The normal development of lymphoid organs such as lymph nodes and Peyer's patches requires LTβR signaling12,13,14,15,16. LTβR provides differentiation signals to lymphoid stromal cells, forming an effective architecture that promotes efficient immune responses17,18,19,20,21. LTβR signaling is essential for the development of HEVs, which serve as the gateway for lymphocytes to enter the lymphoid organs22. LTβR is also required for chemokine secretion, which recruits and organizes B lymphocytes and T lymphocytes and antigen-presenting dendritic cells into germinal centers. Activation of LTβR by its ligand LIGHT or agonistic antibodies has been shown to increase HEV formation and B lymphocyte and T lymphocyte infiltration into tumors23,24,25,26,27,28. However, the formation of mature TLS with germinal center responses was not demonstrated in previous studies. LTβR pathway may require cooperation with innate receptor signals to form mature TLS. The stimulator of interferon genes (STING) is an intracellular danger signal sensor responsible for the induction of type-I interferon (IFN)-stimulated genes, hence providing a crucial bridge between innate and adaptive immunity29,30. Spontaneous or chemotherapy-induced tumor cell death produces cyclic GAMP (cGAMP), an activating ligand for STING, which induces IFN and inflammatory cytokine responses in the tumor microenvironment31,32. A previous study showed that intratumoral administration of STING agonist ADU-S100 induced the formation of lymphoid aggregates composed of CD3+ T cells and dendritic cells in subcutaneous mouse melanoma33. However, these lymphoid structures lacked germinal center-like B cell clusters characteristic to TLS, demonstrating that STING activation alone is insufficient for promoting the development of functional TLS33. To date, there have been no reports of therapeutic induction of TLS in patients' tumors in clinical trials of STING agonists or other cancer therapeutics. The impact of innate immune activation and engagement of the adaptive immune system in the cascade of mechanisms required for the intratumoral TLS formation has yet to be fully explored. Our approach here focuses on defining the relationship of LTβR and STING in inducing functional TLS and antitumor immune responses.

Previously, we conducted a transcriptome analysis of tumor vasculature in human breast cancer, which identified several genes differentially expressed between the endothelium of tumors rich in TLS and those tumors lacking TLS. The differences in gene expression patterns are likely caused by the exposure of endothelial cells to different environmental cues in TLS-rich tumors versus TLS-free tumors. We reasoned the differentially regulated signaling pathways in endothelial cells may identify the nature of these environmental cues. Based on this idea, we performed an upstream prediction analysis of these gene sets using the ingenuity pathway analysis program. Our assessment identified transcripts of type-I IFN and other signaling molecules involved in inflammation and innate immunity, including tumor necrosis factor (TNF), Toll-like receptor 7 (TLR7), TLR9 and lymphotoxin-β (LTβ; Extended Data Fig. 1). Also identified were regulators of B cell recruitment, activation and differentiation as well as generation/maintenance of T cells and germinal center development, such as CXCL13, CD40L, interleukin (IL)-21 and IL-21R, consistent with the development of TLS (Extended Data Fig. 1). The involvement of LTβ and genes responding to LTβR signaling, such as CXCL13, suggest an active role of the LTβ-LTβR pathway in tumor-associated endothelium and lymphoid organogenesis as previously reported. The involvement of type-I IFN suggests both innate and adaptive immune pathways are active in the tumor microenvironment.

Based on these findings, we administered STING agonist ADU-S100 (cGAMP analog) and LTβR agonistic antibody (4H8) to tumor-bearing mice to examine whether activation of STING and LTβR pathways will induce TLS formation in TLS-free tumors by reproducing the microenvironment of TLS-rich tumors. For this study, we treated C57BL/6 mice bearing subcutaneous syngeneic tumors derived from KrasTrp53Pdx1-Cre mice (hereafter KPC tumors). These mice were treated with STING agonist alone, LTβR agonist alone or the two agonists in combination when tumors reached approximately 100 mm in volume. The STING agonist was administered once via intratumoral injection at 2 μg per tumor (day 0), and 100 μg anti-LTβR was administered intraperitoneally every 3-4 days, for a total of four times until day 10. Untreated mice or mice treated with the STING agonist rarely generated lymphoid aggregates in their tumors (Extended Data Fig. 2a). In comparison, mice treated with the LTβR agonist monoclonal antibody or in combination with the STING agonist induced numerous TLS that resembled human cancer TLS composed of dense clusters of B cells surrounded by CD3 T cells and HEV vessels (Fig. 1a-c and Extended Data Fig. 2a,b). All mice in these two treatment groups developed TLS. The majority of B cells in these tumors were found in dense clusters as TLS unlike T cells, which were found in high density around TLS but also observed broadly throughout the tumor area. Many of these B cells expressed the germinal center B cell marker Bcl6 (Fig. 1d) and were in a proliferative state as indicated by Ki-67 staining (Fig. 1e). These lymphoid structures were also positive for follicular markers CD21 and CD23 (Fig. 1f and Extended Data Fig. 2c-e) and contained Bcl6CD4 T cells (Fig. 1d), which are all characteristics of mature TLS in human cancers, distinct from immature lymphocyte aggregates. Notably, neither TLS nor HEVs developed in T cell-deficient nude mice that were treated with agonist combination therapy, and B cell infiltration was extremely rare (Fig. 1g). We depleted CD4 or CD8 T cells individually or together in wild-type animals to further investigate the role of T cells (Extended Data Fig. 3a,b). This study showed that both CD4 T cells and CD8 T cells are essential to TLS formation. The depletion of either CD4 T cells or CD8 T cells nearly abrogated TLS formation (Fig. 1h,i). Interestingly, CD4 T cell depletion did not affect HEV formation while CD8 T cell depletion abrogated HEVs, indicating that the HEV formation alone does not promote TLS development in the absence of CD4 T cells. Depletion of both T cell subsets completely abrogated TLS and HEV formations confirming the result of the nude mouse study (Fig. 1g-i). These observations demonstrate the essential role of T cells in creating the immune landscape supportive of antitumor B cell responses in the form of TLS. TLS did not develop upon combination therapy in B cell-deficient CD79a knockout mice as expected (Fig. 1h,i and Extended Data Fig. 3c). Interestingly, HEVs developed in these mice but at a considerably reduced density, indicating that B cells are also crucial for HEV formation (Fig. 1h,i).

In wild-type mice, TLS and HEV formation was also observed in orthotopic KPC tumors grown in the pancreas (Fig. 2a) and in orthotopic Py230 (MMTV-PyMT) mammary tumors (Fig. 2b) as well as in 'immune cold' orthotopic 76-9 rhabdomyosarcoma grown in the calf muscle upon combination therapy (Fig. 2c). Either LTβR monotherapy or combination therapy induced TLS in KPC and Py230 tumors (Figs. 1c and 2b). In comparison, combination therapy with repeated administration of STING agonist was necessary for TLS development in rhabdomyosarcomas, suggesting the importance of STING activation for TLS induction in 'immune cold' tumors (Fig. 2c).

We next analyzed how tumors respond to the STING agonist treatment. KPC tumors were collected at different time points after agonist treatment, and STING signaling was analyzed by western blot. Intratumoral ADU-S100 induced IRF3 phosphorylation and IFNβ expression within 4 h, indicating the rapid activation of the STING pathway in the tumor microenvironment (Extended Data Fig. 4a). Immunofluorescence of tumor sections confirmed IFNβ induction in macrophages, endothelial cells and possibly other cell types 4 h after ADU-S100 monotherapy or combination therapy (Extended Data Fig. 4b). We also analyzed STING signaling activation by ADU-S100 in different cell types in vitro. The phosphorylation of IRF3 in cultured KPC cells and endothelial cells was analyzed by western blot (Extended Data Fig. 4c). Leukocytes from peripheral blood or peritoneum of normal mice were analyzed by flow cytometry (Extended Data Fig. 4d). The results of these studies suggest that multiple cell types, including tumor cells, endothelial cells, T cells and macrophages/monocytes, could respond to intratumoral ADU-S100 and collectively contribute to the STING-induced immune responses.

The growth of different types of tumors was monitored for 2 weeks from the day of treatment (day 0). The LTβR monotherapy had little or no significant effect, and only a moderate effect of STING monotherapy was observed in the KPC tumors and Py230 mammary tumors at this dose (2 μg intratumoral; Fig. 3a). In comparison, the combination therapy reduced KPC tumor burden by more than 50% and significantly delayed Py230 and 76-9 tumor growth (Fig. 3a). The combination therapy showed no therapeutic benefit in T cell-deficient nude mice (Fig. 3b) suggesting that T cells play an essential role in the immediate-early tumor inhibition by this therapy. The lack of effect in nude mice also indicates that the agonist treatment has little or no direct inhibitory effect on tumor growth. The depletion of CD8 T cells or simultaneous depletion of CD4 T and CD8 T cells resulted in exaggerated tumor growth in both combination therapy-treated and untreated mice, and combination therapy had no tumor suppression effect, demonstrating the importance of CD8 T cells (Fig. 3c). On the other hand, the depletion of CD4 T cells inhibited tumor growth in untreated mice after day 7 (Fig. 3c). This inhibition is likely due to the depletion of CD4 regulatory T cells and resulting potentiation of CD8 T cells, as all CD4 T cells were depleted in this procedure (Extended Data Fig. 3b). In comparison with these mice, the CD4 T cell-depleted, combination therapy-treated mice did not show additional tumor suppression from day 7 (Fig. 3c), which may be related to the inability of CD4 T cell-depleted mice to develop TLS. Therefore, we tested the combination therapy in B cell-deficient CD79a knockout mice to assess the contribution of TLS-B cells. In these mice, combination therapy was unable to continue to reduce tumor burden after day 6, indicating a partial loss of the therapeutic effect (Fig. 3d). These results demonstrated that the initial tumor inhibition is mainly mediated by CD8 T cells, with a delayed and moderate contribution of B cells starting in the second week.

The role of B cells in TLS is thought to be the local development of humoral immunity involving antibody productions and differentiation of memory B cells; hence, contributions of TLS-B cells cannot be fully examined in a short-term, 2-week tumor study. To investigate the importance of TLS and humoral immunity, we used a tumor recurrence model. In this model, we surgically resected the subcutaneous KPC tumors and sentinel lymph nodes after monotherapy or combination therapy on day 14 and reinoculated KPC cells subcutaneously 2-3 weeks later to mimic tumor relapses (Fig. 3e). Remarkably, mice that received the neoadjuvant agonist combination before the tumor resection, which all developed TLS, showed excellent long-term survival with complete disappearance of the tumors (Fig. 3f,g). The reinoculated tumor cells initially grew in the host mice as small papules at the injection sites for several days but then regressed and eventually disappeared, except for a few cases, demonstrating that these animals became resistant to KPC tumors. Mice that received neoadjuvant LTβR monotherapy, which also developed TLS, acquired tumor resistance but of lesser duration (Fig. 3f,g). The neoadjuvant STING monotherapy had little survival benefit, and the reinoculated tumors grew rapidly in most mice, similarly to the untreated control.

To explore the mechanism by which this neoadjuvant therapy confers resistance to future tumor development, we further examined tumor-infiltrating lymphocytes and TLS in the primary tumors on the day of tumor resection. Our flow cytometry analysis showed that tumors of the combination therapy group had a significant increase in B cells, consistent with the presence of TLS in these tumors (Fig. 4a and Extended Data Fig. 5a-c). B cells were prominent (average 27% of infiltrating lymphocytes) in tumors treated with the STING-LTβR combination therapy (Fig. 4a). Mice treated with LTβR monotherapy also showed an increase in B cells. The analyses of early and late activation markers, CD69 and CD44, indicated that activated B cells were substantially increased in tumors after combination therapy but not LTβR monotherapy (Fig. 4a and Extended Data Fig. 5c). In addition, a significant increase in CD73PD-L2CD19 memory B cells was observed in tumors after combination therapy but not LTβR monotherapy (Fig. 4a and Extended Data Fig. 5c). Immunofluorescence of tumor sections confirmed that most B cells in TLS were activated in the combination therapy group (Fig. 4b). In contrast, the fractions of total and activated B cells were not significantly different between the tumor-draining lymph nodes of three therapy groups (Fig. 4a). There was a trend toward an increase in the B cell counts in lymph nodes upon LTβR monotherapy or combination therapy, reflecting a trend toward an increase in total lymphocyte counts in these lymph nodes (Extended Data Fig. 5d). This observation suggests that agonist treatments also influence lymph node responses to tumors. However, CD69, CD73 or Ki-67 staining of CD19 B cell follicles did not indicate increased germinal center formation or activity in the lymph nodes upon combination therapy compared with other groups of tumor-bearing mice (Fig. 5).

Our results suggest that one advantage of combination therapy over LTβR monotherapy is the activation and maturation of tumor-infiltrating B cells. The extensive accumulation of B cells in TLS suggests germinal center-like B cell responses. Immunofluorescence analysis of tumor sections demonstrated considerable accumulation of IgGCD138 B cells within TLS (Fig. 4c), indicating that these B cells were antigen primed, differentiated to antibody-producing plasma cells and have undergone immunoglobulin class switching as expected from the germinal center activities of mature TLS. The presence of CD73 B cells is consistent with the development of memory B cells in these TLS (Fig. 4d). In comparison with B cells, agonist treatments did not significantly increase tumor infiltration of CD4 or CD8 T cells (Extended Data Fig. 6). Likewise, we did not observe significant changes in the T cell fractions in draining lymph nodes (Extended Data Fig. 7).

We next conducted sequencing of total tumor RNA to characterize the immune landscape of the tumors on the day of tumor resection (day 14). This analysis showed an altered tumor transcriptome in the combination therapy group compared with other groups (Fig. 6a). Among the increased transcripts were those of inflammation/innate immunity-related genes Tnf, Lta, Ltb, Light and Tlr7 (Fig. 6b and Extended Data Fig. 8), many of which were also predicted by our pathway analysis of clinical specimens comparing TLS-rich versus TLS-free breast adenocarcinomas. In addition, the transcripts of effector genes crucial to adaptive immunity, Fas ligand (Fasl) and granzyme B (Gzmb), as well as a key chemokine for the recruitment of circulating naive T lymphocytes, Ccl21, were increased (Fig. 6b). The immunosuppressive cytokine Il10 was reduced to half of the level of the untreated control. There was a prominent transcriptomic signature for the immunostimulatory type 1 helper T (T1) cell environment in tumors treated with combination therapy. For instance, T1-inducing transcription factor Tbx21 (Tbet) was increased by twofold. Il12rb2 and Stat4 were both significantly increased (Fig. 6b), which is expected to promote T cell differentiation toward the T1 phenotype by enhancing IFNγ expression. Indeed, Ifng was significantly elevated in the combination therapy group (Fig. 6b). The significantly increased transcripts of IL-21 and IL-21 receptors (Il21 and Il21r) suggest an active pathway that promotes T cell differentiation and IgG production in germinal centers (Fig. 6b). These two genes were also identified in the pathway analysis of individuals with TLS-rich cancer. A follicular B cell marker Cxcr5 (the receptor for CXCL13) and the genes necessary for B cell activation by T cells in germinal centers (Cd40, Cd40lg) as well as the plasma cell and memory B cell markers, Mzb1 and Cd27, were significantly increased and consistent with the B cell differentiation associated with the formation of functional TLS (Fig. 6 and Extended Data Fig. 8). Furthermore, B cell survival factors and their receptors important for humoral immunity (Tnfsf13b/BAFF, Tnfsf13/APRIL, Tnfrsf17/BCMA, Tnfrsf13c/BAFFR, Tnfrsf13b/TACI) were elevated (Extended Data Fig. 8). The effects of LTβR monotherapy were similar but not as robust for many genes.

These findings are consistent with the presence of T cells and B cell activation/differentiation in TLS of these tumors as detected by immunostaining (Figs. 1 and 4) and consistent with the findings from flow cytometry (Fig. 4 and Extended Data Fig. 5). There were also increased transcripts of immune checkpoint molecules PD-1 and PD-L1 as well as H2-K2 (major histocompatibility complex class I; Fig. 6b), consistent with an immune responsive tumor microenvironment created by the combination drug treatment. A gene-set enrichment analysis indicated enhanced inflammatory/innate immune responses as well as adaptive immunity and cytotoxicity by combination therapy (Fig. 6c).

We also found increased expression of class-switched IgG immunoglobulin heavy chains in LTβR monotherapy and combination therapy groups (Fig. 6d) reflecting the accumulation of plasma cells and antibody production in TLS (Fig. 4c). These results recapitulated the abundant expression of IgG in TLS-rich, but not in TLS-free, human pancreatic ductal carcinomas and breast adenocarcinomas (Fig. 6e,f).

The tumor immune environment was further investigated by single-cell RNA-sequencing (RNA-seq) analyses of fluorescence-activated cell sorting (FACS)-sorted tumor-infiltrating CD45 leukocytes on the day of tumor resection. Uniform manifold approximation and projection of different leukocyte subsets showed that the most prominent changes induced by the agonist treatment were the substantial expansions of intratumoral B cells and neutrophils by combination therapy and STING monotherapy, respectively (Fig. 7a and Extended Data Fig. 9a). Combining LTβR agonist with STING therapy canceled the STING-induced neutrophil expansion (Fig. 7a). Further sub-clustering of the B cell population showed that all B cell subtypes, including follicular B cells, memory B cells, plasma cells and long-lived plasma cells, are greatly increased by combination therapy (Fig. 7b,c and Extended Data Fig. 9b). Long-lived plasma cells are terminally differentiated mature B cells that have undergone somatic hypermutations for high-affinity antibody production, and therefore, end products of the germinal center reaction. Our observations support the presence of intratumoral germinal center reactions in this treatment group. Interestingly, there were vast expansions of IgD, IgM and IgG B cells, but the expansion of IgA B cells was limited, suggesting that germinal center B cells of TLS favor class switching to IgG over IgA under combination therapy (Fig. 7c). IgD expression was found exclusively in the follicular B cell population and the expression of class-switched IgG was found in memory B and plasma/long-lived plasma cells (Fig. 7c), again supporting our conclusion for the intratumoral development of mature B cells driven by the germinal center responses taking place within these tumors. The high-affinity IgG-producing long-lived plasma cells could provide long-term protection against future tumor recurrence. LTβR monotherapy had similar but much lesser effects on the expansion of IgG and total B cells (Fig. 7c). STING activation alone did not expand the intratumoral B cell population or increase immunoglobulin expression, indicating that a substantial TLS formation and antibody production require activation of both STING and LTβR pathways.

Unlike the prominent effects on B cell expansion, the agonist treatments did not appear to affect the abundance of total CD4, CD8, γδT, natural killer (NK) T or regulatory T cells in tumors (Extended Data Fig. 10a). However, an analysis of CD4 T cell subsets indicated the emergence of a distinct cluster of naive CD4 T cells, identified by Ccr7 and L-selectin expression (Cd4Ccr7Sell), after combination therapy (Fig. 7d-f). CCR7 is a chemokine receptor for CCL19 and CCL21. L-selectin binds to sulfated sialyl Lewis-X carbohydrate structure (MECA-79 epitope) expressed by HEVs to initiate adhesion and extravasation of circulating lymphocytes. The expansion of the naive CD4 T cell population suggests increased recruitment of these lymphocytes from the circulation, facilitated by increased HEV formation in the tumors of this treatment group. Naive CD4 T cells also increased after LTβR monotherapy but to a lesser extent (Fig. 7d-f), possibly corresponding to the formation of fewer HEVs in this group. Furthermore, we found a considerable increase in type 2 helper T (T2) cells in the tumors upon LTβR or combination therapy (Fig. 7d-f). The T2 cytokines expressed by this T cell subset, such as IL-4, IL-5 and IL-13 (Fig. 7e), are important for the immunoglobulin class switching in B cells and, therefore, important for the maturation of B cells in TLS. In contrast to the naive and T2 populations, the T17 subset of helper T cells diminished considerably by combination therapy (Fig. 7d-f) demonstrating a shift in the T2/T17 balance toward T2 upon combination therapy, likely contributing to the development of humoral immunity in this treatment group.

A subcluster analysis of CD8 T cells identified a cluster of memory CD8 T cells (Extended Data Fig. 10b,c). The relative abundance of memory CD8 T cells increased in treatment groups, especially LTβR monotherapy and combination therapy groups, threefold or more compared with the untreated group (Extended Data Fig. 10b,d). Thus, increased naive CD4 and memory CD8 T cells and expanded T2 and diminished T17 populations were the characteristics of tumor-infiltrating T cells under combination therapy. Overall, the results of our genomic analyses, flow cytometry and immunostaining demonstrated that STING-LTβR combination therapy enhances the adaptive immunity to tumors.

We showed that both LTβR monotherapy and STING-LTβR combination therapy induced TLS in primary KPC tumors; however, only combination therapy significantly increased activated B cells and memory B cells in the tumors, and combination therapy provided much stronger protection against the second tumor challenge than LTβR monotherapy. These results may reflect qualitative differences between TLS (or 'fitness' of TLS) of the two treatment groups and suggest the importance of TLS-driven humoral immunity enhanced by the combination therapy. Therefore, we measured antibodies in the blood of these mice with specificity for tumor cells. Blood plasma was collected from the agonist-treated or untreated mice 2 weeks after the tumor resection (Fig. 8a), diluted at a 1:50 ratio in PBS and incubated in vitro with live KPC cells. The binding of plasma IgG and IgM to the tumor cell surface was determined by flow cytometry using fluorescently labeled goat anti-mouse IgG and IgM antibodies. This study showed that the IgG-bound KPC cells were significantly increased in the combination therapy group, suggesting high-titer and/or high-affinity antitumor IgG productions by B cells of this group (Fig. 8b-d). Specific binding of IgM was undetected in this assay. Clinical studies have shown that patients with cancer exhibiting IgG-bound tumor cells have considerable therapeutic responses to immune checkpoint inhibition and prolonged progression-free survival, signifying the importance of our findings. Our results suggest the presence of antibody-secreting plasma cells residing in the bone marrow of combination therapy-treated mice. Therefore, we analyzed the bone marrow 2 weeks after tumor resection (Fig. 8a). Flow cytometry analysis demonstrated marked increases in Blimp1CD44 cells and CD138CD44 cells in the combination therapy group, but not in other groups, suggesting accumulation of long-lived plasma cells in the bone marrow of these mice (Fig. 8e).

A serum transfer experiment further demonstrated the importance of humoral immunity. We collected serum from mice 2 weeks after the neoadjuvant agonist combination therapy and tumor resection surgery (Fig. 8a). The serum was pooled for each group, and heat inactivation to abolish complement activity was carried out. The serum was aliquoted, and 100 μl was intraperitoneally given to treatment-naive recipient mice on the day of KPC tumor injection to these mice. These mice continued to receive serum transfer two times per week, and tumor growth was monitored. In this study, the mice receiving serum transfer from the donors that underwent tumor resection surgery without neoadjuvant therapy showed a tumor growth rate similar to that of the control mice receiving no serum transfer or serum transfer from healthy mice with sham surgery (Fig. 8f). In comparison, the mice receiving serum transfer from the donors that underwent neoadjuvant combination therapy before the surgery showed significant suppression of tumor growth, demonstrating the presence of antitumor immunity in the serum of neoadjuvant-treated mice (Fig. 8f). To further demonstrate the importance of humoral immunity, we treated B cell-deficient CD79a knockout mice with neoadjuvant combination therapy followed by the tumor rechallenge 2 weeks later. Unlike wild-type mice, these mice failed to reject the reinoculated tumors (Fig. 8g), indicating the essential role of B cells in the therapy-induced development of antitumor immunity. The combined results demonstrate the importance of B cell humoral immunity.

We next examined the role of cell-mediated immunity in the therapeutic effect of the neoadjuvant agonists. KPC tumor-bearing mice were treated with or without neoadjuvant STING-LTβR combination therapy followed by tumor resection surgery. Two weeks later, starting from one day before reinoculation of KPC cells, these mice received intraperitoneal injection of anti-CD8, NK1.1 or Ly6G antibodies to deplete CD8 T cells, NK cells or neutrophils, respectively. In this study, neoadjuvant-treated mice lost the ability to inhibit the growth of reinoculated tumors upon deletion of CD8 T cells or NK cells but not neutrophils (Fig. 8h). These results demonstrate that cellular immunity by CD8 T cells and NK cells is also crucial for the antitumor immunity acquired through neoadjuvant combination therapy.