B-cell non-Hodgkin lymphomas (B-NHL) are a heterogeneous group of hematological malignancies that emerge from different stages of normal mature B-cell differentiation (1). Lymphoma evolution and treatment are complicated by a high prevalence of relapses (2) primarily due to the ability of malignant B cells to interact with protective lymph node (LN) and bone marrow (BM) microenvironments (3–5). In agreement, several studies have correlated BM involvement with worsened prognosis and impaired chemotherapeutic response in B-cell lymphomas (6–8). This review delves into the current knowledge of the BM stromal cell modifications induced by the protumoral niche establishment in B-NHL with a specific focus on diffuse large B-cell lymphoma (DLBCL), follicular lymphoma (FL), and chronic lymphocytic leukemia (CLL). Interestingly, these three B-NHL subtypes displayed various BM involvement, with 11%-34% of DLBCL (9, 10), 70%-80% of FL (11), and virtually all CLL cases showing BM infiltration at diagnosis ( Table 1 ). Moreover, this review highlights the newly described role of extracellular vesicles (EVs) in the seeding of the BM niche. EVs are released during homeostasis and cell activation, with pleiotropic effects on signaling between cells. EV cargos are enriched in nucleic acids, proteins, and lipids. Briefly, the International Society of Extracellular Vesicles had classified EVs into three main groups: i) exosomes, the small vesicles with diameters ≤100-150 nm that are formed inside multivesicular bodies; ii) microvesicles, medium-size vesicles of plasma membrane origin with diameters of up to 1000 nm; and iii) apoptotic bodies, the large vesicles with diameters > 1000 nm that are produced by apoptotic cells (12). Excellent reviews on the biomolecular and functional characteristics of EVs as well as on the techniques used for EV isolation and characterization have recently been published (13, 14).

DLBCL is the most common aggressive B-NHL and accounts for approximatively 24% of new NHL cases (15). Gene expression analysis and study of genomic alterations have identified distinct genetic subtypes in DLBCL, reflecting differential pathogenesis, and associated with distinct clinical behavior (16–19). Interestingly, recent studies have highlighted the impact of tumor microenvironment (TME) heterogeneity on tumor B-cell biological features and on DLBCL patient outcome (20, 21).

FL accounts for about 20% of adult lymphoma and is an indolent disease characterized by prolonged periods of remissions preceding relapses and ultimate transformation into DLBCL in about 30% of cases. The genetic hallmark of FL is the t (14, 18) translocation occurring during the V(D)J recombination of immunoglobulin genes in the BM. The resulting deregulation of BCL2 provides a selective survival advantage to B cells during the germinal center (GC) reaction, triggering illegitimate recirculation of t (14, 18)^pos post-GC B cells detectable in most healthy individuals. Iterative (re)entry of these FL precursor cells inside GC favors accumulation of additional genetic alterations sometimes converging towards overt FL (22). Importantly, FL is the paradigm of a neoplasia fully dependent on a complex microenvironment network that coevolves with tumor B cells to create a tumor supportive niche in both LN and BM (23, 24).

CLL is the most common hematologic malignancy in adults in Western countries. CLL is preceded by a stage of monoclonal B-cell lymphocytosis and is characterized by the accumulation of mature clonal B cells resistant to apoptosis in the blood, BM, and lymphoid organs. Patients with CLL have a heterogeneous clinical course with some never needing treatment, while others require treatment immediately after diagnosis or during illness due to a more symptomatic and unfavorable clinical course. In typical CLL cases, the tumor B cell clone exhibits an abnormal expression of markers like CD5, CXCR4, and ZAP-70, that are used to stratify the disease in conjunction with the mutational status of the BCR reflecting different cell of origin (25, 26). Despite fully disseminated presentation, TME provides crucial survival signals to malignant CLL cells within the proliferation centers of LN and BM (27).

In these three mature B-cell neoplasms, specialized tumor niches support survival, proliferation, and drug resistance of tumor B cells. These highly heterogeneous niches include defective tumor immunity, due to altered recruitment and cell exhaustion of cytotoxic cells, to the amplification of immunosuppressive cells, or to immune escape mechanisms developed by tumor B cell themselves, hampering tumor recognition, immune synapse formation, or anti-tumor cell activation (23, 24, 27). Conversely, fully functional tumor permissive cells, including CD4^pos T cell, myeloid cell, and stromal cell subsets, could be found. The relationship between LN and BM protumoral niches and how the similarities and differences between these microenvironments could impact malignant B-cell features remains elusive. In FL, malignant B cells found in the BM are characterized by a lower cytological grade, a decreased proliferation, and a reduced CD10 expression compared with LN FL B cells (28). Moreover, their gene expression profile reflects their reduced proliferation and active metabolism (29). Finally, somatic hypermutation analysis and targeted deep sequencing demonstrate that different FL B-cell subclones could be detected within LN versus BM, and suggested that FL originates in the LN and infiltrates BM early in the course of the disease, allowing further accumulation of BM-specific mutations (28, 30, 31). Besides the exact cell composition and supportive signals provided by BM niches, a major issue remains to establish how these niches evolve during tumor development, from the pre-tumoral stage to overt lymphoma, during remissions and relapses.

BM constitutes the primary site for the maintenance and differentiation of hematopoietic stem cells (HSCs) and for B-cell lymphopoiesis. Different stromal cell niches dynamically control these processes. Seminal papers have recently proposed a molecular atlas of the BM stromal cells at the single cell resolution, including osteoblasts, perivascular cells, endothelial cells, and mesenchymal stromal cells, providing clues on how various stromal cell subtypes could interact with HSCs and differentiating B-cell subsets (32–34). In the context of B-NHL, dynamic interactions between BM stromal cells and tumor B cells have been described to play a key role in converting the BM TME into a tumor supportive niche (34–36). DLBCL, FL, and CLL display distinct patterns of BM infiltration ( Table 1 ). DLBCL show a mixed pattern of BM involvement that can potentially range from localized focal infiltrates to complete disruption of BM by lymphoma cell proliferation (37). In contrast, FL infiltration is primarily localized to the paratrabecular regions as nodular aggregates admixed with lymphoid-like TME (38). In CLL several BM infiltration patterns can be found including mixed nodular-interstitial, interstitial, and diffuse (39). In each cases, stromal cells exhibiting specific functional phenotype support recruitment, survival, and proliferation of tumor B cells, mimicking the cancer-associated fibroblasts (CAFs) described in solid cancers.

FL tumor B cells could directly contribute to the commitment of BM stromal precursors into an FRC-like phenotype overexpressing CCL2 and IL-8 through TNF-dependent mechanisms (3, 63, 71). Moreover, even if they produce less LT than normal centrocytes, the large number of GC-like B cells ectopically found in invaded FL BM probably contributes to a local overproduction of LT that synergizes with TNF for the induction of lymphoid stroma commitment. However, surrounding non-malignant cells could also participate in the polarization of FL-CAFs. Neutrophils, recruited by IL-8-producing BM FL stromal cells, could in turn contribute to their differentiation into FRC-like cells through activation of the NFκB pathway (63). In addition, LN FL-Tfh overexpress IL-4 which induces a Transglutaminase^hiPodoplanin^lowCD106^hiCXCL12^hi phenotype on human stromal cell precursors. FL-Tfh also produce high amounts of TNF and LT, which sensitize stromal cell precursors to the effect of IL-4, notably through increased expression of the STAT6 signaling molecule (40). Even if fully mature Tfh have not been detected within FL BM, IL-4 and CXCL12 have been shown to be correlated in invaded FL BM (40). Finally, some of the recurrent genetic alterations in FL regulate the re-education of the tumor niche by tumor B cells. In particular, the gain-of-function mutations of the histone methyltransferase EZH2, which occurred early in 20% to 30% of FL, are proposed to uncouple GC B cells from the critical Tfh checkpoint whereas switching them to FDC dependency (99). EZH2-mutated GC B cells downregulate many genes linked to Tfh signaling, fail to engage Tfh, thus limiting recycling toward the dark zone of GC, and survive in the light zone as proliferating centrocytes overexpressing LT, TNF, and BAFFR, all involved in GC B-cell/FDC crosstalk. HVEM loss-of-function mutations detected in about 40% of patients with FL have been associated, in a murine model of FL and in FL patients, with an amplification of Tfh producing large amounts of IL4, TNF, and LT, and able to activate FL-CAF within LN (100). No study had currently evaluated how these genetic events could impact FL TME co-evolution within BM. Even if such data are essentially lacking in the context of DLBCL, some recurrent genetic alterations have been recently associated with a specific TME pattern, with some of them related to overexpression of genes associated with GC-like stroma or extracellular matrix/FRC/CAF genes (21).

Finally, LT produced by CLL cells is involved in the polarization and/or in situ generation of the tumor stromal network and the secretion of CXCL13, IL-6, and IL-8 (74, 79). Moreover, the leukemic clone produces retinoic acid in the stromal microenvironment which contributes, at least in part, to the CXCL13 induction (101).

In addition to the factors described above, tumor derived EVs seem to be involved in the communication between tumor cells and their TME, in particular CAFs. Such mechanism could play a central role in triggering initial induction of tumor-supportive niche within distant sites, including BM.

To date no study has explored the putative involvement of EVs in the induction of a BM lymphoma stromal niche in the context of DLBCL. Moreover, only few studies have investigated the involvement of EVs in the pathophysiology of FL ( Figure 1 ). Recently, FL-derived EVs were shown to modulate the gene expression profile of BM stromal cells, triggering an upregulation of HSC niche factors including CXCL12, angiopoetin-1, KITLG, or IL-7, and increasing the capacity of stromal cells to interact specifically with BM FL B cells and support their survival and their quiescent phenotype (29). Interestingly, the phenotype of EV-treated stromal cells is quite different from that obtained under treatment by TNF/LT or coculture with FL B cells supporting a role of EVs in the activation of BM stromal cells before BM seeding by malignant B cells. In fact, the level of CXCL12 is increased in non-involved BM plasma suggesting that FL EVs could shape the BM stromal niche before BM infiltration by tumor cells or at distance from this BM infiltration (unpublished data). In the same way, the analysis of the gene expression profile of BM stromal cells highlights a continuum ranging from healthy donor BM stromal cells, to stromal cells obtained from FL patients without BM involvement, and finally from FL-invaded BM (29). Altogether these data suggest that EVs could contribute to CXCL12 upregulation in the absence of direct contact with malignant B cells and could then synergize with IL-4 produced by infiltrating T cells admixed with FL B cells to further enhance local CXCL12 production. Interestingly, BM stromal cells activation by FL-derived EVs was shown to rely on TGF-β dependent pathways something that is reminiscent of the role of TGF-β in the B-cell/stromal cell crosstalk within FL LN (42). How TGF-β and STAT6 pathways could synergize for the acquisition of FL CAF phenotype within FL BM remains to be explored.

Bidirectional crosstalk has also been reported between CLL B-cells and their surrounding stroma via EVs ( Figure 1 ). CLL B cells release large amounts of exosomes that show strong expression of CD37, CD9, and CD63. Ibrutinib, a Btk inhibitor, significantly reduces the amount of plasma exosomes in CLL patients. Likewise, in vitro treatment of CLL cells with Idelalisib (a PI3K inhibitor) decreases exosome secretion, something that is not observed during treatment with fludarabine (102). This result highlights the role of the BCR-PI3K pathway in controlling exosome secretion in CLL. Besides BCR itself, CLL supportive TME produces BAFF, APRIL, CD31, and plexin B1 that all protect CLL cells from spontaneous apoptosis by synergizing with BCR signaling (44, 103) and could influence EV secretion. The comparison of the mRNA content of EVs produced by B cells from healthy donors versus patients with CLL, and stimulated or not through the TLR9 pathway, shows enrichment for the kinases of the BCR pathway, LYN, SYK, MAPK1, MAPK2, and the anti-apoptotic proteins BCL2 and BCL3 in CLL-derived EVs. These EVs released by tumor B cells transfer their mRNA content to non-malignant cells in the TME (104). Microvesicles derived from malignant CLL cells and detected in peripheral blood also deliver the receptor tyrosine kinase Axl into BM stromal cells leading to the activation of a AKT/mTOR/p70S6K/HIF-1α axis resulting in an increase in VEGF synthesis (105). This increase in VEGF is associated with an increased neovascularization in medullary (106) and extramedullary tissues, as well as a paracrine pro-survival stimulation of tumor B cells (107). The miRNA content of CLL B cell-derived exosomes is strongly enriched in miR-21, miR-155, miR-146a, miR-148a, and let-7g (108). BM stromal cells treated in vitro with these CLL exosomes acquire an inflammatory pro-tumoral phenotype, while endothelial cells increase their capacity for angiogenesis (108). These effects are consistent with what is known about the effect of miR-21 and miR-146a in the transition from normal fibroblast to CAFs (109–112). Indeed, CLL miR-146a^pos exosomes induce the transition of BM stromal precursors into CAFs showing over-expression of α-SMA and FAP (113). In addition, CLL exosomes show specific enrichment in miR-202-3p, able to decrease expression of Sufu (a component of the hedgehog pathway) in stromal cells and to trigger stromal cell proliferation (114). Finally, EVs isolated from cultures of CLL BM stromal cells induce a significant decrease in spontaneous apoptosis of tumor B cells and an increase in their chemoresistance to several drugs, including fludarabine, ibrutinib, idelalisib, and venetoclax. In addition, these EVs induce changes in the gene expression profile of CLL cells mimicking the transcriptomic signatures obtained after BCR stimulation (115).

Analyzing the deregulation of extracellular proteins or miRNAs in the blood and tumor niches of patients during B cell tumorigenesis is a reliable tool for the identification of new tumor-targeted therapies. For example, the detailed mode of action of the CD30 antibody-drug conjugate Brentuximab vedotin in DLBCL is not well understood since the clinical outcome seems to be partially independent of the CD30 expression on the tumor cells. However, as CD30^pos bystander cells are enriched in the tumor tissue in many cases of DLBCL, CD30 might be released within TME-derived EVs. Thus a model was proposed in which even in the absence of CD30 on the tumor cells, EVs can transport the targeting protein from cells of the TME to tumor cells (116). This model would explain the clinical efficacy of Brentuximab vedotin also in cases of lack of the targeting antigen on tumor cells. In the same way, DLBCL EVs carrying miR-125b-5p can reduce tumor sensitivity to rituximab by inhibiting TNFAIP3 expression and reducing CD20 expression (117). Whether the miR-125b-5p/TNFAIP3 axis can be used as a therapeutic approach for increasing DLBCL sensitivity to anti-CD20 antibodies requires further investigations.

EVs released by B cell could carry CD39 and CD73, two surface molecules known to hydrolyze ATP released by dying cancer cells into adenosine that hijacks CD8 T cell immune activity by binding the A2A adenosine receptors (118). One could speculate that B-cell-derived EVs may have a similar effect. The decrease of B-cell-derived EVs bearing CD73 and CD39 can be achieved by deregulating the docking protein RAB27A (118). This could be performed using an inactivated Epstein–Barr virus carrying siRNA, but it is also possible to generate EVs derived from cell lines producing RAB27A siRNA and to specifically deliver it to tumor cells.

Ultimately, thanks to their molecular structure mimicking the plasma membrane of the cells and their capability to reverse their cargo into target cells, exosomes could be shaped and filled of drug molecules, acting as drug-delivery systems. In fact, cancer vaccine clinical trials relying on the administration of exosomes produced by dendritic cells (Dexosomes), exploited to shuttle antigenic determinants of immune response, were conducted to immunize patients in the context of solid tumors (119–121). In the same way, systemic administrations of TNF-Related Apoptosis-Inducing Ligand (TRAIL)-armed exosomes have shown a great anti-tumor effectiveness against FL/DLBCL cell lines both in vitro and in a mouse model (122).

Despite very interesting recent data highlighting BM as a survival niche for lymphoma B cells, numerous controversies remain open on the role of the BM versus LN niches during the early step of lymphomagenesis or at the stage of post-treatment minimal residual disease that could generate relapse. In FL, both pre-tumoral B cells and early committed precursor cells, that will give rise to overt FL, have been shown to be enriched in BM (22). However, transformation events required iterative passages throughout the GC making it difficult to define precisely whether BM is a primary or a secondary tumor niche. The influence of tumor genetics or patient features on the capacity of tumor B cells to home and develop into BM remains completely unexplored. A major limitation for all BM-dedicated studies is the limited availability of good quality samples to perform phenotypic, transcriptomic, and functional studies and the lack of iterative sampling allowing evaluation of the impact of disease evolution or therapeutic strategies. BM aspirates are scarce and do probably not include the whole diversity of tumor/TME components, in particular stromal cells. Moreover, fixed BM biopsies are very difficult to exploit for spatial transcriptomics and even multiplex immunohistofluorescence approaches. Such technical issue hampers a precise evaluation of spatial heterogeneity in B-cell lymphomas integrating BM as a key tumor site.

Altogether, many evidence support the clinical interest of targeting the crosstalk between BM stromal cells and malignant B cells, through the inhibition of stroma-derived B-cell growth factors, the mobilization of clonal B cells outside their supportive BM niche, or the reprogramming of tumor-supportive stromal cells. Identifying the best therapeutic options, and how to combine them with tumor-targeting drugs or immunotherapy approaches will be the major challenge in the field.

ED wrote the paper, SM reviewed the paper, and KT supervised and wrote the paper. All authors contributed to the article and approved the submitted version.

This work was supported by research grants from Fondation ARC (PGA1 RF20170205386) and the Institut National du cancer (INCA AAP PNP19-009).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.