Fig 2: Coinjection of CXCL13 and CCL21 into orthotopic tumors, and administration of intraperitoneal gemcitabine. (A–D) Representative immunofluorescence images of aggregates or TLSs as detected by the presence of B220 (green), and CD21 (red) with DAPI (blue) staining nuclei in (A) PBS-treated, (B) chemokine-treated, (C) gemcitabine-treated, (D) and C+G-treated mice. (E, F) IF staining of gemcitabine- and C+G-treated mice for granzyme B (green) and CD3 (red) T cells. Insets represent a zoomed-in view of the aggregate. (G) Quantification of the granzyme B+ T cells in gemcitabine alone (G) and in combination with C+G-treated mice. Two-samples Kolmogorov-Smirnov test: ∗∗P < .01. Scale bar, (A–D) 50 μm, (E, F) 250μm .

Fig 3: Immunologically active TLSs are present in a fraction of chemo-naïve human PDAC. (A) Colocalization of T cells (CD3+), B cells (CD20+), and FDCs (CD21+) with high endothelial venules (peripheral node addressin positive [PNAd+]) in dense, compact lymphoid aggregates on consecutive sections defining human TLSs. (B) Human PDAC section stained for CD20 (magenta), CD21 (green), PNAd (red), and DAPI (blue) (upper panel). Sequential section with in situ hybridization of CXCL13 (red) RNA-scope probe in human PDAC patient (lower panel). (C) Frequency of TLSs in human PDAC (n = 56 TMAs, n = 31 full sections). (D) TLS density (expressed as number of CD3+CD20+CD21+ clusters/mm2) in a cohort of human PDAC (n = 17). The dotted line represents cutoff for identification of TLS+ (empty circles) and TLS– (bold circles) PDAC patients. (E) Plot showing the significant correlation of CD20+ B cell density with TLS density. The dotted line represents the cutoff of minimal B cell density needed to induce ectopic lymphoneogenesis. Empty circles represent TLS+ patients, bold circles represent TLS– patients. Spearman r = 0.84, P = .0001. (F) Distribution of TLS stages in patients with different TLS densities (TLS maturation). (G) Human PDAC full section stained for CD3+ T cells, CD20+ B cells, and FDCs, using modified immunohistochemistry stripping and reprobing protocol. Fourth (bottom) panel represents pseudo-color immunofluorescence image of the same sections, showing a cell phenotype map (immune, stromal, and tumor cells) using different colors to better depict spatial distribution. The boxes identify subsequent adjacent panels (I–V). Scale bar: 1000 μm. (I) Higher magnification of a representative area of presence of scattered T cells, with absence of B cells and FDCs. (II) Higher magnification of a representative area of cluster of T cells, with absence of B cells and FDCs. (III) Higher magnification of a representative area of sparse conglomerates of T cells and sparse B cells. FDCs are absent. (IV) Higher magnification of a representative area where T and B cells are clustered but not organized in distinct zones. FDCs are absent (early TLSs).(V) Higher magnification of TLSs, where a cloud of T cells surrounds a core of B cells that includes a FDC network. Those TLSs can show absence or presence of a germinal center; therefore, they are celled primary follicle-like TLSs and secondary follicle-like TLSs, respectively. Each data point represents an individual patient and lines represent median and interquartile range. Empty circles represent TLS+ tumors, bold circles represent TLS– tumors.

Fig 4: Coinjection of CXCL13 and CCL21 into orthotopic tumors. (A) Correlation of CD19+ B cell density with TLS density in orthotopic pancreatic tumors. The dotted line represents the cutoff of minimal B cell density needed to induce lymphoneogenesis. Pink circles represent lymphoid aggregate (LA)– PBS-treated mice, red circles represent LA+ PBS-treated mice, blue circles represent TLS+ CXCL13/CCL21-treated mice, empty blue circles represent stress-induced LA+ CXCL13/CCL21-treated mice. Spearman r = 0.79, P < .0001. (B–G) Flow cytometric analysis, in LA+/– tumors from PBS-treated mice (n = 12) and TLS+/– tumors from CXCL13/CCL21-treated mice (n = 16), of (B) CD19+ B cells out of CD45+ cells, (C) CD3+ T cells out of CD45+ cells, (D) CD4+ helper T cells out of CD3+ T cells, (E) CD8+ cytotoxic T cells out of CD3+ T cells, (F) FOXP3+ regulatory T cells out of CD4+CD3+ T cells, and (G) CD11b+ myeloid cells out of CD45+ cells. The mean percentage of CD19+ cells out of CD45+ cells within the PBS-treated group used to identify stress-induced lymphoid aggregates from potential chemokine-induced TLSs (dotted line in A). (H) Tumor weight in CXCL13/CCL21 intratumoral single or coinjected compared with PBS-treated mice. Each data point represents 1 mouse. Empty circles in CXCL13/CCL21 treated mice represent stress-induced lymphoid aggregates. Two-sample Kolmogorov-Smirnov test. ∗P < .05. ns, not significant.

Fig 5: Artificial induction of TLSs in orthotopic model of PDAC. (A) Ultrasound images of the pancreatic tumor of orthotopic mice preinjection (left panel), during injection (middle panel), and postinjection (right panel) and schema of intratumoral injection of lymphoid chemokines CXCL13 and CCL21 after development of orthotopic PDAC tumor. Red arrow: needle trajectory; red dotted line: tumor; red asterisk: spread of solution. Schema of experiment. (B) Proportion of CD3+, CD19+ CD11b+ and other immune cells out of total CD45+ cells after intratumor lymphoid chemokine injection as assessed by flow cytometric analysis (PBS n = 12, CXCL13/CCL21 n = 16, CXCL13 n = 4, CCL21 n = 4). (C) Representative immunofluorescence staining on sequential sections of lymphoid aggregates as detected by the presence of B cells (B220), the presence of T cells (CD3) (upper panels), and the near absence of FDCs (CD21) (lower panels) with DAPI staining nuclei in PBS (vehicle control) injected mice. (D) Representative immunofluorescence images of TLSs as detected by the presence of B cells (B220), T cells (CD3) (upper panels), and well-formed network of FDCs (CD21) (lower panels) with DAPI staining nuclei in dual chemokine-injected mice, showing mature TLSs. (E) Lymphoid aggregate and TLSs induction in CXCL13/CCL21 intratumoral coinjected mice. (PBS n = 12, CXCL13/CCL21 n = 16). (B) Kruskal-Wallis and (E) chi-square test. ∗P < .05, ∗∗∗∗P < .0001. Scale bar: 50 μm. n.s, not significant.