You are here

Article Text

Article menu
Pancreas
Original research
Mature tertiary lymphoid structures are key niches of tumour-specific immune responses in pancreatic ductal adenocarcinomas
  1. Gabriela Sarti Kinker1,
  2. Glauco Akelinghton Freire Vitiello1,
  3. Ariane Barros Diniz2,
  4. Mariela Pires Cabral-Piccin1,
  5. Pedro Henrique Barbosa Pereira1,
  6. Maria Letícia Rodrigues Carvalho1,
  7. Wallax Augusto Silva Ferreira1,3,
  8. Alexandre Silva Chaves1,
  9. Amanda Rondinelli1,
  10. Arianne Fagotti Gusmão1,
  11. Alexandre Defelicibus1,
  12. Gabriel Oliveira dos Santos4,
  13. Warley Abreu Nunes4,
  14. Laura Carolina López Claro5,
  15. Talita Magalhães Bernardo5,
  16. Ricardo Tadashi Nishio5,
  17. Adhemar Monteiro Pacheco5,
  18. Ana Carolina Laus6,
  19. Lidia Maria Rebolho Batista Arantes6,
  20. Julia Lima Fleck7,
  21. Victor Hugo Fonseca de Jesus8,
  22. André de Moricz5,
  23. Ricardo Weinlich2,
  24. Felipe José Fernandez Coimbra9,
  25. Vladmir Cláudio Cordeiro de Lima8,
  26. Tiago da Silva Medina1,10
  1. 1 International Research Center, A.C.Camargo Cancer Center, São Paulo, Brazil
  2. 2 Hospital Israelita Albert Einstein, São Paulo, Brazil
  3. 3 Evandro Chagas Institute, Ananindeua, Brazil
  4. 4 Department of Pathology, A.C.Camargo Cancer Center, São Paulo, Brazil
  5. 5 Faculty of Medical Sciences, Santa Casa de Misericórdia do Estado de São Paulo, São Paulo, Brazil
  6. 6 Molecular Oncology Research Center, Barretos Cancer Hospital, Barretos, Brazil
  7. 7 Mines Saint-Etienne, Univ Clermont Auvergne, CNRS, UMR 6158 LIMOS, Centre CIS, Saint-Etienne, France
  8. 8 Department of Medical Oncology, A.C.Camargo Cancer Center, São Paulo, Brazil
  9. 9 Department of Surgical Oncology, A.C.Camargo Cancer Center, São Paulo, Brazil
  10. 10 National Institute of Science and Technology in Oncogenomics and Therapeutic Innovation, São Paulo, Brazil
  1. Correspondence to Dr Tiago da Silva Medina, International Research Center, A.C.Camargo Cancer Center, São Paulo, 01508-010, Brazil; tiago.medina{at}accamargo.org.br

Abstract

Objective To better understand the immune microenvironment of pancreatic ductal adenocarcinomas (PDACs), here we explored the relevance of T and B cell compartmentalisation into tertiary lymphoid structures (TLSs) for the generation of local antitumour immunity.

Design We characterised the functional states and spatial organisation of PDAC-infiltrating T and B cells using single-cell RNA sequencing (scRNA-seq), flow cytometry, multicolour immunofluorescence, gene expression profiling of microdissected TLSs, as well as in vitro assays. In addition, we performed a pan-cancer analysis of tumour-infiltrating T cells using scRNA-seq and sc T cell receptor sequencing datasets from eight cancer types. To evaluate the clinical relevance of our findings, we used PDAC bulk RNA-seq data from The Cancer Genome Atlas and the PRINCE chemoimmunotherapy trial.

Results We found that a subset of PDACs harbours fully developed TLSs where B cells proliferate and differentiate into plasma cells. These mature TLSs also support T cell activity and are enriched with tumour-reactive T cells. Importantly, we showed that chronically activated, tumour-reactive T cells exposed to fibroblast-derived TGF-β may act as TLS organisers by producing the B cell chemoattractant CXCL13. Identification of highly similar subsets of clonally expanded CXCL13 + tumour-infiltrating T cells across multiple cancer types further indicated a conserved link between tumour-antigen recognition and the allocation of B cells within sheltered hubs in the tumour microenvironment. Finally, we showed that the expression of a gene signature reflecting mature TLSs was enriched in pretreatment biopsies from PDAC patients with longer survival after receiving different chemoimmunotherapy regimens.

Conclusion We provided a framework for understanding the biological role of PDAC-associated TLSs and revealed their potential to guide the selection of patients for future immunotherapy trials.

  • IMMUNE RESPONSE
  • PANCREATIC CANCER

Data availability statement

Data are available in a public, open access repository. Raw and processed gene expression data generated in this study are available at the Gene Expression Omnibus database (accession GSE226840). We collected publicly available scRNA-seq data from human PDAC,14 20–22 metastatic melanoma,26 non-small cell lung cancer,28 colorectal cancer29 and hepatocellular carcinoma,31 as well as paired scRNA-seq and scTCR-seq data from human endometrial cancer,27 non-small cell lung cancer,27 renal cell cancer,27colorectal cancer,27 squamous cell carcinoma30 and basal cell carcinoma.30

https://doi.org/10.1136/gutjnl-2022-328697

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    WHAT IS ALREADY KNOWN ON THIS TOPIC

    • Only a subset of pancreatic ductal adenocarcinoma (PDAC) patients may derive durable clinical benefit from available immunotherapies.

    • There is a crucial need to identify reliable biomarkers of treatment outcome.

    • Tumour-associated tertiary lymphoid structures (TLSs) predict favourable responses to immune checkpoint blockade in melanoma, soft tissue sarcoma, renal cell carcinoma, urothelial carcinoma, as well as in a pan-cancer cohort.

    WHAT THIS STUDY ADDS

    • Antitumour adaptive immunity can be generated/boosted at the PDAC site within mature TLSs, where B cells undergo antibody-affinity maturation and differentiate towards plasma cells.

    • Mature TLSs are key hubs of lymphocyte communication and support the activation and clonal expansion of tumour-reactive T cells.

    • Fibroblasts promote the expression of the B cell chemoattractant CXCL13 by chronically activated, tumour-reactive T cells in a TGF-β-dependent fashion.

    • CXCL13-producing T cells may support the organisation of mature TLS by guiding the spatial allocation of B cells.

    • PDAC patients with mature TLSs show improved survival after chemoimmunotherapy.

    HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

    • Mature TLSs may guide the design of biomarker-based immunotherapy trials for PDAC patients.

    • Further investigations of the cellular and molecular determinants that create a permissive milieu for TLS assembly and maturation could pave the way to combinatorial therapies that improve the efficacy of immunotherapy in PDACs.

    Introduction

    Pancreatic ductal adenocarcinoma (PDAC) is a leading cause of cancer-related death in Western societies, with a 5-year survival rate of only 10%.1 Despite recent improvements in chemotherapy regimens, surgery remains the only potentially curative option and most survivors belong to the 10%–20% of patients with resectable tumours.2 Over the past years, checkpoint inhibitor-based immunotherapies that boost T cell activity and reinvigorate antitumour immunity have shown remarkable success in a range of tumour types, transforming the standard of care.3 Nevertheless, only a subset of PDAC patients may derive durable clinical benefit from available immune checkpoint inhibitors,4 and identifying robust biomarkers of treatment response is an active area of research.

    Tumour-infiltrating B cells organised in aggregates surrounded by a mantle of T cells, known as tertiary lymphoid structures (TLSs), have been identified in a range of solid tumours, varying in density across tumour types and patients.5 Notably, recent studies reported that tumour-associated TLSs predict favourable responses to immune checkpoint blockade in melanoma, soft tissue sarcoma, renal cell carcinoma, urothelial carcinoma, as well as in a pan-cancer cohort.6–10 These findings may be broadly applicable to other cancer types, as TLSs can be privileged sites of tumour antigen presentation by local dendritic cells and B cells, and also of proliferation and differentiation of T and B cells.11

    Extensive profiling of main leucocyte lineages in PDACs by multiplex immunohistochemistry highlighted the presence of dense tumour-associated lymphoid aggregates primarily composed of expanded memory B cells, CD8 and CD4 T cells, consistent with TLSs.12 As immunosuppressive myeloid cells are a major component of the PDAC microenvironment,13 TLSs may represent sheltered niches for optimal lymphocyte activity. Here, to provide a framework for understanding the functional and clinical relevance of TLSs in PDACs, we dissected the interplay between PDAC-infiltrating T and B cells and investigated the impact of TLS maturation on tumour-specific immunity and patient prognosis.

    Results

    PDAC-infiltrating tumour-reactive T cells express the B cell chemoattractant CXCL13

    To explore the immune composition of the PDAC microenvironment, we mined previously published14 single-cell RNA sequencing (scRNA-seq) data of 24 treatment-naïve, primary human PDACs. By inferring large-scale chromosomal copy-number alterations (online supplemental figure S1A,B), we identified 9659 malignant and 30 657 non-malignant cells (ie, with normal karyotypes), which included four leucocyte populations (online supplemental figure S1C): myeloid cells (n=4708), T cells (n=2131), B cells (n=2441) and plasma cells (n=477). T cells were divided into two major clusters, composed of 909 CD8 and 1025 CD4 conventional (CD4conv) T cells, plus a third minor cluster of 383 regulatory T cells (Tregs) (online supplemental figure S2A). Analysis of transcriptional heterogeneity within CD8 and CD4conv T cell subsets identified robust gene expression programmes that spanned multiple T cell functional states (online supplemental figure S2B and 1A and tables S1–S2), as confirmed by cross-comparison to reference gene signatures (online supplemental figure S3A). States included naïve T cells (CD4conv-P1-Tn.1, CD4conv-P2-Tn.2, CD8-P1-Tn.like), resident memory T cells (CD4-P3-Trm, CD8-P2-Trm.1, CD8-P3-Trm.2), effector and effector memory T cells (CD8-P4-Teff and CD8-P5-Tem, respectively), interferon responses (CD4conv-P4-IFNresp), T cell migration (CD4conv-P5-migra), T cell exhaustion (CD4conv-P6-exh, CD8-P6-exh) and proliferative T cells (CD4conv-P7-cc, CD8-P7-cc). Inferring the transition of T cells between states using trajectory analysis15 further suggested a branched developmental path with cells expressing the T naïve programmes positioned at the opposite end of cells expressing exhaustion programmes (online supplemental figure S3B). Notably, exhaustion programmes of both CD8 and CD4conv T cells were highly similar to those previously found in T cells from multiple cancer types (online supplemental figure S3A), indicating a conserved phenotype likely triggered by chronic T cell receptor (TCR) stimulation. These programmes included multiple immune checkpoints and the tumour-reactivity marker ENTPD1 16,17 (encoding CD39). A subset of exhausted CD4conv T cells coexpressed genes of type I IFN response and cell migration, while some exhausted CD8 T cells coexpressed effector genes (figure 1B), indicating that they still exhibit some functionality.

    Supplemental material

    Figure 1
    Figure 1

    scRNA-seq analysis reveals exhausted PDAC-infiltrating T cells expressing the B cell chemoattractant CXCL13. (A) Heatmaps showing feature scores of selected genes from six expression programmes identified in PDAC-infiltrating CD8 (left) and CD4conv (right) T cells using scRNA-seq data and non-negative matrix factorisation (NMF). Expression programmes related to the cell cycle were omitted. (B) Hierarchical clustering of CD8 (top) and CD4conv (bottom) T cells based on the expression of gene programmes identified. Bar on top of the heatmap shows the expression of CXCL13. Black line indicates CXCL13 + exhausted T cells. (C) The number of CD8 (top) and CD4conv (bottom) T cells found in each tumour and the proportion of CXCL13 + exhausted T cells. PDAC, pancreatic ductal adenocarcinoma; scRNA-seq, single-cell RNA sequencing.

    Figure 2
    Figure 2

    Expression of CXCL13 is enriched in antigen-experienced PDAC-infiltrating T cells. (A) Differential expression analyses comparing CXCL13 + vs CXCL13 - CD8 (left) and CD4conv (right) T cells identified in PDAC scRNA-seq data. (B) Expression of IL7R and CXCL13 by CD8 (left) and CD4conv (right) T cells along the developmental trajectory inferred by Monocle (v2)15 using scRNA-seq data. Lines were fitted using LOESS regression. (C) Expression of gene signatures18 19 derived from tumour-reactive T cells in CXCL13 + vs CXCL13 - CD8 (left) and CD4conv (right) T cells identified in PDAC scRNA-seq data. (D) Flow cytometry analysis of the percentage of CD103+CD39+, PD1+CD39+ and CXCL13+ cells in CD8 (top) and CD4conv (bottom) T cells sorted from PDACs and paired peripheral blood mononuclear cells (PBMCs) samples (n=4), and stimulated overnight with anti-CD3 (5 µg/mL, plate-bound) and anti-CD28 (1 µg/mL). (E) Flow cytometry analysis of the percentage of CXCL13+ and CXCL13+GZMB+ cells in CD103-CD39- vs CD103+CD39+ and PD1-CD39- vs PD1+CD39+ CD8 (top) and CD4conv (bottom) T cells sorted from 5 PDACs and stimulated as in (D). Bars in (D) and (E) denote the mean. T-test (A, C); paired t-test (D, E). P values in (A) were corrected using the FDR procedure ****p<0.0001. FDR, false discovery rate; LOESS, locally estimated scatterplot smoothing; PDAC, pancreatic ductal adenocarcinoma; scRNA-seq, single-cell RNA sequencing.

    Figure 3
    Figure 3

    Identification of a conserved expression profile in CXCL13+ cells across human cancer types. (A) Heatmaps showing differentially expressed genes (absolute log2(fold change) >0.5, FDR-adjusted p<0.05) comparing CXCL13 + vs CXCL13 - tumour-infiltrating CD8 (left) CD4conv (right) T cells from different scRNA-seq datasets. Dataset labels are shown on the right. Genes differentially expressed in less than three datasets were omitted. (B) Combined PCA analysis of tumour-infiltrating CD8 (left) and CD4conv (right) T cells from scRNA-seq datasets in (A). Dot plots showing coordinates for PC1×PC3 with cells coloured by dataset (top) and by the CXCL13:IL7R expression ratio (bottom). (C) Analysis of paired scRNA-seq and scTCR-seq data showing the frequency of non-expanded (clone size=1) and expanded (clone size>1) tumour-infiltrating CD8 (left) and CD4conv (right) T cell clones expressing CXCL13 in different cancer datasets. (D) Co-expression of CXCL13 and the tumour-reactive marker ENTPD1 in non-expanded (clone size=1) and expanded (clone size>1) tumour-infiltrating CD8 (top) and CD4conv (bottom) T cell clones across datasets shown in (C). Summary plots on the left show percentages calculated for each dataset separately. Representative density plots on the right combine clones from all datasets analysed. Dots denote 300 clones randomly selected to illustrate each group. Bars and error bars show the mean and SE. T test (A); Fisher’s exact test (C); one-way repeated measures ANOVA (D). P values in (A) were corrected using the FDR procedure. ***p<0.001. ANOVA, analysis of variance; FDR, false discovery rate; PCA, principal component analysis; -seq, single-cell RNA sequencing; scTCR-seq, single-cell T cell receptor sequencing.

    Interestingly, expression of CXCL13, the main chemokine controlling the recruitment and organisation of B cells in lymphoid tissues, such as TLSs, was strongly enriched among CD8 and CD4conv T cells displaying a prominent exhaustion signal across patients (figure 1B,C). CXCL13 + CD8 T cells upregulated the expression of immune checkpoints, the effector molecules GZMB and IFNG, and IFN response genes; on the other hand, markers of naïve T cells and pre-exhaustion (IL7R and GZMK, respectively) had downregulated expression (figure 2A, online supplemental table S3). Similarly, CXCL13 + CD4conv T cells upregulated the expression of PDCD1, TIGIT, TNFRSF18 and the T cell differentiation transcription factors BATF and TOX, while the expression of IL7R, CCR7 and CD69 was significantly downregulated (figure 2A, online supplemental table S4). Ordering of cells in pseudotime underscored the opposite expression patterns of IL7R and CXCL13 along the inferred T cell developmental trajectory15 (figure 2B). More importantly, CXCL13 + CD8 and CD4conv T cells also upregulated the expression of gene signatures18 19 derived from tumour-infiltrating neoantigen-reactive T cells (figure 2C). Analysis of three additional PDAC scRNA-seq datasets20–22 further corroborated the differentiated phenotype of T cells expressing CXCL13 (online supplemental figure S4). To confirm these findings at the protein level by flow cytometry, we evaluated T cells expressing CD39 and PD1 or the tissue-resident marker CD103. These subsets, which are known to be enriched with tumour-reactive cells,16 23–25 were rare in the blood and abundant in the tumour microenvironment of PDAC patients (figure 2D). Compared with their double negative counterparts, PD1+CD39+ and CD103+CD39+ PDAC-infiltrating T cells had significantly greater expression of CXCL13 and GZMB in response to acute polyclonal TCR stimulation (figure 2E). Finally, we also showed that circulating T cells present a limited capacity to synthesise CXCL13 (figure 2D), indicating that intratumoural T cells have been primed locally to produce the B cell chemoattractant.

    Figure 4
    Figure 4

    Fibroblast-derived TGF-β promotes the expression of CXCL13 by chronically activated T cells in PDACs. (A, B) scRNA-seq analysis of the expression of TGFB genes by different cell types in PDACs. (A) Bubble plot showing the average expression of TGFB1/2/3 types and its correlation with two TLS signature10 35 scores calculated for each tumour as a pseudobulk. Only cellular populations with significant expression of TGFB genes are shown (see online supplemental figure S7A–C). (B) Aggregated expression of TGFB1/2/3 by fibroblasts in tumours grouped according to the percentage CXCL13 + T cells detected. (C) Flow cytometry analysis of CD8 (top) and CD4conv (bottom) T cells sorted from PBMCs of PDAC patients (n=4) and stimulated for 7 days with anti-CD3 (5 µg/mL, plate-bound) and anti-CD28 (1 µg/mL) alone or in the presence of fibroblast conditioned media and the TGF-β inhibitor SB431542. Primary fibroblast cultures were derived from one TLS- and one TLS+ PDAC. (D) H&E staining and immunohistochemistry staining of CD20 and pan-TGF-β in PDAC-associated TLSs. Fibroblast-rich areas surrounding TLSs are highlighted. Bars and error bars (C) denote the mean and SE, respectively. One-way ANOVA (B), paired T test (C). *p<0.05, **p<0.01. ANOVA, analysis of variance; PBMCs, peripheral blood mononuclear cells; PDAC, pancreatic ductal adenocarcinoma; scRNA-seq, single-cell RNA sequencing; TLS, tertiary lymphoid structure.

    Antigen recognition is linked to the expression of CXCL13 by T cells across human cancer types

    To examine whether CXCL13 may be associated with similar T cell states across human cancers, we used published scRNA-seq datasets from tumour-infiltrating lymphocytes of metastatic melanoma,26 endometrial cancer,27 non-small cell lung cancer,27 28 renal cell cancer,27 colorectal cancer,27 29 squamous cell carcinoma,30 basal cell carcinoma,27 and hepatocellular carcinoma,31 encompassing more than 150 000 cells. For CD8 and CD4conv T cells in each dataset, we generated a CXCL13 transcriptional signature by comparing CXCL13 + to CXCL13 - cells. Remarkably, there was an extensive overlap between CXCL13 signatures across tumour types (online supplemental figure S5A), which included multiple molecules that are induced by chronic TCR activation. Overall, CXCL13 + T cells upregulated the expression of inhibitory receptors, the transcription factor TOX, and the late activation markers GZMB and IFNG (figure 3A, online supplemental tables S5 and S6). Among downregulated genes were markers of T naïve (eg, IL7R and S1PR1) and pre-exhaustion (ie, GZMK; figure 3A, online supplemental tables S7 and S8). The relevance of CXCL13 expression by exhausted T cells was further demonstrated by analysing CD8 and CD4conv T cells from all datasets combined (figure 3B). Principal component analysis (PCA) revealed a naïve-to-exhaustion gradual trajectory shared across tumour types, with PC1 and PC3 primarily reflecting the IL7R-CXCL13 expression axis (figure 3B and online supplemental figure S5B).

    Figure 5
    Figure 5

    Characterisation of the CXCL13-CXCR5 axis in PDACs. (A) scRNA-seq analysis of the expression of CXCL13 (top) and CXCR5 (bottom) by different cell types in PDACs. (B) CXCL13-CXCR5 ligand-receptor interaction likelihood between pairs of cell types inferred by CellphoneDB36 using scRNA-seq data. (C) Dot plots showing the likelihood of all potential interaction where the ligand is expressed in CD8/CD4conv T cells and the receptor in B cells. Immunofluorescence staining of (D) CD20 (red), CD3 (green), PD1 (magenta), CD39 (yellow) and DAPI or (E) CD20 (red), CD8 (green), CD4 (cyan), CXCL13 (magenta), CXCR5 (yellow) in PDAC-associated TLSs. White arrows indicate CD3+PD1+CD39+ T cells. (F) Heatmap showing the percentage of different cell types in each PDAC, quantified by scRNA-seq. Samples were ordered according to the abundance of T and B cells. Cell types were ordered according to the correlation between their abundance and the abundance of B cells. Only CD8 and CD4conv T cells showed significant correlations (p<0.05) with B cell abundance. Samples were classified as having low or high abundance of T and B cells. (G) Pseudobulk expression of two TLS gene signatures10 35 in LT and LB low versus high tumours. Pearson’s correlation (F); T-test (G). **p<0.01. LB, B lymphocytes; LT, T lymphocytes; PDAC, pancreatic ductal adenocarcinoma; scRNA-seq, single-cell RNA sequencing; TLSs, tertiary lymphoid structures.

    Tracking T cell clones in endometrial cancer,27 non-small cell lung cancer,27 renal cell cancer,27 squamous cell carcinoma30 and basal cell carcinoma30 using paired scRNA-seq and scTCR-seq data (~100 000 cells) further corroborated the association between antigen recognition, T cell activation and CXCL13 expression. We observed that the expression of CXCL13 progressively increased as cells expanded (figure 3C): 56% of T CD8 and 82% of T CD4conv clones under high expansion (>5 cells) were CXCL13+, with a significantly smaller number (~20%) observed for non-expanded ones. Expanded clones also frequently coexpressed CXCL13 and ENTPD1 (encoding CD39), consistent with tumour reactivity (figure 3D). Notably, the expression pattern of CXCL13 and ENTPD1 gradually changed as clone size increased. In non-expanded clones, CXCL13 and ENTPD1 were expressed independently, but as clones expanded, ENTPD1 + T cells gained the expression of CXCL13 (figure 3D). CXCL13 transcripts were detected in more than 85% of highly expanded clones (>5 cells) that expressed ENTPD1. Altogether, these findings suggest that our observations in PDACs generalise to other cancer types. Importantly, the ability of tumour-reactive T cells to express CXCL13 may lay at the core of TLS formation and the mounting of adaptive immune responses directly at the tumour site.

    Continuous TCR engagement promotes the expression of CXCL13 by PDAC T cells exposed to fibroblast-derived TGF-β

    Previous reports32–34 demonstrate that T cells can produce CXCL13 in response to sustained TCR triggering with concurrent activation of TGF-β receptors. Accordingly, 7d polyclonal TCR stimulation in the presence of recombinant TGF-β induced the expression of CXCL13 by CD4conv and CD8 T cells from the peripheral blood of PDAC patients (online supplemental figure S6A–C, E). CXCL13+ CD8 and CD4conv T cells preferentially expressed PD1, CD39 and CD103 (online supplemental figure S6A,B,D,F). As in the pan-cancer clonotype analysis (figure 3C,D), we also observed that the expression of CXCL13, as well as the coexpression of CXCL13 with PD1, CD39 or CD103, progressively increased with the number of mitotic divisions, reaching maximum levels after three cycles and decreasing around six cycles (online supplemental figure S6G).

    Figure 6
    Figure 6

    scRNA-seq analysis reveals PDAC-infiltrating B cells under GC reaction. (A) UMAP plot of B lineage cells from PDACs coloured by the expression of markers of follicular (SELL, CCR7), GC (AICDA, BCL6) and plasma (MZB1, XBP1) cells. Relative expression of additional B lineage markers is shown in (B). (C) PCA analysis of GC B cells showing PC1×PC2 coordinates with cells coloured by the expression of GC light and dark zones signatures45 (left), as well as light zone (CD83) and dark zone (AICDA) markers (right). Radial projection of PC1×PC2 cartesian coordinates highlights a circular trajectory. (D) K-means clustering splitting cells into light, dark and intermediate GC zones. (E) Expression of dark zone (top) and light zone (bottom) markers by GC B cells in different K-means clusters. GC, germinal centre; PCA, principal component analysis; PDAC, pancreatic ductal adenocarcinoma; scRNA-seq, single-cell RNA sequencing.

    Since TGF-β is abundantly produced by fibroblasts in the PDAC microenvironment (online supplemental figure S7A–C), we hypothesised that it might play a noncanonical role coordinating the production of CXCL13 by tumour-reactive T cells and, consequently, the organisation of tumour-associated TLSs. Analysis of PDAC scRNA-seq data revealed a significant correlation between the expression of TGFB1 and TGFB3 by fibroblasts and the score of TLS gene signatures10 35 calculated for each tumour (figure 4A). Similarly, expression of TGF-β genes by fibroblasts was increased in tumours with higher percentage of CXCL13 + T cells (figure 4B), while no significant association was observed for other TGF-β-producing cell types such stellate and myeloid cells (online supplemental figure S7D). In line with that, TCR engagement along with conditioned media from fibroblasts of a TLS-containing tumour (online supplemental figure S7E) induced, in a TGF-β-dependent manner, the expression of CXCL13 by CD8 and CD4conv T cells from the peripheral blood of PDAC patients (figure 4C). CXCL13 was again coexpressed with CD39, reinforcing the synergic effect of chronic TCR stimulation and TGF-β receptor signalling. Importantly, no effect was observed with conditioned media from fibroblasts of a TLS-absent tumour (figure 4C; online supplemental figure S7E). Last, we demonstrated that TGF-β can be found both within tumour-associated TLSs, with a scattered distribution, and in the surrounding fibroblast-rich connective tissue (figure 4D, online supplemental figure S7F). Collectively, these observations indicate that the organisation and maintenance of TLSs are linked to sustained antigen recognition by T cells in a TGF-β-rich tumour microenvironment, underscoring the potential of such structures as sites where antitumour adaptive immune responses can be generated and/or boosted.

    Figure 7
    Figure 7

    Gene expression profile of PDAC-associated TLSs in different maturation stages. (A) Representative H&E staining and immunohistochemistry staining of CD20 and CD23 in PDAC-associated TLSs. TLSs were classified according to the presence of CD23+ mFDCs. (B) Detection of FCER2 (encoding CD23) expression in normal pancreatic tissue and TLSs with different maturation stages. (C) Gene modules with distinct expression patterns throughout the maturation process, identified using K-means clustering. (D) Heatmap showing the average relative expression of genes in each module by normal pancreatic tissue and CD23neg/int/high TLSs, with selected genes labelled. (E) Functional annotation of genes in each module. (F) Expression of gene signatures18 19 derived from neoantigen-reactive T cells and pan-cancer CXCL13 programmes (online supplemental tables S6 and S7) by normal pancreatic tissue and CD23neg/int/high TLSs. (G) Dot plot showing gene expression levels in CD23neg (x-axis) and CD23high (y-axis) TLSs relative to normal pancreatic tissue (log2(fold change)). Genes are coloured according to the difference in log2(fold change) observed in CD23high vs CD23neg TLSs. Relevant genes composing the signatures shown in (F) are labelled. (H) Number of inferred ligand-receptor interactions in CD23neg/int/high (left) and the proportion of ligands and receptors (right) among genes with greater expression (log2(fold change) >1, p<0.05, t test) in each TLS group compared with normal pancreatic tissue. (I) scRNA-seq analysis of the expression of ligands and receptors detected in CD23high TLSs by PDAC-infiltrating T and B cells. Lines indicate ligand-receptor pairs. Boxplots (F) show the median and IQR with whiskers extending from the fifth to the 95th percentile. Fisher’s exact test (B, H); t test (F). *p<0.05, **p<0.01, ***p<0.001. mFDCs, mature follicular dendritic cells; PDAC, pancreatic ductal adenocarcinoma; scRNA-seq, single-cell RNA sequencing; TLSs, tumour-infiltrating lymphocytes.

    CXCL13-producing T cells as TLS organisers in PDACs

    Evaluation of PDAC scRNA-seq data demonstrated that CD8/CD4conv T cells and B cells are the main cellular populations expressing CXCL13 and its cognate receptor CXCR5, respectively (figure 5A). As cell-cell communication analysis36 indicated a high likelihood of interaction between CD8/CD4conv T cells expressing CXCL13 and B cells expressing CXCR5 (figure 5B,C), we hypothesised that CXCL13+ tumour-reactive T cells play a central role in the spatial organisation of tumour-associated TLSs. Multicolour immunofluorescence of PDACs confirmed the presence of PD1+CD39+CD3+ T cells within well-structured tumour-associated TLSs (figure 5D and online supplemental figure S8A). More importantly, TLSs showed an enriched expression of CXCR5 in the B cell zone, while CXCL13 was abundantly distributed throughout the aggregate (figure 5E and online supplemental figure S8B). This pattern is consistent with the diffusion model described in secondary lymphoid organs where CXCL13 binds to the extracellular matrix, creating short sharp gradients proximal to CXCL13-secreting cells that promote B cell trafficking.37

    Accordingly, we found a significant correlation between the intratumoural abundance of CD8/CD4conv T cells and B cells in the PDAC scRNA-seq data (figure 5F), with highly infiltrated tumours (T/B cell high) showing an increased score of TLS gene signatures10 35 (figure 5G). Similar results were obtained by flow cytometry analysis of 14 treatment-naïve, primary PDACs (online supplemental table S9), with T/B cell high patients presenting an increased density of TLSs in H&E-stained slides of surgical specimens (online supplemental figure S9A,B). The abundance of CD8 T cells, CD4 T cells and B cells in T/B cell high tumours was significantly higher than in adjacent normal tissues, whereas no differences were observed for T/B cell low counterparts (online supplemental figure S9C). CD45+ leucocytes in T/B cell high tumours were also enriched for lymphocytes relative to other immune cell types when compared with T/B cell low tumours (online supplemental figure S9D). Analysis of H&E-stained slides from an independent cohort of 37 treatment-naïve, primary PDACs revealed that TLSs were present in 86% of specimens, with densities ranging from 1.25% to 10% of tumour area (online supplemental figure S9E,F and table S10). TLS densities greater than 5% were observed in 19% of cases, corroborating previous reports12 38 39 that PDACs are not uniformly depleted of lymphocytes. When possible (n=13), we also evaluated adjacent normal tissues, and TLSs were observed in only 15% of cases, consistent with the low abundance of T and B cells quantified by flow cytometry.

    Mature TLSs support antitumour adaptive immunity in PDACs

    Accumulating evidence indicates that tumour-associated TLSs exist under different functional states, according to their structural similarity to secondary lymphoid organs.8 40–42 Fully mature TLSs resemble secondary follicles, presenting well-segregated T and B cell zones and active germinal centres (GCs) with mature follicular dendritic cells (mFDCs). GCs are specialised compartments in lymphoid structures where antigen-activated B cells proliferate, undergo antibody affinity maturation, and differentiate towards antibody-secreting plasma cells. Accordingly, analysis of PDAC scRNA-seq data revealed three main subsets of B lineage cells based on the detection of known markers43: follicular B cells (n=1847), GC B cells (n=593) and plasma cells (n=477) (figure 6A,B). Within GCs, B cells transit cyclically from the dark zone, where they undergo somatic hypermutation (SHM) and immunoglobulin class switch recombination (CSR) of B cell receptors (BCRs), to the light zone, where they interact with FDCs and helper T cells for selection of BCR variants with high affinity to the foreign antigen and deletion of non-reactive clones.44 Notably, by applying PCA to PDAC GC B cells, we identified a characteristic continuum of gene expression states spanning the dark-light zone axis45 (figure 6C). The cyclic transitions found in PDAC GC B cells (figure 6C) were highly consistent with reports from B cells in human tonsil and spleen.46 Dark and light zone cells were oppositely distributed along a circular trajectory and were interpolated by cells in transitional states (figure 6D). Dark zone B cells upregulated the expression of AICDA and CXCR4, while light zone B cells upregulated BCL2A1, CD83 and MYC. Cycling cells were enriched in intermediate zones (figure 6E), as previously described.46

    Next, to understand the role of different TLS functional states in intratumoural immune responses, we used laser-capture microdissection to evaluate the expression profile of PDAC-associated TLSs in three maturation stages8 40–42 (figure 7A,B): (1) the initial stage consisting of dense aggregates of CD20+ B cells with no evidence of CD23+ mFDCs (CD23neg TLSs); (2) primary follicle-like TLSs containing scattered CD23+ mFDCs (CD23int TLSs); and (3) fully mature, secondary follicle-like TLSs displaying GCs with a meshwork of CD23+ mFDCs (CD23high TLSs). K-means clustering identified three main groups of genes with distinct expression patterns throughout the maturation process (figure 7C, online supplemental table S11): early genes that had their expression upregulated in all three maturation stages compared with normal pancreatic tissue; intermediate genes that showed increased expression in CD23int and CD23high TLSs; and late genes that were preferentially expressed in CD23high TLSs (figure 7C,D). Consistent with previous reports on the cellular dynamics of TLS maturation,47 early genes included markers of B cells (eg, MS4A1 and CD79B), T cells (eg, ZAP70 and ITK) and FDCs (CR1 and FCGR2B), as well as molecules involved in lymphocyte survival (eg, BCL2, IL7R), antigen presentation (eg, HLA-DMB and HLA-DRB3) and lymphoid tissue induction (eg, CCL19 and LTB). On the other hand, markers of GC B cells (eg, IL21R and POU2AF1) and FDC maturation (eg, LAMP3 and XCR1) were only differentially expressed in CD23high TLSs (figure 7D–E). Notably, mature TLSs showed increased expression of gene signatures derived from neoantigen-reactive tumour-infiltrating T cells18 19 and CXCL13 + tumour-infiltrating T cells from our pan-cancer analysis (figure 7F). These signatures included genes such as CXCL13, ENTPD1, HAVCR2, TNFRSF18 and GZMA/B (figure 7G, online supplemental tables S5 and S6), supporting the hypothesis that T cells recognising tumour epitopes lay on the core of TLS spatial organisation, creating a privileged niche to propagate effective antitumour responses. Accordingly, molecules associated with chemoattraction, T/B cell activation, type I IFN responses and the complement system were highly enriched among intermediate/late genes. In addition, we demonstrated that mature TLSs have a higher number of predicted ligand-receptor interactions, which indicates that such lymphoid aggregates are central hubs of cell-to-cell communication (figure 7H–I).

    Clinical relevance of PDAC-associated TLSs

    To validate our findings in a larger cohort, we analysed bulk RNA-seq data from 147 primary PDACs in The Cancer Genome Atlas (TCGA) cohort48 using the microenvironment cell populations (MCP)-counter algorithm49 to estimate the abundance of different immune cell types, as well as fibroblast and endothelial cells. Hierarchical clustering based on MCP-counter estimates revealed three main groups of tumours (clusters A, B and C, figure 8A) that have partial overlap with previous PDAC molecular classifications (online supplemental figure S10A). Cluster A reflected malignant cell-rich tumours with high cellularity (online supplemental figure S10A), while cluster C had the highest abundance of leucocytes, especially B lineage cells. An elevated abundance of fibroblast was observed in approximately half of PDACs in cluster B and C. Cluster C tumours also presented increased expression of markers of T cell exhaustion and TLS formation/maturation (eg, CXCL13 and CXCR5; figure 8B), higher incidence of TLSs in H&E-stained slides (figure 8C and online supplemental figure S10A), and a more diverse and clonally expanded repertoire of TCRs and BCRs (figure 8D). We did not identify statistically significant associations between PDAC immune subtypes and the anatomic subdivision of tumours, tumour stage and grade, incidence of point mutations and copy number variations in main driver genes (figure 8A), microsatellite instability status (online supplemental figure S10A) and tumour mutational load (online supplemental figure S10B). Accordingly, modelling of PDAC progression, based on the incidence of non-synonymous mutations in top driver genes together with the expression of immune-related genes/signatures, suggested that TLS neogenesis occurs in early stages of tumourigenesis (online supplemental figure S10C).

    Figure 8
    Figure 8

    Clinical relevance of mature TLSs in PDACs. (A) Abundance of different immune cell types in TCGA PDAC tumours (n=147) estimated by MCP-counter49 using bulk RNA-seq data. Hierarchical clustering depicts three main immune subtypes with distinct levels of lymphocytic infiltration. Immune subtypes are independent (p>0.05) of anatomic subdivision, tumour stage/grade and mutations/copy number variations in main driver genes. (B) Expression of genes related to T cell exhaustion and the formation/maturation of TLSs, and (C) the density of tumour-associated TLSs detected in H&E-stained sections are increased (p<0.05) in cluster C. (D) Clonal count (top) and clonal expansion rate (bottom) of TCR (TRA/TRB chains) and BCR (IgH/IgL chains) sequences in each immune subtype. (E) Representative images of H&E-stained sections from PDACs with immature (left) and mature (right) TLSs. PDACs with mature TLSs were used to derive a GC-specific signature. The circle indicates a GC. (F) GC signature score in PDACs with different degrees of B cell differentiation, as defined by the occurrence of SHM and CSR in IgH sequences. (G) Prognostic significance of indicated signatures evaluated using multivariate Cox regression adjusting for gender, age, tumour grade and stage (left). Kaplan-Meier survival curve and log-rank test comparing patients with high versus low GC signature scores (right). (H) GSEA of the GC signature comparing bulk RNA-seq of tumour samples from PDAC patients that survived >1 year vs <1 year after receiving different chemoimmunotherapy regimens (PRINCE trial).4 (I) GC signature scores in bulk RNA-seq samples from patients that survived >1 year vs <1 year. Patients are coloured by treatment regimens as in (H). (J) Kaplan-Meier survival curves and log-rank test comparing patients with high versus low GC signature scores. Boxplots (D, F, I) show the median and IQR with whiskers extending from the fifth to the 95th percentile. Fisher’s exact test (A, C); one-way ANOVA (B, D, F); logistic regression correcting for biopsy site and treatment regimen (I). *p<0.05, ***p<0.001. ANOVA, tertiary lymphoid structures; BCR, B cell receptor; CSR, class switch recombination; GC, germinal centre; PDACs, pancreatic ductal adenocarcinomas; SHM, somatic hypermutation; TCGA, The Cancer Genome Atlas; TCR, T cell receptor; TLSs, tertiary lymphoid structures.

    Detailed examination of TCGA pathology slides revealed that 25% of tumours in cluster C harboured mature TLS with GC/GC-like structures (figure 8C,E). Since the accurate identification of such structures in clinical routine can be limited due to varying degrees of tissue integrity and preservation, as well as the heterogeneous tissue architecture throughout the tumour bulk, we aimed to define GCs molecularly and infer their presence using gene expression data. To this end, we identified co-expressed genes that specifically reflected the maturation of TLSs, while disregarding signals coming from unrelated inflammation and scattered leucocyte infiltration (online supplemental methods). The signature included B cell markers such as CD79A, MS4A1 (encoding CD20) and SELL; CXCL13 and CXCR5; and GC markers AICDA and IL21R (online supplemental table S12). To validate the correlation between GC signature scores and TLS functionality, we used IgH CDR3 nucleotide sequences assembled by TRUST450 to infer the rates of BCR SHM and Ig CSR in clonal B cells, the two main processes required for the production of high-affinity antibodies. PDAC samples where both SHM and CSR events were detected showed significantly higher GC scores than those with no clonal B cells, with clonal B cells but without SHM and CSR, and with clonal B cells and SHM only (figure 8F). In addition, increased GC signature scores were associated with improved patient overall survival, independent of age, sex, tumour grade and stage, demonstrating more prognostic power than the estimated abundance of B lineage cells and T cells (figure 8G).

    Finally, to test the potential of our GC signature to guide clinical decision-making, we explored a bulk RNA-seq dataset (PRINCE trial)4 from pretreatment biopsies of first-line metastatic PDAC patients that received nivolumab (nivo; anti-PD-1) and/or sotigalimab (sotiga; CD40 agonistic antibody) with gemcitabine/nab-paclitaxel (chemotherapy). The expression of GC signature genes was significantly enriched in patients that survived more than 1 year across all three treatment settings (figure 8H–I). More importantly, patients with high GC score showed longer overall survival (figure 8J), independent of de novo/recurrent staging at initial diagnosis, prior chemotherapy usage and biopsy site (online supplemental tables S13–S15). These observations indicate that by blocking T cell exhaustion and activating antigen-presenting cells and B cells, respectively, anti-PD1 and CD40 agonistic antibodies may boost the generation of adaptive immunity in pre-existing, mature tumour-associated TLSs.

    Discussion

    As several recent studies demonstrated the value of TLSs as predictors of response to immune checkpoint inhibitors,6–10 deciphering the full repertoire of antitumour immunity orchestrated within these lymphoid structures has become an area of increasing investigation. Here, we corroborated previous findings12 38 39 51 showing that, despite the myeloid cell-rich, immune suppressive ecosystem of many PDACs, a significant subset of tumours is enriched with activated T and B cells organised in TLSs. We also demonstrated that PDAC-associated TLSs can develop into fully mature structures containing antigen-experienced CXCL13-producing T cells that may function as lymphoid tissue organisers. Notably, our findings underscore that B cell proliferation, SHM, and Ig CSR actively take place in GCs within mature TLSs, emphasising that adaptive immune responses can also be generated or boosted at the tumour site. In addition, identification of highly similar subsets of clonally expanded CXCL13 + T cells across multiple cancer types indicates a conserved link between tumour-antigen recognition and lymphocyte compartmentalisation into sheltered hubs.

    Consistent with previous observations,32–34 our findings showed that the expression of CXCL13 by both CD8 and CD4conv T cells depends on two main factors: chronic TCR activation and TGF-β signalling. TGF-β, classically viewed as an immunosuppressive cytokine, is also a potent inducer of the tissue-homing molecule CD103 in activated T cells.16 52 53 Given its abundance in most tumour microenvironments, we hypothesise that TGF-β is critical for the retention of tumour-reactive T cells and the development of ectopic lymphoid tissues, a molecular mechanism that seems to be conserved across cancer types. Despite the limited number of samples analysed, our data also indicate that TGF-β-producing fibroblasts play a central role in promoting the production of CXCL13 by PDAC T cells and are enriched in TLS+ tumours, supporting further investigations of the crosstalk between fibroblast subpopulations and local immune responses. Accordingly, tumour-associated fibroblasts with features of lymphoid tissue organisers were shown to orchestrate the development of TLSs in murine melanoma,54 while spatial transcriptomics of renal clear cell carcinomas revealed a high expression of a fibroblast signature in TLS areas.55

    Although TLSs have been found in many cancer types,5 their functional capacity under different maturation stages remains poorly understood. Here, we demonstrated that fully developed aggregates are key niches of B cell differentiation towards plasma cells, which may culminate with the local production of antitumour antibodies that can cover antigen-expressing cells for opsonin-mediated phagocytosis, antibody-dependent cellular cytotoxicity and complement cascade lysis, as elegantly demonstrated for renal clear cell carcinoma.55 Notably, our data also indicate a connection between the compartmentalisation of lymphocytes, which promotes optimal activation and clonal expansion of tumour-reactive T cells, and the expression of CXCL13 by chronically activated, tumour-reactive T cells, which in turn contributes to the spatial allocation of B cells and the maturation of TLSs.

    By analysing paired RNA-seq data and H&E-stained sections from TCGA PDAC patients, we defined a refined gene signature that specifically reflects tumours harbouring mature GC+ TLSs. Such molecular definition of TLS allows us to infer their presence more accurately using gene expression data, bypassing the need of high-quality pathology slides from surgical specimens. Importantly, analysis of pretreatment biopsies from metastatic PDACs showed that the GC signature score associates with survival after chemotherapy with CD40 agonist and/or anti-PD1, indicating that mature TLSs may guide selection of patients for biomarker-based immunotherapy trials. In line with that, lymphoid tissue neogenesis at the tumour site is suggested to facilitate adaptive antitumour immunity through different mechanisms,56 including bypassing the trafficking of dendritic cells and lymphocytes to and from secondary lymphoid organs, and increasing the likelihood of encounters between tumour-associated antigens and rare matching lymphocytes.

    Altogether, we provided an in-depth view of PDAC-infiltrating lymphocytes, from cell spatiality to their network of interaction, revealing at unprecedented depth the benefit of lymphocyte compartmentalisation within TLSs for antitumour immunity and immunotherapy response, especially when they reach a stage of complete maturation. Our work also supports further investigations of the cellular and molecular determinants that create, in a subset of PDACs, a permissive milieu for lymphocyte infiltration and TLS assembly. This knowledge could pave the way to combinatorial therapies aiming to reshape suppressive microenvironments and induce the de novo formation and maturation of tumour-associated TLSs to improve responses to immunotherapy in PDAC patients.

    Materials and methods

    Described in online supplemental material.

    Data availability statement

    Data are available in a public, open access repository. Raw and processed gene expression data generated in this study are available at the Gene Expression Omnibus database (accession GSE226840). We collected publicly available scRNA-seq data from human PDAC,14 20–22 metastatic melanoma,26 non-small cell lung cancer,28 colorectal cancer29 and hepatocellular carcinoma,31 as well as paired scRNA-seq and scTCR-seq data from human endometrial cancer,27 non-small cell lung cancer,27 renal cell cancer,27colorectal cancer,27 squamous cell carcinoma30 and basal cell carcinoma.30

    Ethics statements

    Patient consent for publication

    Ethics approval

    This study involves human participants and was approved by Institutional Research Ethics Committee (ID 2712/19). Participants gave informed consent to participate in the study before taking part.

    Acknowledgments

    We thank Ana Paula M. S. Silva, Keila T. Macedo, Nathalia V. Araujo, Severino S. Ferreira, Fernanda A. Pintor, Clara M. Cavalcanti and Thiago P. A. Aloia for the technical assistance. Frozen PDAC samples were provided by the A. C. Camargo Cancer Center Biobank.

    References

    1. ↵
      2. Siegel RL ,
      3. Miller KD ,
      4. Jemal A
      . Cancer statistics, 2018. CA Cancer J Clin 2018;68:7–30. doi:10.3322/caac.21442
      OpenUrlCrossRefPubMed
    2. ↵
      2. Gillen S ,
      3. Schuster T ,
      4. Meyer Zum Büschenfelde C , et al
      . Preoperative/neoadjuvant therapy in pancreatic cancer: a systematic review and meta-analysis of response and resection percentages. PLoS Med 2010;7:e1000267. doi:10.1371/journal.pmed.1000267
    3. ↵
      2. Pardoll DM
      . The blockade of immune checkpoints in cancer immunotherapy. Nat Rev Cancer 2012;12:252–64. doi:10.1038/nrc3239
      OpenUrlCrossRefPubMedWeb of Science
    4. ↵
      2. Padrón LJ ,
      3. Maurer DM ,
      4. O’Hara MH , et al
      . Sotigalimab and/or nivolumab with chemotherapy in first-line metastatic pancreatic cancer: clinical and immunologic analyses from the randomized phase 2 Prince trial. Nat Med 2022;28:1167–77. doi:10.1038/s41591-022-01829-9
      OpenUrlCrossRef
    5. ↵
      2. Dieu-Nosjean M-C ,
      3. Giraldo NA ,
      4. Kaplon H , et al
      . Tertiary lymphoid structures, drivers of the anti-tumor responses in human cancers. Immunol Rev 2016;271:260–75. doi:10.1111/imr.12405
      OpenUrlCrossRefPubMed
    6. ↵
      2. Helmink BA ,
      3. Reddy SM ,
      4. Gao J , et al
      . B cells and tertiary lymphoid structures promote immunotherapy response. Nature 2020;577:549–55. doi:10.1038/s41586-019-1922-8
      OpenUrlCrossRefPubMed
    7. ↵
      2. Petitprez F ,
      3. de Reyniès A ,
      4. Keung EZ , et al
      . B cells are associated with survival and immunotherapy response in sarcoma. Nature 2020;577:556–60. doi:10.1038/s41586-019-1906-8
      OpenUrlCrossRefPubMed
    8. ↵
      2. Vanhersecke L ,
      3. Brunet M ,
      4. Guégan J-P , et al
      . Mature tertiary lymphoid structures predict immune checkpoint inhibitor efficacy in solid tumors independently of PD-L1 expression. Nat Cancer 2021;2:794–802. doi:10.1038/s43018-021-00232-6
      OpenUrl
    9. ↵
      2. Gao J ,
      3. Navai N ,
      4. Alhalabi O , et al
      . Neoadjuvant PD-L1 plus CTLA-4 blockade in patients with cisplatin-ineligible operable high-risk urothelial carcinoma. Nat Med 2020;26:1845–51. doi:10.1038/s41591-020-1086-y
      OpenUrlPubMed
    10. ↵
      2. Cabrita R ,
      3. Lauss M ,
      4. Sanna A , et al
      . Tertiary lymphoid structures improve immunotherapy and survival in melanoma. Nature 2020;577:561–5. doi:10.1038/s41586-019-1914-8
      OpenUrlPubMed
    11. ↵
      2. Kinker GS ,
      3. Vitiello GAF ,
      4. Ferreira WAS , et al
      . B cell orchestration of anti-tumor immune responses: a matter of cell localization and communication. Front Cell Dev Biol 2021;9:678127. doi:10.3389/fcell.2021.678127
      OpenUrl
    12. ↵
      2. Liudahl SM ,
      3. Betts CB ,
      4. Sivagnanam S , et al
      . Leukocyte heterogeneity in pancreatic ductal adenocarcinoma: phenotypic and spatial features associated with clinical outcome. Cancer Discov 2021;11:2014–31. doi:10.1158/2159-8290.CD-20-0841
      OpenUrlAbstract/FREE Full Text
    13. ↵
      2. Karamitopoulou E
      . Tumour microenvironment of pancreatic cancer: immune landscape is dictated by molecular and histopathological features. Br J Cancer 2019;121:5–14. doi:10.1038/s41416-019-0479-5
      OpenUrlCrossRefPubMed
    14. ↵
      2. Peng J ,
      3. Sun B-F ,
      4. Chen C-Y , et al
      . Single-Cell RNA-seq highlights intra-tumoral heterogeneity and malignant progression in pancreatic ductal adenocarcinoma. Cell Res 2019;29:725–38. doi:10.1038/s41422-019-0195-y
      OpenUrlCrossRefPubMed
    15. ↵
      2. Qiu X ,
      3. Mao Q ,
      4. Tang Y , et al
      . Reversed graph embedding resolves complex single-cell trajectories. Nat Methods 2017;14:979–82. doi:10.1038/nmeth.4402
      OpenUrlCrossRefPubMed
    16. ↵
      2. Duhen T ,
      3. Duhen R ,
      4. Montler R , et al
      . Co-expression of CD39 and CD103 identifies tumor-reactive CD8 T cells in human solid tumors. Nat Commun 2018;9:2724. doi:10.1038/s41467-018-05072-0
    17. ↵
      2. Lee YJ ,
      3. Kim JY ,
      4. Jeon SH , et al
      . CD39 + tissue-resident memory CD8 + T cells with a clonal overlap across compartments mediate antitumor immunity in breast cancer. Sci Immunol 2022;7:74. doi:10.1126/sciimmunol.abn8390
      OpenUrl
    18. ↵
      2. Zheng C ,
      3. Fass JN ,
      4. Shih Y-P , et al
      . Transcriptomic profiles of neoantigen-reactive T cells in human gastrointestinal cancers. Cancer Cell 2022;40:410–23. doi:10.1016/j.ccell.2022.03.005
      OpenUrlCrossRef
    19. ↵
      2. Lowery FJ ,
      3. Krishna S ,
      4. Yossef R , et al
      . Molecular signatures of antitumor neoantigen-reactive T cells from metastatic human cancers. Science 2022;375:877–84. doi:10.1126/science.abl5447
      OpenUrlCrossRefPubMed
    20. ↵
      2. Lin W ,
      3. Noel P ,
      4. Borazanci EH , et al
      . Single-cell transcriptome analysis of tumor and stromal compartments of pancreatic ductal adenocarcinoma primary tumors and metastatic lesions. Genome Med 2020;12:80. doi:10.1186/s13073-020-00776-9
    21. ↵
      2. Raghavan S ,
      3. Winter PS ,
      4. Navia AW , et al
      . Microenvironment drives cell state, plasticity, and drug response in pancreatic cancer. Cell 2021;184:6119–37. doi:10.1016/j.cell.2021.11.017
      OpenUrlCrossRef
    22. ↵
      2. Steele NG ,
      3. Carpenter ES ,
      4. Kemp SB , et al
      . Multimodal mapping of the tumor and peripheral blood immune landscape in human pancreatic cancer. Nat Cancer 2020;1:1097–112. doi:10.1038/s43018-020-00121-4
      OpenUrl
    23. ↵
      2. Balança C-C ,
      3. Salvioni A ,
      4. Scarlata C-M , et al
      . PD-1 blockade restores helper activity of tumor-infiltrating, exhausted PD-1hicd39+ CD4 T cells. JCI Insight 2021;6:e142513. doi:10.1172/jci.insight.142513
    24. ↵
      2. Thommen DS ,
      3. Koelzer VH ,
      4. Herzig P , et al
      . A transcriptionally and functionally distinct PD-1+ CD8+ T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade. Nat Med 2018;24:994–1004. doi:10.1038/s41591-018-0057-z
      OpenUrlCrossRefPubMed
    25. ↵
      2. Hanada K-I ,
      3. Zhao C ,
      4. Gil-Hoyos R , et al
      . A phenotypic signature that identifies neoantigen-reactive T cells in fresh human lung cancers. Cancer Cell 2022;40:479–93. doi:10.1016/j.ccell.2022.03.012
      OpenUrl
    26. ↵
      2. Sade-Feldman M ,
      3. Yizhak K ,
      4. Bjorgaard SL , et al
      . Defining T cell states associated with response to checkpoint immunotherapy in melanoma. Cell 2018;175:998–1013. doi:10.1016/j.cell.2018.10.038
      OpenUrlCrossRefPubMed
    27. ↵
      2. Wu TD ,
      3. Madireddi S ,
      4. de Almeida PE , et al
      . Peripheral T cell expansion predicts tumour infiltration and clinical response. Nature 2020;579:274–8. doi:10.1038/s41586-020-2056-8
      OpenUrlPubMed
    28. ↵
      2. Guo X ,
      3. Zhang Y ,
      4. Zheng L , et al
      . Publisher correction: global characterization of T cells in non-small-cell lung cancer by single-cell sequencing. Nat Med 2018;24:1628. doi:10.1038/s41591-018-0167-7
    29. ↵
      2. Zhang L ,
      3. Yu X ,
      4. Zheng L , et al
      . Lineage tracking reveals dynamic relationships of T cells in colorectal cancer. Nature 2018;564:268–72. doi:10.1038/s41586-018-0694-x
      OpenUrlPubMed
    30. ↵
      2. Yost KE ,
      3. Satpathy AT ,
      4. Wells DK , et al
      . Clonal replacement of tumor-specific T cells following PD-1 blockade. Nat Med 2019;25:1251–9. doi:10.1038/s41591-019-0522-3
      OpenUrlCrossRefPubMed
    31. ↵
      2. Zheng C ,
      3. Zheng L ,
      4. Yoo J-K , et al
      . Landscape of infiltrating T cells in liver cancer revealed by single-cell sequencing. Cell 2017;169:1342–56. doi:10.1016/j.cell.2017.05.035
      OpenUrlCrossRefPubMed
    32. ↵
      2. Kobayashi S ,
      3. Watanabe T ,
      4. Suzuki R , et al
      . TGF-Β induces the differentiation of human CXCL13-producing CD4 (+) T cells. Eur J Immunol 2016;46:360–71. doi:10.1002/eji.201546043
      OpenUrlCrossRefPubMed
    33. ↵
      2. Workel HH ,
      3. Lubbers JM ,
      4. Arnold R , et al
      . A transcriptionally distinct CXCL13+CD103+CD8+ T-cell population is associated with B-cell recruitment and neoantigen load in human cancer. Cancer Immunol Res 2019;7:784–96. doi:10.1158/2326-6066.CIR-18-0517
      OpenUrlAbstract/FREE Full Text
    34. ↵
      2. Yoshitomi H ,
      3. Kobayashi S ,
      4. Miyagawa-Hayashino A , et al
      . Human SOX4 facilitates the development of CXCL13-producing helper T cells in inflammatory environments. Nat Commun 2018;9:3762. doi:10.1038/s41467-018-06187-0
    35. ↵
      2. Dieu-Nosjean M-C ,
      3. Goc J ,
      4. Giraldo NA , et al
      . Tertiary lymphoid structures in cancer and beyond. Trends Immunol 2014;35:571–80. doi:10.1016/j.it.2014.09.006
      OpenUrlCrossRefPubMed
    36. ↵
      2. Efremova M ,
      3. Vento-Tormo M ,
      4. Teichmann SA , et al
      . CellPhoneDB: inferring cell-cell communication from combined expression of multi-subunit ligand-receptor complexes. Nat Protoc 2020;15:1484–506. doi:10.1038/s41596-020-0292-x
      OpenUrlCrossRefPubMed
    37. ↵
      2. Cosgrove J ,
      3. Novkovic M ,
      4. Albrecht S , et al
      . B cell zone reticular cell microenvironments shape CXCL13 gradient formation. Nat Commun 2020;11:3677. doi:10.1038/s41467-020-17135-2
    38. ↵
      2. Hiraoka N ,
      3. Ino Y ,
      4. Yamazaki-Itoh R , et al
      . Intratumoral tertiary lymphoid organ is a favourable prognosticator in patients with pancreatic cancer. Br J Cancer 2015;112:1782–90. doi:10.1038/bjc.2015.145
      OpenUrlCrossRefPubMed
    39. ↵
      2. Castino GF ,
      3. Cortese N ,
      4. Capretti G , et al
      . Spatial distribution of B cells predicts prognosis in human pancreatic adenocarcinoma. Oncoimmunology 2016;5:e1085147. doi:10.1080/2162402X.2015.1085147
    40. ↵
      2. Lynch KT ,
      3. Young SJ ,
      4. Meneveau MO , et al
      . Heterogeneity in tertiary lymphoid structure B-cells correlates with patient survival in metastatic melanoma. J Immunother Cancer 2021;9:e002273. doi:10.1136/jitc-2020-002273
    41. ↵
      2. Posch F ,
      3. Silina K ,
      4. Leibl S , et al
      . Maturation of tertiary lymphoid structures and recurrence of stage II and III colorectal cancer. Oncoimmunology 2018;7:e1378844. doi:10.1080/2162402X.2017.1378844
    42. ↵
      2. Siliņa K ,
      3. Soltermann A ,
      4. Attar FM , et al
      . Germinal centers determine the prognostic relevance of tertiary lymphoid structures and are impaired by corticosteroids in lung squamous cell carcinoma. Cancer Res 2018;78:1308–20. doi:10.1158/0008-5472.CAN-17-1987
      OpenUrlAbstract/FREE Full Text
    43. ↵
      2. Patil NS ,
      3. Nabet BY ,
      4. Müller S , et al
      . Intratumoral plasma cells predict outcomes to PD-L1 blockade in non-small cell lung cancer. Cancer Cell 2022;40:289–300. doi:10.1016/j.ccell.2022.02.002
      OpenUrlCrossRefPubMed
    44. ↵
      2. Stebegg M ,
      3. Kumar SD ,
      4. Silva-Cayetano A , et al
      . Regulation of the germinal center response. Front Immunol 2018;9:2469. doi:10.3389/fimmu.2018.02469
      OpenUrlCrossRefPubMed
    45. ↵
      2. Victora GD ,
      3. Dominguez-Sola D ,
      4. Holmes AB , et al
      . Identification of human germinal center light and dark zone cells and their relationship to human B-cell lymphomas. Blood 2012;120:2240–8. doi:10.1182/blood-2012-03-415380
      OpenUrlAbstract/FREE Full Text
    46. ↵
      2. Milpied P ,
      3. Cervera-Marzal I ,
      4. Mollichella M-L , et al
      . Human germinal center transcriptional programs are de-synchronized in B cell lymphoma. Nat Immunol 2018;19:1013–24. doi:10.1038/s41590-018-0181-4
      OpenUrlCrossRef
    47. ↵
      2. N J ,
      3. J T ,
      4. Sl N , et al
      . Tertiary lymphoid structures and B lymphocytes in cancer prognosis and response to immunotherapies. OncoImmunology 2021;10. doi:10.1080/2162402X.2021.1900508
    48. ↵
      2. Raphael BJ ,
      3. Hruban RH ,
      4. Aguirre AJ , et al
      . Integrated genomic characterization of pancreatic ductal adenocarcinoma. Cancer Cell 2017;32:185–203. doi:10.1016/j.ccell.2017.07.007
      OpenUrlCrossRefPubMed
    49. ↵
      2. Becht E ,
      3. Giraldo NA ,
      4. Lacroix L , et al
      . Estimating the population abundance of tissue-infiltrating immune and stromal cell populations using gene expression. Genome Biol 2016;17. doi:10.1186/s13059-016-1070-5
    50. ↵
      2. Song L ,
      3. Cohen D ,
      4. Ouyang Z , et al
      . TRUST4: immune repertoire reconstruction from bulk and single-cell RNA-seq data. Nat Methods 2021;18:627–30. doi:10.1038/s41592-021-01142-2
      OpenUrlCrossRefPubMed
    51. ↵
      2. Gunderson AJ ,
      3. Rajamanickam V ,
      4. Bui C , et al
      . Germinal center reactions in tertiary lymphoid structures associate with neoantigen burden, humoral immunity and long-term survivorship in pancreatic cancer. OncoImmunology 2021;10. doi:10.1080/2162402X.2021.1900635
    52. ↵
      2. Zhang N ,
      3. Bevan MJ
      . Transforming growth factor-β signaling controls the formation and maintenance of gut-resident memory T cells by regulating migration and retention. Immunity 2013;39:687–96. doi:10.1016/j.immuni.2013.08.019
      OpenUrlCrossRefPubMedWeb of Science
    53. ↵
      2. Ferreira C ,
      3. Barros L ,
      4. Baptista M , et al
      . Type 1 Treg cells promote the generation of CD8+ tissue-resident memory T cells. Nat Immunol 2020;21:766–76. doi:10.1038/s41590-020-0674-9
      OpenUrlPubMed
    54. ↵
      2. Rodriguez AB ,
      3. Peske JD ,
      4. Woods AN , et al
      . Immune mechanisms orchestrate tertiary lymphoid structures in tumors via cancer-associated fibroblasts. Cell Rep 2021;36:109422. doi:10.1016/j.celrep.2021.109422
      OpenUrlPubMed
    55. ↵
      2. Meylan M ,
      3. Petitprez F ,
      4. Becht E , et al
      . Tertiary lymphoid structures generate and propagate anti-tumor antibody-producing plasma cells in renal cell cancer. Immunity 2022;55:527–41. doi:10.1016/j.immuni.2022.02.001
      OpenUrl
    56. ↵
      2. Schumacher TN ,
      3. Thommen DS
      . Tertiary lymphoid structures in cancer. Science 2022;375:eabf9419. doi:10.1126/science.abf9419

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

    Footnotes

    • Twitter @Pedrohbarbosa01, @ClaudioVladmir

    • Contributors GSK and TSM conceived and designed the study. GSK, GAFV, MPCP, PHBP, MLRC, WASF, ASC, AR and AFG processed patient samples and performed in vitro experiments. ABD performed immunofluorescence staining. ACL and LMRBA performed Nanostring nCounter experiments. GSK analyzed scRNA-seq, flow cytometry, Nanostring and bulk RNA-seq data. AD assembled TCRs/BCRs and JLF modeled tumour progression from TCGA bulk RNA-seq data. TMB, RTN, AMP, AM and FJFC collected surgical specimens. TMB and VHFJ collected clinical data. GOS, WAN, LCLC, VHFJ, FJFC and VCCL provided clinicopathological guidance to the study. GOS, WAN and LCLC performed histological analyses. RW provided resources and experimental support. GSK, GAFV, VCCL and TSM interpreted the results. GSK, GAFV and TSM wrote the manuscript with input from all other authors. TSM is the guarantor of this study.

    • Funding The work was funded by the São Paulo Research Foundation (FAPESP, grant 18/14034-8 to TSM), the National Council for Scientific and Technological Development (CNPq, grant 465682/2014-6 to TSM) and the National Institute of Science and Technology in Oncogenomics and Therapeutic Innovation (INCITO, grant 14/50943-1 to TSM). GSK, GAFV, MLRC and LMRA were supported by fellowships from FAPESP (19/25129-2, 20/10299-7, 21/00643-5 and 21/04100-6, respectively). PHBP, ASC and AR were supported by fellowships from the Coordination for the Improvement of Higher Education Personnel (CAPES).

    • Competing interests VCCL: honoraria (educational presentations and participation in scientific events) from Astra-Zeneca, MSD, BMS, Roche, Amgen, GSK, Lilly; advisory board from Astra-Zeneca, MSD, BMS, Pfizer, Janssen, Amgen.

    • Provenance and peer review Not commissioned; externally peer reviewed.

    • Supplemental material This content has been supplied by the author(s). It has not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been peer-reviewed. Any opinions or recommendations discussed are solely those of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and responsibility arising from any reliance placed on the content. Where the content includes any translated material, BMJ does not warrant the accuracy and reliability of the translations (including but not limited to local regulations, clinical guidelines, terminology, drug names and drug dosages), and is not responsible for any error and/or omissions arising from translation and adaptation or otherwise.