Insights into Tumor-Associated Tertiary Lymphoid Structures: Novel Targets for Antitumor Immunity and Cancer Immunotherapy

It is well appreciated that tumor immune infiltrates are prognostic indicators for patient survival and response to immunotherapies. Infiltrates enriched for CD8⁺ T cells have a positive prognostic value, whereas those enriched for myeloid cells have a negative prognostic value (1, 2). Given the general view that endogenous antibody responses contribute little to antitumor immunity, it is surprising that B lymphocytes also have a positive prognostic value (3). Also, CD8⁺ T cells disseminated throughout the tumor parenchyma have a stronger prognostic value than those confined to the perivascular space surrounding intratumoral blood vessels (4) or those aggregated immediately outside of the tumor mass (5). On the other hand, immune infiltrates organized into TA-TLS also have significant positive prognostic value. These structures have been extensively reviewed previously (6, 7) and we have incorporated the details of these reviews by reference. In this Cancer Immunology at a Crossroads, we summarize what is known about the composition, organization, function, and mechanisms underlying tumor-associated tertiary lymphoid structure (TA-TLS) formation, and highlight issues that remain to be understood to harness these structures for therapeutic purposes.

TA-TLS were initially described in melanoma and in non–small cell lung cancer (NSCLC), and are documented in a variety of primary and metastatic tumor types (6, 7). Histologic elements most frequently used to identify human TA-TLS include one or more of the following: tumor vessels expressing peripheral node addressin (PNAd), mature dendritic cells (DC) expressing lysosome-associated membrane glycoprotein (DC-LAMP), dense aggregates of T and/or B cells, follicular helper T cells (T_FH), and cells resembling follicular DCs (FDC; refs. 6, 7; Table 1). Most TA-TLS are organized “classically,” with distinct T-cell/DC and B-cell/FDC compartments (Table 1), and one or more of the homeostatic chemokines CCL19, CCL21, CXCL12, and CXCL13, which organize the secondary lymphoid organ (SLO) microarchitecture, are documented in TA-TLS by immunohistochemistry (IHC; refs. 6, 7; Fig. 1). Composite gene signatures are used for the detection of TA-TLS (Table 1). Expression of the plasma cell–specific marker B-cell maturation antigen (BCMA) is associated with the presence of TA-TLS in ovarian cancer (8). A more comprehensive 19-gene signature identifying B cells and T_H1 T cells is associated with the presence of TA-TLS in gastric cancer (9), and an eight-gene signature identifying T_FH cells is predictive of the presence of TA-TLS in breast cancer (10). A 12-chemokine gene signature is also predictive of the presence of TA-TLS in colorectal (11), melanoma (11), breast (12), and hepatocellular carcinoma (13). Finally, a nine-gene signature has been identified by comparing CD8⁺/CD20⁺ and CD8⁺/CD20^neg melanomas (14). Collectively, these components provide a baseline for identifying TA-TLS. However, most human studies have relied on only one or a small number of these markers (6). In addition, the criteria typically used are largely oriented toward elements that support antitumor immunity, although regulatory T cells (Treg) have occasionally been reported previously (15, 16). A particularly interesting study demonstrates distinct Treg-, T_H1-, and T_H17-biased profiles in TA-TLS associated with response or lack of response to a mesothelin vaccine (17). However, it is still unknown whether other immunosuppressive cells (myeloid-derived suppressor cells, some populations of innate lymphocytes or natural killer T cells) can be present in TA-TLS. Thus, there is likely to be significant unappreciated heterogeneity in TA-TLS cellular composition. Also, some reports have identified loose aggregates of lymphocytes as TA-TLS (13, 18), although they are typically tightly aggregated structures. Whether these are nascent or senescent TA-TLS, or something altogether different, remains unclear. We believe that the field as a whole should move to utilize a comprehensive set of markers to identify these structures to ensure that their heterogeneity, function, and prognostic value can be more thoroughly evaluated.

Despite the limited characterization in many studies, TA-TLS from different tumor types vary in cellular composition and organization. B cells with immature, naïve, activated, memory, and plasma cell phenotypes are evident to varying extents in different TA-TLS (6). In NSCLC, TA-TLS contain large numbers of mature DC-LAMP⁺ DCs (19, 20), but these are absent in those associated with lung metastatic renal cell carcinoma (21). T_FH cells are common features of breast cancer TA-TLS (22), whereas those associated with prostate (15) and lung metastatic colorectal cancer (16) contain large numbers of Tregs. While this could indicate that TA-TLS in different tumor types or anatomic locations contain different T-cell subpopulations, none of these studies evaluated both. The majority of studies have identified human TA-TLS as having a peritumoral location, whereas a smaller number have identified TA-TLS as intratumoral, usually in addition to peritumoral structures (Table 1; Fig. 1; ref. 6). However, in germ cell tumors (23), hepatocellular carcinoma (13), and lung metastatic renal cell carcinoma (24), TA-TLS are largely intratumoral, and exhibit a nonclassical organization lacking discrete T- and B-cell compartments. However, peritumoral TA-TLS are sometimes inside the tumor albeit near the tumor-invasive margin and sometimes fully outside the margin (25). This distinction may have important consequences for TA-TLS function. In addition, there is concern that the peritumoral TA-TLS identified in lymph node metastases may represent residual lymphoid follicles (14). In human melanoma, TLS frequently develop in metastatic lesions, but are largely absent from primary tumors, despite the presence of a PNAd⁺ vasculature (26). Similarly, TA-TLS are found in intraperitoneal, but not in subcutaneous murine tumors (27–29), and are observed frequently in primary breast tumors but largely absent in metastatic brain lesions (30). Thus, TA-TLS presence and structural organization are associated with tumor microenvironment and anatomic location, although the factors responsible, and the overall impact on TA-TLS functionality remains to be determined.

It has been suggested that TA-TLS serve as sites for sustained generation of in situ immune responses that are focused toward tumor antigens (6, 7), but evidence for this remains somewhat limited. We demonstrated that tumor vessels expressing PNAd, a hallmark of TA-TLS, supported infiltration of naïve T and B cells (27, 29), and TA-TLS in NSCLC contain large accumulations of naïve T and B cells (20, 31, 32). Thus, TA-TLS could promote a continual influx of naïve cells for sustaining immunity. Whereas TA-TLS vary in the number of mature DCs they contain, larger numbers of mature DCs are associated with larger numbers of T cells with a T_H1/cytotoxic immune profile (19, 20), suggesting active antigen presentation to T cells in TA-TLS. A linear relationship between the density of TA-TLS and the levels of intratumoral activated T and/or B cells has also been described previously (6, 7). While this observation may suggest that TA-TLS support tumor-infiltrating lymphocyte (TIL) development, our own work (29), described below, demonstrates that TIL support TA-TLS development.

Aggregation of tumor-associated B cells into a follicle-like structure is one of the most dramatic and defining features of TA-TLS. Although B cells can be positive or negative mediators of antitumor immunity (33), B cells in classically organized TA-TLS often express markers associated with germinal center activity (32, 34–36), and show higher degrees of clonal amplification, rearranged immunoglobulins, somatic hypermutations, and isotype switching than those in tumor parenchyma (26, 37), suggesting active antitumor humoral responses in these structures (Fig. 1). T_FH cells are also commonly found in the B-cell compartment of TA-TLS (38). These observations suggest that TA-TLS support the activation and differentiation of B cells into antibody-producing cells, and they are consistent with the idea that TA-TLS promote the in situ generation of tumor-specific antibody that augments antitumor immunity.

Despite these intriguing observations, there are a number of unaddressed issues that may limit the contribution of TA-TLS to producing effective antitumor immunity. Lymphatic vessels are rarely reported in TA-TLS, leading to uncertainties about tumor-antigen and cellular transport to these structures. Given the relatively small size of TA-TLS, it is unclear what fraction of the resident naïve T and B cells are tumor-antigen specific, and how frequently they turn over. Also, it is unclear whether DCs in TA-TLS come from tumor parenchyma, adjacent tissue, or blood-derived inflammatory monocytes. Similarly, it is unclear whether and how these cells acquire tumor antigen, and if DCs in TA-TLS are more mature than those in tumor parenchyma or tumor-draining SLO. We do not know how the effector and exhaustion marker profiles of T cells in TA-TLS compare with those in the surrounding tumor parenchyma and tumor-draining SLO. We also do not know how immunosuppressive elements in the tumor microenvironment, such as hypoxia, indoleamine 2, 3-dioxygenase, nitric oxide, arginase, TGF-β, and immune checkpoint ligands, impact immune responses that occur in TA-TLS. Finally, it is unknown whether B cells in TA-TLS support antitumor immunity through tumor-antigen presentation to T cells, secretion of proinflammatory cytokines (33), or the direct action of in situ produced antibody, or whether regulatory B cells, which can enhance tumor development by suppressing antitumor immunity (33), exist in TA-TLS. Because B cells are a major component of TA-TLS, it is particularly important to have a more comprehensive understanding of the function(s) of these cells in these structures.

The prognostic impact of TA-TLS has been extensively evaluated. Several studies have pointed to a significant relationship between the densities of TA-TLS and overall patient survival (6, 7), although there are exceptions (refs. 13, 39, 40; Table 1). Because TA-TLS are associated with higher densities of CD8⁺ TIL, it remains possible that TA-TLS are simply proxies for more robust intratumoral T cell effector activity. However, multivariate studies in NSCLC (20) and colorectal cancer (41) have established that the prognostic value of TA-TLS independent of TIL density. Intratumoral TA-TLSs are more significantly associated with enhanced patient survival than peritumoral TA-TLS in pancreatic cancer (42) and early-stage hepatocellular carcinoma (43). Also, patients with oral squamous cell carcinoma whose tumors contained higher proportions of classically organized TA-TLS tended to survive longer, although this was not statistically significant (44). Two studies show that germinal centers within TA-TLS determine their prognostic value in colorectal (35) and lung squamous cell carcinoma (36). However, it is important to establish more generally whether peritumoral and intratumoral, or classically and nonclassically organized TA-TLS differ in their association with patient survival (Fig. 1).

An area of immense interest is whether TA-TLS are associated with patient responsiveness to cancer therapies (Table 1). Interestingly, the presence of TA-TLS was initially associated with a favorable response to neoadjuvant chemotherapy in breast cancer (10, 45). Similarly, densities of TA-TLS in HER2⁺ breast cancer strongly correlate with disease-free survival and responsiveness to adjuvant trastuzumab (46). The presence of B cells and/or TA-TLS prior to treatment is associated with favorable responses to checkpoint blockade immunotherapies in melanoma (14, 47, 48) and soft-tissue sarcoma (49), and one of these studies presents evidence suggesting that immunotherapy might increase TLS density (47). Immune checkpoint treatment of murine tumors increases the number and size of TA-TLS, and promotes a classical organization, in association with diminished tumor outgrowth (29). These observations establish that TA-TLS may be important predictors of patient response to chemotherapy and immunotherapy, along with overall intratumoral CD8⁺ TIL, mutational burden, and programmed death ligand 1 (PD-L1) expression. At the same time, there is a suggestion that TA-TLS may be the site at which these therapies act. However, the range of tumors in which TA-TLS are identified is larger than the range that responds to immune checkpoint blockade. Whether this is a consequence of additional regulatory mechanisms, and whether these operate within the TA-TLS, remains to be determined. As above, it is important to establish more generally whether peritumoral and intratumoral, or classically and nonclassically organized TA-TLS, differ in their association with treatment responses.

Given the considerations above, it is highly attractive to develop immunotherapeutic approaches that induce or augment TA-TLS formation. Interestingly, vaccination induces TLS formation in association with pancreatic tumors (17) and human papilloma virus–driven cervical intraepithelial neoplasia (50). Transgenic overexpression of lymphotoxin-β receptor ligands (51), injection of recombinant LIGHT (CD258; ref. 52), intratumoral administration of CCL21 (53), and intratumoral injection of DCs engineered to overexpress T-bet (54), IL36 (55), or CCL21 (56, 57) all induce TA-TLS in murine tumors. However, there are limited reports of spontaneous TA-TLS development in murine tumors (27, 29, 41, 58), and the mechanisms driving their formation have mostly remained unknown. In murine melanoma, we demonstrate that spontaneous development of intratumoral PNAd⁺ CCL21⁺ vasculature is controlled by effector CD8⁺ T cells and natural killer cells secreting lymphotoxin-α₃ and IFNγ (27). We show that these cells, together with B cells expressing lymphotoxin-α₁β₂, act coordinately as surrogate lymphoid tissue inducer cells to drive TA-TLS development (29). We also demonstrate that a population of cancer-associated fibroblasts (CAF) form a reticular network co-extensive with the immune cells in TA-TLS, and also express high levels of the B-cell attracting chemokine CXCL13, and the B-cell survival factors BAFF and APRIL (27–29). The CXCL13 receptor CXCR5 is critical for intratumoral B-cell accumulation and TA-TLS formation (29). These findings demonstrate a novel and previously undescribed role of CAF as surrogate lymphoid tissue organizer cells that orchestrate TA-TLS development.

CAF populations in TA-TLS are also associated with human melanoma (29). However, in breast cancer, T_FH cells are reported as a CXCL13 source (10), raising the possibility that these cells also aid in TA-TLS development or long-term maintenance. TA-TLS in late-stage NSCLC are associated with a distinct population of CXCL13-producing CD8⁺ T cells (59). Conversely, TA-TLS associated with early-stage NSCLC contain a population of type 3 innate lymphocytes (60). This suggests that the cells responsible for initiating and maintaining TA-TLS in NSCLC may evolve over time, although the direct activity of these cell populations has not yet been demonstrated. However, it is unknown whether these cells exist in TA-TLS from other tumor types, and whether different anatomical compartments contain tissue-resident fibroblasts that are more prone to become organizer-like cells. In addition, it is unknown how CAFs or other cell populations may resemble follicular DCs and fibroblastic reticular cells, which, respectively, promote the formation of distinct T- and B-cell compartments in SLO and are likely to be necessary for TA-TLS with a classical organization (Fig. 1). It is entirely unknown what promotes the development of TA-TLS in peritumoral and intratumoral locations. To the extent that CAFs promote TA-TLS development, it is unknown how this may be related structurally to the formation of desmoplastic stroma by other CAF populations. Overall, while suggesting some general molecular mechanisms may operate to promote TA-TLS development and maintenance, these results also suggest that the cellular sources of these molecules may vary based on tumor type, anatomic location, and/or tumor evolution.

Based on the preponderance of evidence, it seems highly desirable to induce and/or augment TA-TLS development as a new aspect of cancer immunotherapy, either alone or in combination with immunotherapy or chemotherapy. However, given the significant number of studies that show negative prognostic associations, and the heterogeneity of TA-TLS organization, consideration should also be given to optimizing their functionality. Cellular and molecular mechanisms responsible for spontaneous TA-TLS formation, and strategies to induce these structures, are described in murine models. However, much remains to be done to understand the overall heterogeneity and functionality of these structures in human tumors, and the tumor, tissue, and temporal elements that may control these properties. These in turn will enable more refined understanding of the properties of TA-TLS that are of greatest value in determining patient survival, and applicability to the widest array of different cancer types.

V.H. Engelhard reports grants and personal fees from Agenus Inc. outside the submitted work. No potential conflicts of interest were disclosed by the other author.

This work was supported by the U.S. Public Health Service (USPHS) Grants CA78400 and CA181794 (to V.H. Engelhard). A.B. Rodriguez was supported by USPHS Training Grant T32AI007496, and was a recipient of the University of Virginia School of Medicine Wagner Fellowship.