs with VEGFC caused a strong VEGFR3 downregulation, which could be substantially ameliorated by a concomitant LTα1β2 stimulation (Fig. 6G). We conclude that LTβR signaling might interfere with VEFR-3 endocytosis or recycling, a general mechanism of receptor tyrosine kinase regulation (34).

In whole-tissue homogenates from lymphoma-bearing LNs, a 4-fold upregulation of VEGFC protein expression was detected. The active variant of the VEGFC protein is generated after enzymatic cleavage of a propeptide. Upon lymphoma challenge, there was increased conversion to the mature protein in spleen and LNs, indicating that lymphoma had induced enzymatic activities necessary for generating biologically active VEGFC (Fig. 6H and I). Gene expression of collagen- and calcium-binding EGF domains 1 (Ccbe1) was strongly enhanced in Eμ-Myc lymphoma cells, but not in LN stroma (Fig. 6J). CCBE1 protein localizes together with pro-VEGFC and ADAMTS3 on ECM and EC surfaces, where a cleavage complex is assembled allowing the mature VEGFC to activate VEGFR3 (35). Importantly, fully processed VEGFC is not only lymphangiogenic, but also angiogenic (36, 37).

VEGFR3 is activated by VEGFC and -D and is highly expressed in leading-edge ECs that undergo sprouting (38). We detected Vegfr3 gene expression in BECs and LECs already in steady-state conditions, and observed a 2-fold induction in Vegfr3 expression in LECs upon lymphoma challenge (Fig. 6K). This upregulation is consistent with VEGFR3 signaling in LECs as a driving force for lymphangiogenesis. Vegfr2 mRNA was upregulated in LECs and BECs, which corresponded to a modest increase in VEGFR2 protein expression on BECs, but not in LECs (Supplementary Fig. S7F and S7G). To confirm the VEGFR3 specific pro-proliferative activity in vitro, HUVECs grown in spheroids were treated with VEGFA, VEGFC, and LTα1β2. In the presence of a VEGFR2–blocking antibody, VEGFC plus LTα1β2 still stimulated the formation of sprouts, which was abrogated in the combination of VEGFA and LTα1β2 (Supplementary Fig. S7H). Next, inducible Vegfr3-GFP reporter mice (Tg(Flt4-tm2.1-cre ERT2) Rosa26-mTmG) were transplanted with lymphoma cells. The highest frequency of GFP signal was found in LECs, followed by BECs, as confirmed by anti–VEGFR3 antibody detection (Fig. 6L; Supplementary Fig. S7I). We conclude that in vivo even a small population of VEGFR3 carrying BECs can gain a leading role in lymphoma-induced angiogenesis, independent from VEGFR2 activity.

Treatment with the VEGFR3 kinase inhibitor SAR131675, which has a much higher selectivity for VEGFR3 than for VEGFR1/2, inhibited both lymphoma progression and vascular expansion stimulated by lymphoma (Fig. 6M). A direct antilymphoma effect of the drug was excluded since spleen located Eμ-Myc cells were spared (Supplementary Fig. S7J). Concomitantly, expansion of the lymphatic vasculature (Lyve1⁺) was retarded (Supplementary Fig. S7K and S7L). Treatment with another VEGFR3 kinase inhibitor, MAZ51 decreased the density of the vascular surface (CD31⁺) in the T-cell zone 0.5-fold (Supplementary Fig. S7M). In fibrosarcoma, SAR131675 had no effect on tumor development (Fig. 6N).

Taken together, VEGFR3⁺ BECs in angiogenic vessels sensitized the vasculature for proliferation and differentiation inhibition by specific kinase inhibitors.

We elucidated the mechanisms regulating the angiogenic switch and the morphogenesis of new blood vessels in aggressive B-cell lymphoma. Using the transgenic Eμ-Myc mouse model, we detected a substantial morphological similarity with human aggressive lymphoma characterized by a high MVD, a feature that has an adverse prognostic impact on patient survival (2, 3). Lymphoma B cells provide growth stimuli that induce a marked proliferation of BECs. This lymphoma cell cross-talk with BECs in turn translates to vessel hypersprouting and capillary growth.

The clinical importance of angiogenesis for growth of solid tumors is well recognized (4, 39, 40). Therapeutic concepts from solid tumors targeting the VEGFA/VEGFR1/2 axis have been adopted for DLBCL and mantle cell lymphoma combination therapies, resulting in rather disappointing clinical outcomes (6, 7). Here, we show that BEC proliferation and angiogenesis proceeded in a VEGFR2–independent manner, a finding that would not have been anticipated from precedents set from studies of solid malignancies. Instead, we identified VEGFR3 as the leading mediator of blood vessel expansion, a signaling process fueled by lymphoma cell- and FRC-derived VEGFC. Although a much smaller fraction of BECs than LECs expressed VEGFR3, treatment with VEGFR3 inhibitors was sufficient to normalize angiogenesis and attenuate lymphoma expansion.

It is worthy to note that a limitation of pharmacologic inhibitor administration to block receptor tyrosine kinase activity is the specificity of their activity. Selectivity of VEGFR3 drug targeting might vary with the active concentration of SAR131675, MAZ51, or any other tyrosine kinase inhibitor, respectively (41). In addition, ligand-independent activation of VEGFR3 in angiogenesis may lead to inefficiency of an anti–VEGFR3 antibody treatment (42). In contrast, a BEC-specific genetic deletion of the VEGFR3 would allow more definite conclusions regarding the mechanism of lymphoma-induced angiogenesis. Here, we chose a drug application in our animal model that has the potential to reveal a therapeutic solution directly translatable into clinical studies.

The overexpression of the VEGFR3 ligands VEGFC by lymphoma cells and macrophages was interpreted to be the inducer of the lymphatic vasculature, but not of blood vessels (9, 21, 43). The VEGFC/VEGFR3 axis is important for lymphatic sinus growth and increased lymph flow that correlated with metastasis of solid tumors and lymphoma cell dissemination, respectively (19, 41). In contrast, we showed previously that the preferred dissemination route of transplanted Eμ-Myc B cells involves the LN T-cell zone, which can be accessed via HEVs (13). We also demonstrate structural remodeling and transcriptional reprogramming of lymphoma-exposed BECs, which are the major structural constituents of microvessels. To bring lymphangiogenesis and blood vessel growth in agreement, we envision that our intravenous transfer model reports on early steps of lymphoma B-cell homing and niche formation. These early steps are decisive for further LN remodeling and subsequently, lymphatic sinus growth and lymphoma cell dissemination. In LN, lymphatic and blood vasculature growth proceed in a VEGFR3–dependent manner.

Our results draw parallels with infections and solid tumors, but they also reveal clear distinctions from both. In clear contrast to many solid tumors and their LN metastasis, lymphoma-challenged BECs acquired gene signatures that were not enriched for Notch, Vegfa, or Hif1α pathway genes. LN remodeling in inflammation proceeds in sequential overlapping phases (12, 15, 44). In the initiation phase, a rapid blood vasculature growth is dependent on IL1β-secreting DCs that stimulate VEGFA release from FRCs (45). In contrast, we here examined LN from lymphoma challenged mice at days 8 to 12, when VEGFA has likely declined and instead, VEGFC has gained a leading role.

The TF HIF1α is induced by hypoxia and functions as the main transcriptional regulator of tumor-associated vascular remodeling (46). In conjunction with the weak pimonidazole staining seen in lymphoma-bearing LNs, our data point to a micromilieu in which hypoxia is not a driving force for angiogenesis. We conclude that LNs have a distinct vascular microanatomy that involves highly specialized HEVs and capillaries that can accommodate an autochthonous tumor without changing the tissue-specific oxygen metabolism.

Consistent with the inflammatory response upon infection, Eμ-Myc lymphoma-challenged LNs rapidly begin HEV remodeling (15, 33). One study shows that blocking VEGFA or its cognate receptor VEGFR2 does not inhibit LCMV-induced vascular expansion in LNs (15). This observation parallels a mechanism that we detected in lymphoma-induced angiogenesis; we also could not find evidence for a functional role of VEGFA in this process. During viral infection, B cell–derived LTα1β2 contributes to remodeling of the HEV network in the acute phase (15). The infection model examined the role of activated B cells in an environment perturbed by virus-induced structural disruption of the reticular network. In the current study, LT-expressing lymphoma B cells were being added to a largely intact LN microstructure. Thus, a functional B cell–dependent LTα1β2–LTβR signaling loop was present. Immunopharmacologic interference with LTβR signaling on BECs retarded angiogenesis substantially and delayed progression of B-cell lymphoma. Although LTα1β2 did not induce HUVEC proliferation and sprouting alone, we envision that it acts as a morphogen stimulating HEV or capillary branching and thus, may synergize with the growth factor VEGFC.

Interestingly, VEGFD–overexpressing tumor cells induced downregulation of BMP4 in HEV cells (47). Loss of BMP4 expression resulted in severe remodeling of HEVs, in contrast to our lymphoma model a key mediator of this process, VEGFD, was not increased in tumor-exposed LNs. Likewise, BMP4 gene expression in BECs was not significantly altered, indicating that these linked pathogenetic factors are unlikely to cooperate in lymphoma-induced vascular remodeling.

We note that branching points after LTβR-Ig treatment were strongly diminished. Regarding the hierarchy of this interaction, it has been shown that LTβR stimulation of murine FRC-type cells upregulates VEGF expression, suggesting that the LTβR-LTα1β2 axis on FRCs regulates VEGF levels and thus, EC proliferation (44).

Despite large similarities to infection-induced LN remodeling, we note that inflammation is not the cause for our observations. A comparison of our transcriptome dataset with microarray data from published results characterizing stromal response to inflammation showed no enrichment of an inflammation signature (8).

Morphologic analysis revealed a hypersprouting phenotype of lymphoma-associated vessels. GSEA of BECs did not indicate an enrichment of a tip cell or Notch signaling pathway gene signature, which would have been expected from ontogeny or other tumor angiogenesis models (48–50). Angiogenic sprouting is led by tip cells that sense a VEGFA guide cue and produce Dll4. This Notch ligand acts on stalk cells preceding the tip that proliferate but are prevented from adopting a tip cell differentiation (22). It has also been shown that VEGFR3 localizes in endothelial sprouts of several solid tumors and in the developing mouse retina in leading tip cells. In the absence of VEGFR2 activity, VEGFR3–transmitted signals can sustain some degree of angiogenesis (38). When Notch signaling is disrupted, increased sprouting in conjunction with induced Vegfr3 expression occurs (51). We note that Esm1 and Dll4 expression normally associated with a tip cell phenotype was irregular in lymphoma vasculature, contrasting to studies in highly ordered developing tissues like retina (42, 49, 52). Thus, in mice unopposed VEGFR3 activity that is released from a Notch signaling brake can result in a hyperactive sprouting phenotype, as seen in our Eμ-Myc lymphoma model.

In summary, we showed that lymphoma angiogenesis proceeds in unique transcriptional and morphogenic programs, which are clearly distinct from solid tumor and inflammation-induced structural remodeling. Identification of the VEGFR3 and the LTβR signaling pathways as regulators of lymphoma angiogenesis strongly suggests that combinatorial therapies targeting vessel densities should be reconsidered.

No potential conflicts of interest were disclosed.

Conception and design: M. Gloger, M. Zschummel, G. Lenz, U.E. Höpken, A. Rehm

Development of methodology: M. Gloger, L. Menzel, K. Gerlach, T. Kammertöns, M. Zschummel, G. Lenz, H. Gerhardt, A. Rehm

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): M. Gloger, L. Menzel, A.-C. Vion, I. Anagnostopoulos, K. Gerlach, T. Hehlgans, U.E. Höpken, A. Rehm

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): M. Gloger, L. Menzel, M. Grau, A.-C. Vion, M. Zapukhlyak, M. Zschummel, G. Lenz, H. Gerhardt, U.E. Höpken, A. Rehm

Writing, review, and/or revision of the manuscript: M. Gloger, L. Menzel, M. Grau, A.-C. Vion, I. Anagnostopoulos, M. Zapukhlyak, T. Kammertöns, G. Lenz, H. Gerhardt, U.E. Höpken, A. Rehm

Administrative, technical, or material support (i.e., reporting or organizing data, constructing databases): I. Anagnostopoulos, A. Rehm

Study supervision: U.E. Höpken, A. Rehm

The study was funded by a grant from Deutsche Krebshilfe to A. Rehm and U.E. Höpken (Grant No. 111918). We thank Daniel Besser (MDC) for help in bioinformatic analysis; Christian Friese (Charité, Berlin) for expert technical assistance; Anje Sporbert and Matthias Richter for microscopic image analysis (Advanced Light Microscopy Platform; MDC); Ralf Adams (MPI, Münster, Germany); and Taija Mäkinen (Uppsala University, Sweden) for Cre-deleter mouse strains.

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.