Tumors represent a complex microenvironment that not only includes tumor cells but also stromal cells, blood vessels, and different immune cell populations, reflecting the capacity of the immune system to sense tumor cells (1). The major role played by the immune system in tumor surveillance and control has been extensively documented. Since the discovery of T cells, B cells, and antibodies specific for tumor antigens (2–6), several clinical studies have clearly demonstrated that a high density of T-cell or B-cell subsets within the tumor microenvironment is associated with increased patient survival (1, 7, 8), such as in colorectal cancer (9), primary cutaneous melanoma (10), breast cancer (11–14), non-small cell lung cancer (NSCLC) (15, 16), head and neck cancer (17), and ovarian cancer (18–20).

Regarding the development of adaptive anti-tumor immune responses (but it is also true for any other kind of disorders), the dogma was that infiltrating anti-tumor immune cells are educated to recognize tumor antigens and proliferate at distance of the tumor site, in conventional secondary lymphoid organs (SLOs), and are then recruited to the tumor microenvironment to exert their anti-tumor activity. SLOs are highly specialized tissues that arise at predetermined locations during embryogenesis and persist all over the life. They include lymph nodes (LNs), spleen and mucosal-associated lymphoid tissues (MALTs), namely Peyer’s patches, tonsils, and nasal-associated lymphoid tissues (NALTs) or bronchus-associated lymphoid tissues (BALTs). However, it is now established that immune responses can also take place at distance of SLOs, in tertiary lymphoid structures (TLS), also called tertiary lymphoid organs (TLOs) or, in the case of the lung, (T)i-BALTs, for (tumor-) induced BALTs (21–23). As opposed to SLOs, TLS are defined as transient ectopic lymphoid organizations that can develop after birth in non-lymphoid tissues, in situations of chronic inflammation. This phenomenon is usually termed lymphoid “neogenesis” by opposition to lymphoid organogenesis, which refers to SLO development (24, 25). TLS have already been observed in several infectious diseases, graft rejection, autoimmune disorders [including rheumatoid arthritis (RA) and multiple sclerosis], allergy, and cancer (25–27). They have also been described in different lung inflammatory diseases, such as fibrosis, pneumonia, hypersensitivity pneumonitis, diffuse pan-bronchiolitis, and tobacco-induced inflammation (28), and in the adventitia of human atherosclerotic arteries (29–31). It has been proposed that the presence of TLS can be observed in every situation when the inflammatory response initiated is unable to eradicate the associated antigen (32). This inability might be the result of an efficient escape of a pathogen antigen to immune surveillance, and/or of a permanent replenishment of a self or tumor antigen by the tissue. The difficulty in eradicating antigens would lead to a chronic inflammation, characterized by a sustained infiltration of immune cells, and to the development of TLS as sites for the generation of a local more efficient immune response. Because these structures turn off after the resolution of the inflammation, they have been proposed as a genuine inflammation marker (27). In autoimmune diseases and transplantation, the presence of TLS is respectively associated with disease exacerbation, poor outcome (33), chronic rejection, and anti-graft responses (34–37). On the opposite, T- and B-cell responses initiated within TLS during infections are associated with pathogen clearance, increased survival, and reduced morbidity (21). Finally, an increasing number of studies demonstrated over the past decades that the presence of TLS in the tumor microenvironment is associated with anti-tumor immune responses and prolonged patient survival (22, 27, 38–44). Yet, the events that influence immune cell infiltration and TLS formation are still incompletely understood.

In this review, we discuss the factors favoring the recruitment of immune cells and their organization in TLS by comparison with what is known about conventional SLOs. We detail the role played by TLS B cells in anti-tumor immunity, both as tumor antigen-specific antibody-secreting cells and antigen-presenting cells (APCs). Finally, we raise the question of the capacity of chemotherapeutic agents and/or new immunotherapeutic drugs to induce the development of TLS within tumors, and how to take advantage of what we know now from these structures to develop new therapeutic tools.

It has been clearly established that tumor-infiltrating B cells, and in particular cancer-associated TLS B cells, contribute to tumor prognosis and are able to produce antibodies locally. Thus, the important question to address is what these antibodies do, what their specificity is, and whether produced antibodies may contribute to the clinical outcome. In the 1980s, the knowledge that tumors can elicit spontaneous humoral responses spurred the development of autologous typing to identify novel molecular targets (129, 130). One of the most prolific methods to discover tumor-associated antigens was serological expression cloning (SEREX), which originally used serum of a cancer patient to screen for targets expressed in the autologous tumor after expression cloning (131). This unbiased method led to the discovery of a series of novel immunogenic targets, many of which were validated to be only eliciting antibody responses in the context of tumors (132). NY-ESO-1, discovered by SEREX, emerged as a model human tumor antigen, with restricted expression in normal tissues to germline cells, but widespread aberrant expression in various tumor types, and therefore named a cancer-testis antigen (133). Members of this family include MAGE, LAGE, XAGE, SSX, and other antigens, all of which are intracellular molecules, with cytoplasmic and/or nuclear expression. These antigens are controlled by epigenetic mechanisms and can be induced by hypomethylation or deacetylation of promoter sequences (134). It is thought that as tumors grow, some cells may be releasing antigens and initiating B-cell responses, and it is well established that detection of antibodies against these tumor antigens is associated with tumor antigen expression, even though there is no evidence that resulting antibodies have a direct functional role against tumors (135). Not much is known about what initiate immunogenicity, and whether it happens at the tumor site from local TLS, or whether more classic SLOs are responsible, after APCs carrying tumor extracts migrate there. A positive correlation has been described between size of LN B-cell areas and tumor-infiltrating B cells in head and neck cancer patients treated with a mixture of cytokines (136), suggesting that an interplay may be happening across SLOs and local tumor environment. To address the question of the structure initiating humoral responses, one would need a dynamic follow-up of antigen and uptake, and a better characterization of TLS versus LN versus circulating B cells for their specificity and chemokine receptor expression.

While B cells at the tumor site may recognize tumor antigens, it is important to consider their role and potential clinical benefit in a dynamic fashion as tumors evolve over time. Expression of most cancer-testis antigens tends to increase with stage and grade of disease, as shown for melanoma, lung, myeloma, and ovarian tumors among others, thereby highlighting more aggressive, poorly differentiated tumors with usually worse outcome (137), but that also become immunogenic as a result of new antigenicity. A study has found that NY-ESO-1 protein expression in primary breast carcinoma and metastases correlates with brisk plasmacytic B-cell infiltration (138). It is likely however that by the time, the tumor has reached a large enough size to generate B-cell response, the resulting antibodies are insufficient to affect the disease. Recently, tumor-infiltrating B cells were found increased in prostate cancer tissue with high intra-tumoral density of B cells related to higher stage and potential for recurrence or progression (139). In addition, B cells from ascites of ovarian cancer are able to produce antibodies to known tumor antigens, though the presence of B cells in ascites has been shown to increase with stage and to correlate with poor prognosis, but to decrease by chemotherapeutic treatment (140).

It is therefore important to view antibody responses as diverse repertoires with polyfunctionality, not all of it associated with better prognosis. Historically, circulating antibodies to tumor-specific antigens were seen as a sign that cellular immunity was skewed in favor of humoral rather than cellular responses, while CTLs were preferred for their direct anti-tumor capacity (103, 141). Mutated TP53 was one of the first tumor-associated antigens to be identified serologically (142), but circulating CTL responses against TP53 appear to be rare (143). Despite abundant literature emphasizing the split between B cell and Th2-mediated immunity versus cytotoxic CD8⁺ T cell and Th1 immunity, this dichotomy may not be so clear for all tumor-associated antigens. In particular, NY-ESO-1, one of the most spontaneously immunogenic antigens from the cancer-testis family, is an example for which presence of specific circulating antibodies typically implies concomitant presence of circulating effector-like CD4⁺ and CD8⁺ T cells, in an integrated fashion (144, 145). Therefore, antibodies have been proposed as a surrogate for T-cell immunity to NY-ESO-1. We showed that patients seropositive for NY-ESO-1 have NY-ESO-1-specific T cells accumulating at the tumor site in ovarian cancer patients, though their activity was blunted by expression of co-inhibitory molecules CTLA-4, PD-1, and LAG-3 (146). This may explain the lack of anti-tumor effect, and therapeutic strategies to reverse such responses will be discussed later. Nevertheless, these results suggest that seropositivity could be a surrogate for the presence of tumor-specific T cells infiltrating tumors.

Serum reactivities against tumor antigens have been described in many cancer types, such as anti-NY-ESO-1 antibodies in sarcoma (147) and primary breast cancer (148), anti-XAGE-1b antibodies in NSCLC (3), anti-NY-ESO-1, anti-LAGE-1, and anti-P53 antibodies in ovarian cancer (18, 149). Very recently, Ohue et al. showed that cancer patients with serum antibody-reactivity against the tumor antigen XAGE1 had prolonged overall survival as compared with seronegative patients (3). Similarly, antibodies to MUC1 appeared to confer better survival in various tumors, while p53 antibodies did the opposite (150). Circulating antibody responses to NY-ESO-1, SOX2, Ubiquilin, and others have been proposed as biomarkers for early detection of NSCLC (151, 152), though few have been validated to date (153). Presence of circulating autoantibodies to tumor antigens has also been well described in paraneoplastic syndrome, where specificities shared between tumors and the nervous system abruptly lead to autoimmune pathology, albeit with some anti-tumor benefit (154).

Some very recent studies have provided evidence that tumor-specific humoral responses can be generated in situ, within TLS. To characterize the specificity of local B cells at the tumor site, we performed ex vivo cultures of NSCLC-infiltrating B cells and showed that they were able to produce high levels of IgG and IgA directed against tumor antigens of the cancer-testis antigen family, such as LAGE-1, MAGE antigens, and NY-ESO-1, but also P53. The reactivity we observed was polyclonal, as a same patient could display reactivities against several different tumor antigens (39). The presence of IgA at the tumor site is intriguing, as it differs from the usual IgG detected in the periphery, and detailed repertoire analyses are still needed to distinguish potential differences locally and systemically. More evidence of antigen-specific local B cells was found in ductal breast carcinoma where infiltrating B cells from TLS-like structures were microdissected and showed polyclonality and high mutation rates, with some dominant clones shared between various clusters of B cells suggesting affinity maturation in situ and seeding from site to site (155). Finally, Maletzki et al. observed in colorectal cancer that tumor-infiltrating B cells produced IgG able to bind to the cell surface of several tumor cell lines (156).

Once again though, mere presence of antibodies may not necessarily be sufficient to predict outcome, as not all B-cell responses are of equal quality. In lung cancer, a recent report indicated increased circulating regulatory B cells [Bregs, reviewed in Ref. (157)] but decreased Tregs compared to healthy donors, highlighting a novel population of B cells to consider. It is likely that the context of the original antigen stimulation (viral versus self, chronically versus acutely inflamed) will drive the differentiation of the B-cell repertoire, and this opens up strategies for manipulation with immunotherapies.

It is assumed that the presence of TLS within tumors (except in renal cell carcinoma) is always associated with prolonged survival (27). Thus, a big challenge in the near future will be to develop new therapeutic tools, or to determine therapeutic combinations, able to favor the formation of such lymphoid structures in cancer patients.

Several studies have already reported the development of TLS as a consequence of antigen challenge during anti-tumor vaccination protocols. Lutz et al. for example observed the induction of TLS in pancreatic ductal adenocarcinoma (PDAC) after vaccination with GM-CSF-secreting pancreatic tumor vaccine (GVAX), a granulocyte-macrophage colony-stimulating factor (GM-CSF)-secreting, allogeneic PDAC vaccine (158). While PDAC is classically considered as a non-immunogenic neoplasm, tumors from patients vaccinated with GVAX presented organized and functional TLS as soon as 2 weeks following GVAX treatment. These TLS displayed all features of an ongoing immune response, including GC B-cell proliferation, T-cell activation, and Th1 transcription factor T-bet expression, associated with the induction of IFNγ-expressing effector T-cell responses. The number of TLS detected within PDAC tumors tended to increase when patients were treated with GVAX combined to the Treg-depleting chemotherapy cyclophosphamide, in accordance with studies showing that Tregs can suppress the development of TLS (88, 98, 99, 159). More interestingly, a downregulation of genes involved in the Treg pathway and a concomitant upregulation of genes involved in the Th17 pathway was detected in TLS of patients who survived longer and had enhanced immune responses. These results are in accordance with studies claiming that, as opposed to Tregs, Th17 cells and Th17 cytokines promote TLS development (92, 93, 96, 97). Similarly, Maldonado et al. observed the formation of TLS in patients with high-grade cervical intraepithelial neoplasias (CIN2/3) who received intramuscular therapeutic vaccination targeting HPV16 E6/E7 antigens (160). These TLS were associated with HEVs, immune cell activation and Th1 polarization signatures, and TCR clonality. In both studies, it seems that vaccination-induced antigen-specific T cells generated systemically were able to massively infiltrate the tumor microenvironment, and to initiate a local inflammatory reaction, further favoring the recruitment of immune cells and TLS neogenesis.

Besides vaccination protocols, which aim to increase humoral and cellular immune responses against particular tumor antigens, several studies demonstrated that some chemotherapeutic drugs have the capacity to enhance tumor immunogenicity and recognition by the immune system. Kroemer and Zitvogel groups notably have extensively documented how anthracyclines and oxaliplatin are able to trigger an immunogenic form of tumor cell death, and thus to promote anti-tumor immune responses, through three molecular mechanisms (161): first, the translocation of calreticulin on the tumor cell surface, which results in tumor engulfment by DCs; and secondly, the release by dying tumor cells of HMGB1 protein and ATP, two danger signals that bind TLR4 and P2X7 on DCs, respectively, and increase their ability to present tumor antigens and to activate anti-tumor CD8⁺ T-cell responses (162–164). Similarly, some DNA demethylating agents showed capacities to increase tumor immunogenicity, by favoring the expression of cancer-testis antigens in cancer cells (165). Ayyoub et al. for example demonstrated that DNA demethylation induced by 5-aza-2′-deoxycytidine (5-AZA-CdR) increased or induced mRNA expression of several cancer-testis antigens in sarcoma cell lines, including MAGE-A10, SSX, NY-ESO-1, and LAGE-1, resulting in increased tumor cell recognition by CTLs (147). Considering the potential role played by DCs and B cells in TLS induction and/or maintenance, we can speculate that chemotherapeutic drugs favoring DC and/or B-cell activation may also have a beneficial effect on the development of TLS within the tumor.

Other developed strategies aim to directly induce the formation of TLS within tumor tissue, by targeting at the tumor site signals or chemokines known as favoring immune cell recruitment and/or lymphoid neogenesis. Schrama et al. used the property of LTα to promote TLS development in a mouse model of melanoma (166). They developed an anti-disialoganglioside GD2 specific antibody-LTα fusion protein, and demonstrated that this fusion protein was able to favor subcutaneous tumor and established pulmonary metastasis eradication, and to prolong survival. Twenty-eight days after treatment, they could observe the presence of T cells and B cells organized in lymphoid-like patterns within tumors, characterized by the presence of MHC class II⁺ cells resembling DCs, the presence of CD62L⁺ cells in close contact with PNAd⁺ HEVs. Lymphoid neogenesis in the treated mice was associated with anti-tumor T-cell clonal expansion. Similarly, Liang et al. used the properties of the lymphoid chemokine CCL21 to favor the recruitment of immune cells at the tumor site (167). In a mouse model of hepatocellular carcinoma, they showed that the transduction of tumor cells with a recombinant adeno-associated virus (rAAV) expressing CCL21, or intra-tumoral injections of rAAV-CCL21, allowed extensive infiltration of the tumor site by CD11c⁺ DCs and activated CD3⁺ CD69⁺ T cells. Finally, Irvine et al. discussed the possibility of using tissue engineering, and matrix or scaffold-based approaches in order to create an implantable microenvironment that would support the recruitment of immune cell and/or favor lymphoid neogenesis through, for instance, the slow release of chemoattractants (168).

We can hypothesize that strategies aiming at increasing or re-activating exhausted T-cell responses, by targeting immunoregulators such as CTLA-4 or PD-1, may also have bystander effects on TLS development. Beyond immune checkpoint inhibitors, next generation immunomodulators may also involve costimulatory agonists that can target both T and B cells, including 4-1BB, CD27, HVEM, and CD40 (169). Cross-linking capacity of B cells was shown to be important for the therapeutic effect of immunomodulatory antibodies such as anti-CD40 (170). These novel strategies will need to be studied in detail for their role on local immune infiltration and organization, through the design of presurgical, neoadjuvant clinical studies. Another immunomodulatory approach, beside systemic administration of antibodies targeting immune pathways, may be to proceed with intra-tumoral injections of pathogen-associated molecular patterns (PAMPs) or cytokines, to change the inflammatory environment of the tumor. BTLA was described as an inhibitor of some B-cell functions that is upregulated in the context of immunotherapies including CpG administration, but could be actionable with blocking of its interaction with natural ligand HVEM (171). Intra-tumoral injection of IL-12 was also shown to activate B cells and lead to the polarization of immunoglobulin secretion to Th1-associated IgG1, along with IFNγ production by B cells (17).

The detection of a humoral anti-tumor immune response initiated within TLS and its characterization open the way to the development of new vaccine and antibody-based therapeutic strategies. As discussed above, reactivity against tumor antigens belonging to the cancer-testis antigen family is a feature shared by several cancer types, making these antigens as potential targets in future protocols of immunotherapy. Despite their intracellular expression, vaccines or monoclonal antibodies directed against some of these antigens have already been validated in preclinical models. Noguchi et al. for example demonstrated in mice the anti-tumor effect of an anti-NY-ESO-1 antibody when associated with chemotherapy (172). In this study, they observed not only the generation of an anti-NY-ESO-1 CD8⁺ T-cell immune response, but also the differentiation of other antigen-specific CD8⁺ T cells, thanks to antigen spreading and DC activation by anti-NY-ESO-1 immune complexes (173).

MAGE-A3 and NY-ESO-1 have entered into clinical trials for various cancers, and belong to the top 10 of the cancer antigens referenced in the “Prioritization of Cancer Antigens” project proposed by the National Cancer Institute (NCI) (174–177). Several studies have already reported the development of both humoral and cellular specific immune responses in NY-ESO-1-vaccinated cancer patients (144, 178). In one study, it was suggested that patients with the highest humoral responses following vaccination with NY-ESO-1 protein and CpG were more likely to have associated induction of CD8⁺ T cells (179).

Finally, as described above, NY-ESO-1 is known to elicit spontaneous immune responses though their clinical benefit is not clear, since the effector T cells at the tumor site may be dysfunctional due to immune checkpoint inhibition (146). We asked if some of the success of CTLA-4 blockade could be attributed to a change in immune responses against NY-ESO-1, since it is a frequent antigen in melanoma. We found that patients with pre-existing serological immunity to NY-ESO-1, coupled with detectable CD8⁺ T cells in the periphery, were nearly twice as likely to experience clinical benefit to ipilimumab compared to patients without NY-ESO-1 immunity (180). Circulating antibody titers, CD8⁺ T-cell repertoire, and polyfunctionality to NY-ESO-1 increased after treatment, but it is still unknown how ipilimumab affected local tumor B and T-cell responses. However, the clinical benefit suggests that having the tools to recognize NY-ESO-1 at baseline will make it more likely that immunomodulatory treatment will work.

Tertiary lymphoid structures in human tumors contribute to an environment conducive to the generation of B-cell diversity, with the help of T cells and DCs that leads to accumulation of tumor-specific immune responses inside or in proximity to the tumor. The presence of such lymphoid structures, including their B-cell component, underlines a population of patients that have already been primed to react against their tumors, albeit immunosurveillance has failed by the time the tumor is clinically detectable likely due to many counter-regulatory mechanisms. Nevertheless, TLS, along with T cells, mature DCs, and B cells, often predict eventual outcome in various diseases such as NSCLC, which suggests that having a primed tumor immune microenvironment, i.e., local immunocompetence, prior to treatment may facilitate various interventions and slow down progression. However, with progression also comes a series of counter-regulatory mechanisms that hamper the orchestrated anti-tumor immune responses, including the B-cell repertoire.

Many questions remain to be answered however, such as identifying whether immune responses are originally primed locally in TLS and become systemic, or conversely are generated in classic lymphoid organs and migrate to the disease site. Investigating differences in antigen cross-presentation, in antigen specificity locally and systemically, in quality and function between tumor B cells, LN B cells, and circulating B cells should help bring some answers.

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.