CD86⁺ Antigen-Presenting B Cells Are Increased in Cancer, Localize in Tertiary Lymphoid Structures, and Induce Specific T-cell Responses

The prognostic significance of B cells and plasma cells in human cancers has increasingly been recognized (1–5). B cells and plasma cells in the tumor microenvironment may interact with T cells to induce or enhance antitumor immune responses. Tumor-associated CD138⁺ plasma cells are present in several types of cancer (4, 6–13). Accordingly, antibodies against tumor-associated antigens are detected in sera of patients with cancer, underlining the contribution of a tumor-directed humoral immune response to antitumor immunity (9, 11, 12).

The role of B cells in the antitumor immune response is also highlighted by studies reporting on B cell–rich clusters in the tumor microenvironment, which show similarities in structure and composition to lymphoid follicles in secondary lymphoid organs. Presence of these tertiary lymphoid structures (TLS) in the tumor microenvironment has been associated with a superior prognosis in several types of cancers (14–16). Functionally, TLSs may serve as sites for local generation of tumor antigen–specific B and T cells. This includes differentiation and maturation of B cells and potentially also presentation of antigens to T cells, as B cells isolated from TLSs of patients with gastric cancer induce increased proliferation of T cells in vitro (17, 18). Accordingly, activated CD86⁺ B cells have been shown to be increased in the tumor microenvironment in patients with colorectal cancer, gastric cancer, and head and neck squamous cell carcinoma (HNSCC; refs. 4, 9, 11, 18).

CD21 is a complement receptor that binds split products of complement 3. Lowering the threshold antigen concentration for B-cell activation and uptake of opsonized antigen by follicular dendritic cells (DC) are major functions of CD21 (19). CD21 expression decreases upon B-cell activation and differentiation into class-switched, pathogen-specific B cells in infectious diseases. B cells with strong antigen-presenting capacity are enriched in this class-switched CD21⁻ subset (20, 21). We previously demonstrated that antigen-presenting B cells can be further identified on the basis of coexpression of CD86. CD21⁻CD86⁺ B cells had the highest antigen-presenting capacity, showed similarities to CD40-activated B cells, and were termed antigen-presenting B cells (BAPC). These cells are IgM⁻IgD⁻ class switched, show high expression of MHC molecules, and possess an immunostimulatory capacity. They are also increased after vaccination and under inflammatory conditions (22). Their counterpart, CD21⁺CD86⁺ B cells (Breact) are IgM⁺IgD⁺ naïve B cells, express MHC molecules, but only show a moderate capacity for presentation of antigen. They can differentiate into BAPCs upon prolonged stimulation. In contrast, the majority of CD21⁺ B cells are naïve, non–class-switched, do not express the costimulatory molecule CD86, do not possess an immunostimulatory capacity, and are decreased under inflammatory conditions (22).

We hypothesized that BAPCs might be increased in tumor-infiltrating lymphocytes (TIL) of solid cancers and contribute to antitumor immune responses by local induction of tumor-specific T-cell responses in the tumor microenvironment. We performed a prospective analysis of tumor-infiltrating B cells and TLSs in 10 different cancer types and analyzed the antigen-presenting capacity of BAPCs from patients with cancer to demonstrate their potential to induce antitumor T-cell responses.

A total of 237 patients with cancer across 10 different tumor types were included: HNSCC (n = 46), non–small cell lung cancer (NSCLC; n = 28), hepatocellular carcinoma (HCC; n = 12), breast cancer (n = 16), testicular germ cell carcinoma (TGCT; n = 4), ovarian carcinoma (OVCA; n = 7), urothelial carcinoma (UCC; n = 6), colorectal cancer (n = 40), esophagogastric adenocarcinoma (EGA; n = 46), and renal cell carcinoma (RCC; n = 32). Samples were collected between 2011 and 2019. Resection of a previously untreated primary tumor of the different cancer types and an age >18 were main inclusion criteria. Patients with autoimmune disease or immunosuppressive therapy were not eligible. Peripheral blood, tumor-draining lymph nodes (TDLN), fresh tumor and normal tissue, and formalin-fixed paraffin-embedded (FFPE) tumor tissue were collected. Peripheral blood mononuclear cells (PBMC) from 33 age- and sex-matched healthy donors were included as controls for flow cytometry. Disease stage was assessed according to the 8th edition of the tumor–node–metastasis classification (23). Patient characteristics are summarized in Table 1. PBMCs were isolated by density-gradient centrifugation with Pancoll human (PAN-Biotech). Single-cell suspensions from tissues were generated by mechanical (gentleMACS Dissociator; catalog no. 130-093-235; Miltenyi Biotech) and enzymatical (320 U/mL collagenase IV and 100 U/mL DNAse-I, Applichem) processing in C-Tubes (Miltenyi Biotech). On the gentleMACS, the predefined human tumor programs 1–3 were used. The protocol included incubation for 20 minutes at 37°C in a rotation incubator before gentleMACS human tumor programs 2 and 3. Finally, single-cell suspensions were filtered through a 100 μm and a 70 μm cell strainer (Corning). A maximum of 5–10 × 10⁶ cells per 1.8 mL cryotube (CryoPure 2.0 mL, Sarstedt) were resuspended in FBS (Pan-Biotech) + 10% DMSO (Sigma-Aldrich) and stored in liquid nitrogen until further use. The study was conducted in accordance with the Declaration of Helsinki, written informed consent was signed by all patients, and this study was approved by our institutional ethics committee (no. 17-282).

T cells and B cells were isolated from single-cell suspensions generated from tumor-draining lymph nodes of patients with colorectal cancer. T cells were isolated by negative selection using the Pan T-cell isolation kit (human; Miltenyi Biotech). B-cell subsets (CD86⁻ B cells, CD21⁺CD86⁺, and CD21⁻CD86⁺ B cells) were isolated by positive selection using CD19 MicroBeads (human; Miltenyi Biotech) and by sorting on a FACSAria III (BD FACSARIA III cell sorter, RRID:SCR_016695, BD Biosciences) from single-cell suspensions of tumor-draining lymph nodes of patients with colorectal cancer. Antibodies used for sorting of B cells (Aqua Zombie, CD45, CD20, CD21, and CD86) are listed in Supplementary Table S1. A total of 5 × 10⁴ T cells were cocultured with 1 × 10⁴ autologous B-cell subsets in the presence of a carcinoembryonic antigen (CEA) peptide pool [PepMix Human (CEA); JPT Peptide Technologies] at a concentration of 0.1 μmol/L. AIM-V medium (Thermo Fisher Scientific) was supplemented with anti-CD28 (0.1 μg/mL; Mabtech). Anti-CD3 (1 μg/mL; Mabtech) was used as positive control. Cells were cultured for 20 hours at 37°C, 5% CO₂ on a precoated anti-human IFNγ-FITC FluoroSpot plate (Mabtech). FluoroSpot analysis was performed on an AID iSpot FluoroSpot reader (AID). The number of spots and representative pictures were assessed using the AID EliSpot Reader Software v7.0 (AID).

PBMCs, single-cell suspensions, or isolated B cells were first stained with Zombie dye (BioLegend) and human FC Receptor Binding Inhibitor (Thermo Fisher Scientific) for 15 minutes in PBS. Surface staining was performed using a mastermix for 20 minutes at 4°C in CellWash (BD Biosciences) using standard 96-well plate with U-bottom (BRAND) using the following antibodies: CD20, CD19, CD45, CD21, CD86, CD3, CD4, CD8 (for detailed antibody information, see Supplementary Table S1). Fluorescence-Minus-One was performed to control for false-positive staining. Dead cells were excluded using Zombie dye (BioLegend). For staining of CEA-specific B cells, CEA protein (Sino Biological) was biotinylated using EZ-Link NHS-Biotin Reagent (Thermo Fisher Scientific) according to manufacturer's protocol. Biotinylated CEA protein was preincubated with Streptavidin-PE (BioLegend) in a ratio of 1:3 for 10 minutes and used for staining. B cells isolated from healthy control PBMCs were used as negative controls (Supplementary Fig. S1). Data were acquired on a Gallios flow cytometer (RRID:SCR_016702, Beckman Coulter) and analyzed using the Kaluza software (Kaluza, RRID:SCR_016182; Version 2.1; Beckman Coulter). The gating strategy is shown in Supplementary Fig. S2.

Whole slides of tumor specimens were stained with CD20, CD3, or CD86 on a Leica BOND-MAX platform according to the manufacturer's protocol (for detailed antibody information, see Supplementary Table S1). Detection of primary antibodies was performed with 3,3′-Diaminobenzidine (DAB). High-resolution images were captured using a Leica SCN300 slide scanner (20× objective). Tumor areas on CD20 slides were delineated by two experienced pathologists (A. Quaas and S. Eidt) and digitally transferred to the QuPath v2.0 software (RRID:SCR_018257). A total of 300 and 2,000 μm margins based on the tumor outline were automatically generated with QuPath. Clusters of CD20⁺ cells in the respective areas were manually counted to assess the number of TLSs per area.

Significant differences were calculated by two-tailed t test or one-way ANOVA, where appropriate, using GraphPad Prism v8 software (GraphPad Prism, RRID:SCR_002798). Figures were generated with GraphPad Prism software and Adobe Photoshop CS2 (Adobe Photoshop, RRID:SCR_014199). P values of statistical significance were marked with asterisks as indicated in figure legends. Mean values and SDs were calculated from the indicated number of independent experiments (n).

TILs in single-cell suspensions of untreated patients with cancer from 10 different types of cancers (n = 237), corresponding PBMCs (n = 232), normal tissues (NT; n = 137), and PBMCs of healthy controls (HC; n = 33) were analyzed by flow cytometry for the presence of CD20⁺CD19⁺ B cells (Fig. 1A and B). B cells were not significantly increased in PBMCs of patients with cancer compared with HC PBMCs, but TILs contained significantly more B cells than NT. Processing of 71 of 137 (51.8%) corresponding NT did not reveal sufficient lymphocytes for detailed flow cytometric analyses. B-cell infiltration also differed across cancer types, with TILs of HNSCC, breast cancer, and colorectal cancer showing the highest B-cell content (Fig. 1B).

When assessing B cells according to their expression of CD21 and CD86 (Fig. 1A), site-specific variation in the percentages of tumor-infiltrating BAPCs (CD21⁻CD86⁺ B cells) became apparent (Fig. 1C). Tumors of patients with RCC contained less BAPCs (1.2% ± 0.9) than those of UCC (7.8% ± 4.6, P = 0.0003), OVCA (6.5% ± 3.6, P = 0.0081), or breast cancer (4.9% ± 4.9, P = 0.0343). The highest BAPC-fraction in TILs was observed in patients with UCC, whereas it was lower in EGA (2.2% ± 1.7; P = 0.0020), colorectal cancer (3.0% ± 2.4, P = 0.0128), HNSCC (3.3% ± 2.4, P = 0.0298), and NSCLC (3.0% ± 2.6, P = 0.0131).

In addition to the described site-specific variations of BAPCs in the tumor microenvironment, we also compared their abundance in PBMCs. Most cancer types showed a significant increase of BAPCs in TILs compared with PBMCs of HC or patients with cancer (Fig. 2A). Three of 10 cancer types differed from this infiltration pattern: colorectal cancer and EGA showed increased percentages of BAPCs not only in TILs, but also in PBMCs compared with HC, thus demonstrating a systemic BAPC increase. Along with the overall low B-cell infiltration in RCC, we found no increase of BAPCs in PBMCs or TILs of these patients. CD21⁺CD86⁺ Breact cells were also significantly increased in TILs compared with HC in all cancer types, except TGCT (Fig. 2B). HNSCC and colorectal cancer showed the highest infiltration of Breact cells in TILs (14.0% ± 7.2 and 16.0% ± 9.2, respectively; Fig. 2C).

We used IHC on FFPE tumor samples to elucidate the spatial distribution of BAPCs. We detected colocalization of CD86⁺ cells in areas with a high density of CD20⁺ cells, which were surrounded by CD3⁺ cells (Fig. 3A). Thus, CD86⁺ B cells accumulate in areas of TLSs with similarities to germinal centers in secondary lymphoid organs. To examine this relation between BAPCs and TLSs in the whole cohort, we first performed digital image analyses of TLSs in the tumor microenvironment across all 10 cancer types. TLSs were defined as highly dense CD20⁺ cell clusters, as specified by other publications (24–26). Tumor areas were determined by experienced pathologists, and the invasive margin (the area of 300 μm surrounding the tumor) and a greater margin (the area of 2,000 μm surrounding the tumor) were digitally delineated (Fig. 3B, top left). The surrounding margins contained significantly more TLSs/mm² than the tumor area itself, and densities of TLSs in the 300 and 2,000 μm margins were higher in samples of advanced tumor stages (Supplementary Fig. S3A and S3B). Significant differences in the number of TLSs/mm² between cancer types were detected (Fig. 3B), and the frequencies of TLS-high and TLS-low patients depended on the cancer type (Supplementary Fig. S3C). Whereas in some types of cancer, patients with high numbers of TLSs (>median of tumor type) and low numbers of TLSs (<median of tumor type) are almost equally distributed (e.g., tumor area for breast cancer, HNSCC, and colorectal cancer), in other cancer types, the majority of patients were classified as TLS low (e.g., tumor area RCC 70% or HCC 87% TLS-low patients).

IHC analyses of CD86⁺ B cells in tumor tissues indicated that BAPCs infiltrating the tumor microenvironment localized in TLSs. To confirm this, we correlated the percentage of BAPCs with the number of TLSs/mm² and found a weak correlation in the 300 and 2,000 μm margins (Fig. 4A). This correlation was more pronounced when dividing patients into BAPC-high (%BAPC >median +15%), BAPC-low (%BAPC <median − 15%), or BAPC-intermediate (%BAPC median ± 15%) and TLS-high (>median) or TLS-low (<median; Fig. 4B) groups.

Results from flow cytometry also demonstrated that with increasing tumor infiltration by CD8⁺ T cells, the percentage of BAPCs increased. When assessing CD8⁺ T cells (within CD3⁺ TILs) and BAPCs (within CD19⁺CD20⁺ TILs), there was a significant correlation between CD8⁺ T cell–high/BAPC-high patients and CD8⁺ T cell–low/BAPC-low patients. This correlation was not found for bulk CD19⁺CD20⁺ B cells, CD21⁺CD86⁺ Breact, or CD21⁺CD86⁻ B cells (Fig. 4C). Because CD8⁺ T cells have been shown to colocalize with activated B cells in TLSs and increase patient survival (2, 17), these results provide further evidence for a role of local antigen presentation by BAPCs in the tumor microenvironment.

To investigate the functional role of BAPCs in patients with cancer, we first analyzed the BAPC, Breact, and CD86⁻ B-cell subsets for the presence of tumor antigen–specific B cells. For this, we produced antigen-biotin tetramers, which consisted of one fluorescently labeled streptavidin molecule and four biotinylated CEA molecules (Fig. 5A). This tetramer complex is able to cross-link a CEA-specific B-cell receptor (BCR), thus allowing for detection of CEA-specific B cells in flow cytometry. The cancer testis antigen CEA was chosen as model antigen because it is expressed in around 50% of patients with colorectal cancer and has previously been shown to induce tumor-specific T-cell responses (27, 28). We tested for CEA-specific B cells in the TDLNs of patients with colorectal cancer. In 5 of 10 TDLNs from patients, we detected CEA-specific B cells, with a range of 0.25% to 1.9% of CD20⁺CD19⁺ B cells being CEA positive (Fig. 5B). BAPCs and Breact cells contained more CEA-specific B cells than the CD86⁻ B-cell subpopulation (Fig. 5C).

Next, we investigated whether BAPCs from patients with colorectal cancer were able to induce an antitumor T-cell response. We performed FluoroSpot analyses using sorted BAPCs, Breact cells, and CD86⁻ B cells from 3 patients with colorectal cancer. Because the B-cell content in tumor derived, single-cell suspensions was too low for in vitro assays, B cells were sorted from TDLNs and coincubated together with negatively isolated autologous CD3⁺ T cells in the presence of a CEA-peptide pool. One of these patients showed CEA-specific T cells and was used to demonstrate antigen-presenting capacity of BAPCs. When T cells were cocultured with the CEA peptide pool and BAPCs, an IFNγ response was induced, which was not seen when T cells were cocultured with the CEA peptide pool and Breact cells or CD86⁻ B cells (Fig. 5D). Taken together, these results suggest a functional role for BAPCs as antigen-presenting cells in patients with cancer.

The role of B cells in antitumor immune responses has been highlighted by studies reporting a correlation between B-cell abundance and patient prognosis or survival (1–4). Especially with regard to TLSs in patients with cancer, their antigen-presenting function is increasingly gaining attention because TLSs may serve as site for local interaction of B and T cells. We here demonstrated that the CD21⁻CD86⁺ B-cell subpopulation (BAPC) was increased in tumor tissue of nine different cancer types and colocalized with T cells in TLSs. BAPCs from patients with cancer also induced a tumor antigen–specific T-cell response in vitro.

Studies comparing the immune infiltrate in different types of cancer are scarce, with most studies focusing on only one tumor type (29). One comprehensive study compared patterns of TILs in 33 different cancer types on the basis of gene expression data compiled from database The Cancer Genome Atlas (30). These data on lymphocyte infiltration coincides with the pattern of B-cell infiltration we described here. One exception is RCC, which is generally considered to be among the top third of cancer types with regard to immune infiltration (30, 31), but ranked lowest in our study concerning B cells, especially BAPCs. Another study on B-cell infiltration in RCC tumors supports our data by demonstrating that only 29% of patients show positive staining for CD19⁺ B cells in tumor tissue (32).

We found CD86⁺ B cells colocalized in TLSs and that the density of BAPCs was increased in TLS-high samples. We also related the presence of high percentages of BAPCs to high tumor infiltration by CD8⁺ T cells, which could be another indicator for their antigen-presenting function in patients with cancer. Colocalization of B cells with CD8⁺ T cells has been shown to result in increased patient survival compared with presence of CD8⁺ T cells alone (2, 17). Taken together, these results suggest local antigen presentation as important mode of action underlying the positive effects of TLS in cancer.

To facilitate determination of TLS abundance in a large patient cohort, we limited our IHC analysis to the identification of TLSs by dense CD20 clusters. In addition to CD20⁺ B cells, mature TLSs are defined by the presence of CD3⁺ T cells, follicular DCs, and high endothelial venules (16). Nevertheless, dense accumulation of CD20⁺ B cells is a major component and this method has been used in other studies to investigate TLSs and their relevance for patient outcome (24–26). Available studies were confined to selected tumor types and, to our knowledge, we here compare for the first time the density of TLSs in different cancer types and different compartments of the tumor microenvironment. Our finding that high TLSs correlated with higher Union for International Cancer Control (UICC) stage also adds to the hypothesis that TLS formation increases with continuous exposure of immune cells to tumor antigens (33–35).

As expected from our previous study (22), BAPCs were potent antigen-presenting cells. In a proof-of-principle experiment, BAPCs isolated from TDLNs induced a tumor antigen–specific response in autologous T cells from a patient with colorectal cancer. Although Breact cells, in general, possess the potential to present antigen (22), they did not induce a CEA-specific T-cell response. Because this cell type only recently encountered its antigen, they might lack germinal center selection, which was also reflected in the lower percentages of CEA-specific B cells compared with BAPCs. A large fraction of tumor-associated B cells was CD86⁻. These cells consist mainly of CD21⁺CD86⁻ B cells, which probably correspond to naïve non–class-switched B cells in healthy individuals, but could also contain B cells with tumor-promoting properties (20). Isolation of antigen-specific B cells needs high amounts of patient-derived material from individuals with a response to the tested tumor antigen. Hence, experiments showing antigen specificity and antigen presentation were only possible in a small subset of included patients. This is a limitation of our work and further studies are needed to confirm antigen specificity of tumor-associated B cells. Usage of specific combinations of different fluorophores, as described for tetramer staining of T cells, could allow staining of multiple shared antigens and additional controls. Its establishment and application should be considered in future studies (36–38).

The importance of B cells as antigen-presenting cells is often doubted. However, several publications highlight their antigen-presenting capacity. In transgenic mice, B cells were shown to either enhance CD4⁺ T-cell responses when DCs are present or, depending on the type of antigen, B cells were the dominant antigen-presenting cell (39, 40). B cells can activate CD8⁺ and CD4⁺ T cells by presentation of antigen via HLA-I (cross-presentation) and HLA-II, respectively (20, 41–43). First evidence suggests that both mechanisms could be relevant in the context of antitumor immune responses. In patients with ovarian cancer, the predominant function of TLS-associated B cells was found to be antigen presentation (to CD8⁺ T cells), instead of production of antibodies (17), and in patients with gastric cancer, B cells isolated from TLSs have upregulated co-stimulatory molecules and MHC molecules and induce proliferation of CD4⁺ T cells (18). Especially in the context of immune checkpoint inhibition, studies have been published that underline the relevance of B cells, independent from their antibody-producing function. Checkpoint inhibition can induce proliferation of the CD21^lowIgD⁻ B-cell population, resulting in increased percentages of CD21^low B cells in tumor samples of patients with melanoma (1, 44). Activated B cells with high expression of MHC class II and hypermutated immunoglobulin genes are found to colocalize with CD8⁺ T cells in the tumor microenvironment (2, 17). These results are also supported by in vitro experiments, in which B cells enhance T-cell activity under PD-1 blockade after being stimulated with culture medium of autologous melanoma cells (5).

Taken together, we showed that in addition to the production of antibodies, presentation of antigens to T cells is an important mode of action of tumor-associated B cells across different types of cancer. Abundance of B cells and TLSs in the tumor microenvironment can influence clinical efficacy of immune checkpoint inhibition, and combined immunotherapies aiming to enhance B-cell responses appear promising.

C. Bruns reports personal fees from Medtronic and Promedicis, and grants from Intuitive and SIRTex outside the submitted work. H.A. Schlößer reports grants from German Research Foundation, German Cancer Aid (numbers 70113702 and 70113009), and European Union Regional Fund for Development (EFRE) during the conduct of the study, as well as other support from AstraZeneca and personal fees from Bristol Myers Squibb outside the submitted work. No disclosures were reported by the other authors.

K. Wennhold: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing. M. Thelen: Data curation, formal analysis, validation, investigation, methodology, project administration. J. Lehmann: Formal analysis, investigation, visualization, writing–review and editing. S. Schran: Formal analysis, investigation. E. Preugszat: Investigation. M. Garcia-Marquez: Conceptualization, supervision, investigation, writing–review and editing. A. Lechner: Formal analysis, investigation, writing–review and editing. A. Shimabukuro-Vornhagen: Conceptualization, writing–review and editing. M.S. Ercanoglu: Software, investigation, writing–review and editing. F. Klein: Conceptualization, methodology, writing–review and editing. F. Thangarajah: Validation, investigation, writing–review and editing. S. Eidt: Formal analysis, investigation, visualization, writing–review and editing. H. Löser: Data curation, investigation, visualization, methodology. C. Bruns: Conceptualization, investigation, writing–review and editing. A. Quaas: Conceptualization, formal analysis, validation, investigation, visualization, writing–review and editing. M. von Bergwelt-Baildon: Conceptualization, formal analysis, supervision, funding acquisition, validation, investigation, methodology, writing–review and editing. H.A. Schlößer: Conceptualization, resources, formal analysis, supervision, funding acquisition, validation, investigation, visualization, methodology, writing–original draft, writing–review and editing.

The authors thank the CECAD Imaging Facility and Peter Zentis for their support. They thank the technicians Wiebke Jeske and Pauline Volkmar for their support.

Research reported in this article was supported the German Cancer Aid (numbers 70113702 and 70113009), the Center for Molecular Medicine Cologne, the German Research Foundation (number 325827080), and the European Union Regional Fund for Development (EFRE; number 0801306).

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.