Animal procedures were carried out in accordance with the U.K. Home Office Animal and Scientific Procedures Act 1986 and the European Directive 2010/63/EU, reviewed and approved by the UK Home Office (project licenses 70/7411, 70/7449 and PBE3719B3). Male and female μMT mice (C57BL/6 background) were generated as described (35). Male and female KPC mice (23) were generated in house by crossing Lox-STOP-Lox (LSL) Kras^G12D/+ and LSL-Trp53^R172H/+ (C57BL/6/129/SVJae) with Pdx-1-Cre (C57BL/6) mice. Control mice were aged-matched healthy Pdx-1-Cre. For orthotopic experiments using anti-CD20, 10–12-week-old female C57BL/6 mice (Charles River, U.K.) were used.

The KPC-derived PDAC cell line (TB32048) previously generated from a female C57BL/6 KPC mouse was provided as a generous gift from David Tuveson laboratory. TB32048 was cultured for 3–4 passages at maximum 80% confluency in 10% fetal calf-serum (#A15-104, GE Healthcare) in DMEM (#E15-810, PAA) + 100 μg/ml penicillin/streptomycin (#P11-G10, PAA) in a T175 flask in standard conditions (37°C, 5% CO₂) and tested regularly in house for mycoplasma (#rep-pt1, Invivogen and #LT07-710, Lonza). Cells were dissociated using 0.1% trypsin (PBS) (#594-18C, Sigma) for 10 min at 37°C and resuspended in PBS and BD Matrigel™ Basement Membrane Matrix High Concentration (#354248, BD Biosciences) in a ratio of 1:1. 1,000 cells in 5 μl was injected into the pancreas using a Hamilton® syringe, 700 series (#10100332, Fisher Scientific). For the experiments used to assess CD138^hi cells and anti-CD20 at day 10, 10,000 cells in 30 μl Matrigel, or Matrigel only sham surgery was performed. The peritoneal wall was sutured using 6/0-gauge coated vicryl sutures™ (#W9500T, Ethicon) and skin closed using two 9 mm Clay Adams Clips (#IN015A, VetTech Solutions) and an Autoclip® applier (#IN015B, VetTech Solutions).

Mice with orthotopic tumors were treated with either a B-cell-depleting anti-CD20 (mouse IgG2a, clone 18B12) or IgG2a isotype control (clone WR17), produced in house (36). Treatments were administered on day−8,−2, 14, with orthotopic surgery performed on day 0. Mice were culled on days 26–27. KPC mice were treated with either 250 μg of anti-CD20 antibody (mouse IgG2a) to deplete B cells or with IgG2a isotype control (anti-ragweed 1428). Both antibodies were provided as a gift from Genentech, however, for the last batch of experiments, the Genentech isotype control was not sufficient and a mouse IgG2a from BioXcell was used instead (clone C1.18.4, lot 5518/1214) (#BE0085, BioXcell). KPC mice were palpated at least once per week for a pancreatic tumor from approximately 70 days old. When a tumor was detected, treatment with either PBS, anti-CD20 or mouse IgG2a commenced between 1 and 7 days later. Treatments were given in 200 μl sterile PBS once per week for 5 (n = 20) or 6 weeks (n = 8) and mice were culled a week later, at which point most had developed small tumors (n = 20). The Genentech antibodies were also used for the orthotopic experiment, where anti-CD20 or IgG was administered at day 10 and mice culled at day 23 (endpoint).

Organs were passed through a 70-μm-pore size cell strainer (#11597522, Fisher). Red blood cell lysis was performed for blood and spleen samples using RBC Lyse buffer (#555899, BD Biosciences) for 10 min at RT. Tumors were digested in 5 ml 2.0 mg/ml collagenase (#C9263, Sigma) (DMEM), 0.025 mg/ml DNase (#D4513, Sigma) at 37°C under agitation for 20 min and passed through a 70 μm cell strainer to achieve a single cell suspension. 0.5–2.0 × 10⁶ cells were incubated with anti-CD16/32 FcR Block (1:200) (#553142, BD Biosciences) for 15 min followed by 50 μl containing labeled antibodies (Supplementary Table S1) for 30 min, all at 4°C. Cells were washed in FACS buffer (5% BSA, 2mM EDTA, PBS) and incubated with fixable viability dye (FVD506) at 1:200 in PBS (#65-0866, eBioscience) at 4°C for 20 min. After further washing in FACS buffer, they were incubated for 20 min in 100 μl of 2% paraformaldehyde/FACS buffer (#BX1143CB0201, Adams) at 4°C. Cells were washed and analyzed on BD LSR Fortessa™. For qPCR B cells were sorted from spleens and tumors as DAPI⁻ CD45⁺ CD19⁺ (with the exclusion of CD138⁺ in the spleens) using fluorescence-activated cell sorting. Cells were stained as above but approximately 100 μl of FcR block and 100 μl 2X antibody master mix was used per 10 × 10⁶ cells. The viability dye DAPI (#D9542, Sigma) was added at 2 μg/ml immediately before sample acquisition on the BD FACS Aria II and samples were collected in 1 ml sterile FBS before RNA extraction.

All immunofluorescence (IF) was carried out at RT. 4–7 μm sections were prepared from frozen samples, which were thawed then fixed with 4% paraformaldehyde solution (#BX1143CB0201, Adams) for 20 min at RT. Sections were washed 3x in 0.1% PBS-Tween 20 (PBS-T) and permeabilized in 0.1% TritonX100 solution for 5 min. Then they were washed and blocked using a blocking buffer of 5% goat serum, 2.5% BSA in PBS for 1 hr. B220, E-Cadherin, EpCAM, IgG1, IgG2a/b, IgG3, and IgM were added in blocking buffer for 1 hr (details in Supplementary Table S2 and isotypes in Supplementary Table S3) then washed 3x in PBS-T, 1x PBS, 1x H₂O for 5 min. Sections were stained with DAPI (1:10,000 5 min in H₂0) and mounted with Prolong Gold Antifade with DAPI (# P36931, Invitrogen) and stored at 4°C. Images were acquired on LASER scanning microscopes 510 and 710 (Zeiss) and processed using Image J. The colocalization coefficient was generated on the raw LSM files on Zen 2009 software.

Immunohistochemistry (IHC) was carried out at RT. Frozen sections were thawed and fixed for 20 min in 4% paraformaldehyde solution, then permeabilized in 0.1% TritonX100 solution for 5 min and washed once in 0.1% PBS-Tween (PBS-T). One hundred μl of Dako dual endogenous enzyme block (#S2003, Dako) was added for 15 min at RT and washed 1x in PBS-T, followed by a blocking step in 100 μl 5% goat serum, 2% BSA in PBS (blocking buffer) for 1 hr. The primary antibody (B220, CK8) in blocking buffer was added for 1 hr (details in Supplementary Table S2 and isotypes in Supplementary Table S3). Slides were washed 3x in PBS-T, then secondary anti-rat or anti-rabbit antibodies (1:200) conjugated to biotin in 100 μl blocking buffer were incubated for 1 hr (Supplementary Table S4). Slides were washed 3x in PBS-T and 1x in PBS. Endogenous peroxidase activity was blocked by a 20 min incubation in 250 ml 99.9% Methanol (#M/4000/PB17, Fisher Scientific) and 5 ml 30% H₂O₂ (#H/1800/15, Fisher Scientific). Slides were then washed 3x in PBS-T. The VECTASTAIN Elite ABC kit (standard) (#PK-6100, Vector Labs), was prepared as per manufacturer's instructions. Briefly 2 drops of solution “A” and 2 drops of solution “B” were added to 5 ml of PBS, vortexed and stored on ice prior to using, then incubated for 30 min and washed 2x in PBS-T and 1x in PBS. SIGMAFAST™ 3,3'-Diaminobenzidine (DAB) tablets (#D4293, Sigma) were prepared as per manufacturer's instructions. Briefly one DAB tablet and one Tris/peroxide tablet was added to 5 ml doubled distilled H₂0 and filtered (0.45 μm) before use. DAB was applied to sections for approximately 4–10 min). Slides were washed in double distilled H₂O. Hematoxylin (#1.05175.2500, VWR) was applied for 1 min then washed until clear in distilled water. Tap water was used to blue tissue for 30 s−1 min. Slides were then dehydrated by 3 min in 95% ethanol, 2x 3 min in 100% ethanol and 5 min in xylene. Coverslips were mounted using DPX solution (#06522, Sigma). B220 and CK8 were quantified using Definiens “Tissue Studio” 64 software.

Sorted B cells were vortexed for 1 min in 350 μl of RLT buffer per 3 × 10⁶ cells and stored at −80°C. RNA extraction was performed using Qiagen RNeasy micro kit (#74004, Qiagen) as per manufacturer's instructions. RNA concentration and integrity were measured using the Nanodrop and the integrity and concentration of the RNA was in some cases further confirmed using an Agilent RNA 6000 Pico Kit (#5067-1513, Agilent), according to manufacturer's instructions. For experiments using the Qiagen RT2 profiler, cDNA synthesis was performed using the RT2 First Strand Kit (#330401, Qiagen) as per manufacturer's instructions and qPCR performed using the RT² Profiler™ PCR Array Mouse Cancer Inflammation & Immunity Crosstalk (#PAMM-181Z-E-4, Qiagen) performed on the C1000 Touch Thermal Cycler (#CFX384, Bio- Rad), using cycling conditions recommended by the manufacturer. All additional qPCR was performed using reagents from Applied Biosystems. For additional qPCR, RNA was extracted as before. cDNA was generated using the High Capacity reverse transcription kit (#4368814, Applied Biosystems) as per manufacturer instructions. qPCR reaction was performed using a master mix of 9 μl cDNA, 10 μl iTaq (#1725134, Bio-Rad) and 1 μl primer (Applied Biosystems). The primers used were Cxcl2 Mm00436450_m1, Ebi3 Mm00469294_m1, Gapdh Mm99999915_g1, Il10 Mm00439614_m1, Il12a Mm00434169_m1, Il12b Mm00478374_m1, and Ptgs2 Mm00478374_m1. The reaction was performed on a Step One Plus Thermal Cycler (Applied Biosystems) using the following cycling conditions: 2 min 50°C x1, 10 min 95°C x1, 15 s 95°C x40 cycles, 1 min 60°C. The raw Ct values were exported for subsequent analysis. The ΔCt was calculated by subtracting the Ct of the housekeeping genes Gapdh from each Ct value i.e., ΔCt = Ct Gene – Ct Gapdh and then expressed as 2^∧(−ΔCt). For data generated using the RT2 Profiler array, a heatmap was generated in the statistical programming language R (version 3.1.3), using the gplots package. It was constructed using row z-scores of log2 transformed gene expression values. Color-coding corresponds to row z-scores, columns represent individual samples and a row dendrogram was included to illustrate clustering relationships of gene expression.

Plasma and pancreas or pancreatic tumor were harvested from healthy Pdx-1-Cre mice and KPC mice at endpoint. Protein lysates of pancreas and tumor tissue were prepared by cutting and weighing the tissue on dry ice, then lysing in ice-cold lysis buffer: 150 mM NaCl, 20 mM Tris pH7.5, 1 mM EDTA, 1 mM EGTA, 1% Triton X-100 with 2 protease inhibitor tablets (#1836170, Roche), 200 μl phosphatase inhibitor II (#P-5726, Sigma) and 200 μl phosphatase inhibitor III (#P-0044, Sigma) per 10 ml of lysis buffer. Lysis buffer was added 1 ml per 75 mg tissue and digested in M-tubes (#130-093-236, Miltenyi) in a gentleMACS™ Dissociator on protein_01 setting (#130-093-235, Miltenyi). Samples were centrifuged at 244 rcf for 2 min and stored on ice, then sonicated in 10 s bursts at 40% amplitude whilst on ice. Samples were then rotated at 4°C for 30 min and centrifuged at 16.1 rcf for 15 min at 4°C. The final supernatant was aliquoted and stored at −80°C and the pellet discarded. The concentration of immunoglobulins IgA, IgG1, IgG2a, IgG2b, IgG3, and IgM was detected using the Mouse Isotyping Panel 1 kit (#K15183B-1, Mesoscale Discovery (MSD), as per manufacturer's instructions and analyzed on an MSD Model S1250 2400. The concentration of IgE was determined using the Mouse IgE ELISA MAX Standard (#432401, Biolegend). For the MSD plate, plasma samples were diluted at 1:5,000 and protein lysates were diluted 1:200 in 1% BSA/PBS. For the IgE ELISA, the plasma and protein lysates were diluted at 1:100 in 1% BSA/PBS.

The concentration of C1q-Ig immune complexes was analyzed using the Mouse CIC Ig's (total A+G+M) ELISA kit (#5900, Alpha Diagnostic International). Serum was prepared by taking blood from healthy and KPC mice at endpoint. Blood was allowed to clot for 1 hr at RT followed by centrifugation at 16.1 rcf for 10 min, serum was removed and stored at −80°C. Protein lysates of pancreas tissue were made as described above (see immunoglobulin concentration in plasma and pancreas). Samples were thawed on ice and diluted 1:100 in H₂O, 100 μl were used for the ELISA assay (in duplicate), which was performed as per manufacturer's instructions. Optical density (OD) was measured at 450 nm.

In order to detect intracellular cytokines IFNɤ and TNFα and proliferation (Ki67) in CD4⁺ and CD8⁺ T cells (antibody details in Supplementary Table S1), 2 × 10⁶ cells from digested orthotopic tumors were cultured in a 96-well U-bottomed plate for 5 hr with Cell Stimulation Cocktail (#00-4970-03, eBioscience), with a final concentration of 0.081 μM phorbol 12-myristate 13-acetate (PMA) and 1.34 μM ionomycin (#00-4980-03, eBioscience). After 1 hr incubation, cell transport inhibitor cocktail was added, at a final concentration of 1.06 μM Brefeldin and 2.0 μM Monensin (#00-4980-03, eBioscience) for the remaining 4 hr incubation. After 5 hr, the cells were transferred to a v-bottomed plate on ice and stained for flow cytometry. To detect cytoplasmic IL-10 in B cells, cells from digested pancreatic tumors were plated in 10 μg/ml LPS (#L2630, Sigma), PMA and Brefeldin for 5–12 hr, at which point they were stained for flow cytometry.

Cells were prepared for flow cytometry as previously described above. The intracellular fixation and permeabilization buffer kit (#88-8824, eBioscience) was used for intracellular (cytoplasmic) staining. After extracellular staining and viability dye stain had been performed, 100 μl of intracellular (IC) fixation buffer were added for 1 hr at RT. Following this, the IC buffer was washed twice with permeabilization buffer. The intracellular antibody was diluted in permeabilization buffer for 20–60 min. Cell pellets were resuspended in 100 μl FACS buffer and stored at 4°C until acquisition on the flow cytometer, performed on the same or the following day.

Cells were prepared for flow cytometry as previously described. For intranuclear staining of markers such as FOXP3 and Ki67 (antibody details in Supplementary Table S1), the Foxp3/transcription factor fixation/permeabilization concentrate and diluent kit (#00-5523, eBioscience) were used. After extracellular staining and viability dye stain had been performed, 100 μl of fixation/permeabilization buffer was added for 1 hr at RT, then washed twice with 100 μl of permeabilization buffer. The intranuclear antibody mix was diluted in permeabilization buffer and incubated for 60 min. The cells were then washed with 100 μl permeabilization buffer and with 100 μl FACS buffer. Cell pellets were resuspended in 100 μl FACS buffer and stored at 4°C until acquisition on the flow cytometer, performed on the same or the following day.

The percentage of PDAC in the tissue of KPC tumors post anti-CD20 treatment was analyzed blind by a Consultant Histopathologist, assessing whole sections (Dr. Jacqueline R. McDermott, Department of Pathology, University College London Hospital). Similarly, PanIN lesions were quantified by analyzing 10 high-powered fields at 20X magnification, blind.

Immunofluorescence images were analyzed using Image J (Java) and bright field images were analyzed using Panoramic Viewer (3DHISTECH) software and Zen 2 (Carl Zeiss). Graphic representation of data and statistical analysis was performed using GraphPad Prism Version 5. Data was tested for normality using the Kolmogorov-Smirnov test. If the data were normally distributed, then an unpaired t-test or One-Way ANOVA was used with Bonferroni's post-test. Non-parametric data were tested using a Mann-Whitney test or Kruskal-Wallis and Dunn's post-test. A cut off p < 0.05 was used to define significance.

Additional tables detailing the antibody clones and respective product codes are available in Supplementary Material.

In order to dissect the discrepancy regarding the role of B cells in human and murine models of PDAC, we started by assessing the total B cell infiltration in KPC and orthotopic tumors (Figures 1A,B and Supplementary Figure S1A). We observed a substantial infiltration of B cells into KPC tumors, which was significantly higher (6-fold) than that seen in orthotopic tumors at endpoint, both in density (Figure 1A) and proportion of immune infiltrate (Figure 1B). The B cell proportion observed in KPC tumors was comparable to that reported in PDAC patients (27). B cells largely organized in clusters in the stroma of KPC tumors (Supplementary Figure S1B), whereas orthotopic tumors were poorly infiltrated (Supplementary Figure S1C). Furthermore, B cell infiltration significantly correlated with T cell infiltration in KPC tumors (Figure 1C) but not in orthotopic (Supplementary Figure S1D), which showed lower infiltration of both B and T cells. We wondered if the substantial infiltration in KPC tumors was a result of antigen encounter. Following antigen encounter, B cells become activated and can enter germinal centers, leading to B cell proliferation and differentiation to either memory cells or plasma cells producing high-affinity immunoglobulin clones (37, 38). In line with this, we found a significant increase in the GL7^hi B cell frequency, a marker of activated B cells (Figure 1D), in both the tumor (Figure 1E) and the tumor-draining mesenteric lymph nodes of KPC mice (Figure 1F) when compared to healthy murine pancreas or lymph nodes, suggesting an active B cell response within the tumor in this model. However, we did not see a significant increase in GL7^hi B cells in the spleen (Figure 1F). Most GL7^hi B cells were positive for CD95 (Figure 1G and Supplementary Figure S1E), indicating they were part of a germinal center. This was further confirmed in additional samples with the alternative gating strategy of CD95⁺ CD38⁻ staining in KPC tumors (Supplementary Figure S1F). Other subsets such as B1 cells (CD11b⁺) (mean 7.6% ± SEM 0.9) and IgG1⁺(9.2% ± 3.3), IgG2a/b⁺(4.1% ± 1.1) and IgA⁺ (21.5% ±18.8) memory B cells were also present, whereas IgG3⁺(1.0% ± 0.3) and IgE⁺ (1.1% ± 0.3) memory B cells and IL-10⁺ (0.7% ± 0.4) regulatory B cells were present in very low numbers in KPC tumors (Figure 1H and Supplementary Figures 1G–M). Plasma cells were only present in KPC tumors at a very low density, not significantly different from healthy pancreas, which normally lacks plasma cells (Supplementary Figure S1N), whereas they were significantly increased in the spleen (Figures 1I,J). However, plasmablasts were not significantly upregulated in the spleen (Supplementary Figure S1O). In contrast, in a small cohort of orthotopic mice, no expansion of splenic plasma cells was observed, when compared to aged-matched healthy and sham surgery controls (Supplementary Figure S1P).

Next, we wanted to assess if the plasma cells were producing immunoglobulins (Igs) and if these were reaching the tumor, therefore we analyzed all subclasses in the blood and tumor lysates of KPC mice. We found a slight increase of IgGs in KPC mice blood, which reached statistical significance only for IgG2b (Figure 1K). Crucially, we found a significant increase in all but one subclass (IgG2b) in KPC tumors compared to healthy pancreas (Figure 1L). The range of immunoglobulin subclasses produced indicated a broad-range of cytokine and antigen exposure. Interestingly, IgG1 is the most dominant subclass found in the blood (Figure 1K) and was similarly most strongly increased in the tumor; suggesting a Th2-cytokine dominance may be present during B-cell class switching, which has been described for PDAC (39). Additionally, we found a significant increase in tumor IgG2a (Figure 1L), regarded as the most pro-inflammatory subclass, providing the most effective antiviral or pathogenic response (40, 41). Although at a lower concentration than the other subclasses, IgE was also significantly increased within the tumor (Figure 1L), confirming data from a small cohort of human PDAC patients, where patient-derived IgE was found upregulated and exerted antitumoral activity via ADCC (33). Immunoglobulins target their bound antigen and this can lead to their degradation/elimination. One of the possible mechanisms by which they do this is through the formation of immune complexes (IC). Classically, during an infection, IC lead to activation of the complement cascade and the lysis of a target cell, or they target the antigen for phagocytosis (42). In cancer, however, IC have been shown to be tumor-promoting, by stimulating macrophage IL-10 secretion and potentiating an immunosuppressive microenvironment (43, 44). In order to assess if IgG were bound in immune complexes, thus contributing to an immunosuppressive environment in KPC mice, we analyzed IC content in the serum and tumors with a C1q-IgG ELISA, as previously reported (45). However, we found no evidence of increased IC formation at either site (Figure 1M and Supplementary Figure S1Q).

Given the increase in IgGs in KPC pancreatic tumors, we wanted to both confirm this finding and compare this to orthotopic tumors. We found pronounced immunoglobulin deposition in KPC tumors (Figures 2A,B and Supplementary Figures S2A,B), in line with human data (27). In contrast, immunoglobulin deposition was markedly weaker to absent in orthotopic tumors (Figures 2A,B), suggesting a much-reduced B cell activation in this model that has not progressed to plasma cell differentiation. IgG2a is the only immunoglobulin subclass which can bind the FcɤRI as a monomer and not in an immune complex. Interestingly, we found macrophages upregulated FcɤRI at the tumor site (Figure 2C) and it colocalized with IgG2a/b deposition (Figures 2D,E), suggesting IgG2a binds as a monomer to FcɤRI on tumor-infiltrating macrophages in KPC tumors.

Previous analysis of PDAC patient serum revealed tumor-derived immunoglobulins recognize predominantly intracellular/nuclear antigens (32). In order to assess if this was also the case for KPC-derived Ig, we incubated sections of healthy pancreas with the plasma of either healthy or KPC mice. We found that only KPC-derived IgG2a/b bound strongly to nuclear components of healthy pancreatic cells (Figures 2F,G) and crucially not to cells in other organs (Supplementary Figure S2C). We repeated this assay on glycine-treated KPC tumors, in order to remove native immunoglobulins, and again found nuclear binding of KPC-derived IgG2a/b (Figures 2H,I). This suggested to us that a B cell response is mounted against pancreatic intracellular antigens during KPC PDAC, accurately recapitulating observations in human PDAC (32, 33).

Taken together, our data indicate B cells heavily infiltrate KPC tumors, become activated, enter the germinal center reaction and produce substantial immunoglobulin deposits at the tumor site. Crucially, germinal center B cells are observed mainly within or near the tumor in the draining lymph nodes but not in the spleen, suggesting a local B cell activation to tumor antigens. More importantly, the orthotopic model appear different, possibly because the faster tumor kinetics does not allow sufficient time to see a full B cell involvement in these tumors.

The protumoral activity of B cells in other murine models has been partly attributed to their role as Bregs secreting cytokines IL-35 (46), IL-10 (47), and TGF-β (48), all described to induce immunosuppression. We found IL-10⁺ Bregs to represent only a small proportion of B cells infiltrating tumors (Figure 1H).

Therefore, to further investigate the phenotype of the large infiltration of B cells seen in KPC tumors and how they might influence the tumor microenvironment, we profiled the expression of 86 genes relating to cancer inflammation and immune cross-talk from tumor-infiltrating B cells, comparing them to splenic B cells both from KPC mice and healthy controls. Genes found to be significantly different are reported in the heatmap in Figure 3A.

Interestingly, splenic B cells showed a rather similar phenotype, whether derived from a KPC or healthy mouse, indicating the overall B cell population in the spleen largely retains its phenotype even in the presence of a pancreatic tumor, whereas B cells infiltrating the tumor develop a distinctly different profile. For the intra-tumoral B cells, there were numerous interesting features such as significantly upregulated proinflammatory cytokines (Spp1, Il6, Csf2, Vegfa, Ccl4), the inflammatory enzyme COX2 (Ptgs2), as well as chemokines involved in the recruitment of T cells in a germinal center (Cxcl1, Cxcl2, and Cxcl5) and macrophages/DCs (Ccl2, Cxcl12, Ccl20) (Figure 3A). On the other hand, genes associated with immunosuppression were downregulated, such as PDL1 (Cd274), as well as the IL-35 component Il12a (Figure 3A). The gene Il10 was not upregulated in infiltrating B cells but rather expressed at low levels across all three groups (Supplementary Figure S3A).

Although the B cell infiltration is far weaker, we wanted to understand if orthotopic intratumoral B cells developed a different profile from KPC intratumoral B cells. Thus, a number of genes were selected for further qRT-PCR and compared against original and independent KPC samples with the addition of B cells from orthotopic tumors. Upregulation of Ptgs2 and Cxcl2 was confirmed in additional KPC tumor-derived B cells and was also observed in orthotopic B cells (Figures 3B,C). Although significantly downregulated in the array, we found a weak but not significant downregulation of Il12a in KPC and in orthotopic samples (Figure 3D). Additionally, Ebi3, which together with Il12a forms the immunosuppressive IL-35 (28), showed no upregulation in tumor B cells (Figure 3E). Taken together, these results indicate that KPC tumor B cells do not express IL-35. Furthermore, the immunosuppressive cytokine Il10 was not upregulated in the array (Supplementary Figure S3A) or in additional KPC tumor B cells by qRT-PCR (Figure 3F), it was actually downregulated in orthotopic B cells (Figure 3F), further confirming the overall B-cell population within tumors holds a more proinflammatory and likely immunostimulatory phenotype. On the other hand, Il12b showed no difference in KPC samples analyzed by qRT-PCR and was weakly downregulated in orthotopic-derived B cells (Figure 3G), in contrast to the array which showed a certain level of upregulation (Figure 3A). It is not surprising that some results from an array are not subsequently validated (49, 50). This is why we further assessed important genes by qPCR and the results confirmed our initial findings.

Our data indicate that the PDAC tumor microenvironment uniquely shapes the phenotype of infiltrating B cells, demonstrated in KPC and similarly in orthotopic models. In both cases, infiltrating B cells possess a remarkably different phenotype to splenic B cells in tumor-bearing mice. This suggests a unique proinflammatory phenotype is acquired by B cells upon entry into the tumor microenvironment and the main difference between KPC and orthotopic tumors is likely to be the kinetics of this B-cell infiltration. Importantly, these results indicate that spleen-derived and tumor-derived B cells should not be used interchangeably in future studies.

Although in this study we have demonstrated a relative paucity of B cells in orthotopic tumors, it has been suggested that B cells may promote tumor growth in orthotopic models of PDAC (27, 28). To address this, we investigated the effect of B-cell loss in syngeneic orthotopic tumors through two methods; B-cell deficiency in μMT mice vs. transient B-cell depletion with anti-CD20 mAb. Injection of tumor cells in B-cell deficient, μMT mice (35) (Figure 4A) led to reduced tumor growth compared to WT littermates (Figure 4B), in agreement with others using similar PDAC cell lines and B-cell deficient models (27, 28). Owing to the loss of the B cell fraction (Figure 4C) and immunoglobulins (Supplementary Figure S4A), we then analyzed the tumor immune infiltrate and found comparable infiltration of MDSCs (CD45⁺, CD11b⁺, Ly6G⁺, Ly6C⁺), tumor associated macrophages (TAMs: CD45⁺, CD11b⁺, Ly6G⁻, Ly6C⁻, F4/80⁺, MHC II⁺), monocytes (CD45⁺, CD11b⁺, Ly6G⁻, Ly6C^hi), and dendritic cell subsets (CD45⁺, CD11b⁺, Ly6G⁻, Ly6C⁻, F4/80⁻, MHC II⁺, CD11c⁺) within the tumor between μMT mice and WT littermates (Figure 4D). CD86 expression on TAMs and DCs, as well as TAM mannose receptor (MR) expression was also unchanged, indicating B-cell loss does not influence their polarization to a more immunomodulatory phenotype (Supplementary Figure S4B). Furthermore, the ex vivo cytokine production, degranulation activity and proportion of Tregs in the tumor-infiltrating T cell population was not altered (Supplementary Figure S4B). However, μMT mice have an altered immune system, with well-described skewing toward Th1 responses (51, 52). Indeed, in our experiments, we confirmed μMT mice with orthotopic tumors had a higher proportion of CD8⁺ to CD4⁺ T cells in the spleen (Figure 4E), whilst no change in T cell proportions in the tumor itself (Figure 4F).

Since our data indicated B cells to infiltrate orthotopic B cells poorly, we speculated that the reduced tumor growth observed in μMT mice was due to their intrinsically different immune system and especially their T cell phenotype. In order to confirm this finding, we used an alternative method to study B-cell loss in orthotopic tumor development. To this end, we depleted B cells in WT mice using an anti-CD20 antibody both before and during tumor progression (Figures 4G,J). We observed complete ablation of B cells at endpoint (Figure 4H and Supplementary Figure S4C). In contrast to our results obtained with μMT mice, B-cell depletion with anti-CD20 did not cause any reduction in tumor mass in either experimental setting (Figures 4I,K), further suggesting that intrinsic characteristics of B-cell deficient μMT mice may contribute to the effect on orthotopic tumor development, rather than the loss of B cells per se.

Our data thus far indicated that, unlike in orthotopic PDAC, proinflammatory B cells heavily infiltrate KPC PDAC and these tumors display substantial deposition of immunoglobulins. If B cells exerted a protumor effect, then B cell deletion from KPC mice should result in reduced tumor growth. To this end, KPC mice with early stage, palpable PDAC tumors were treated for 5–6 weeks with B-cell depleting anti-CD20 antibody, in order to assess any potential therapeutic benefit (Figure 5A).

KPC tumors were then harvested and their immune infiltrate and tumor content analyzed. Treatment with anti-CD20 effectively depleted B cells in the pancreatic tumors (Figure 5B), spleen, MLN and blood (Supplementary Figure S5A). Plasma cells downregulate CD20 following differentiation, thus our treatment should not deplete pre-existing plasma cells. However, we found a significant decrease of plasma cells in the spleen (Supplementary Figure S5B), indicating that B cell depletion successfully blocked further plasma cell differentiation taking place during tumor progression. Accordingly, anti-CD20 treatment also strongly diminished immunoglobulin deposition within the tumor (Figure 5C and Supplementary Figures S5C,D). Nonetheless, tumor weight was not altered (Figure 5D). It is important to note that the KPC model is characterized by a highly heterogenous tumor growth, which varied independent of treatment. To overcome this heterogeneity, we randomly assigned mice to treatment arms and performed further histological analysis to determine if more subtle changes had occurred in response to B cell depletion. We assessed the prevalence of each PanIN subtype (Figure 5E) or percentage of PDAC content (Figure 5F and Supplementary Figure S5E), which also revealed no difference between treatment groups. Further staining for the epithelial marker cytokeratin 8 (CK8) confirmed this result (Figure 5G and Supplementary Figure S5F). We also analyzed the tumor microenvironment and found that the infiltration of T cells and myeloid subsets into the tumor was unchanged by B-cell depletion, as measured by flow cytometry (Supplementary Figure S5G). Taken together, these data indicate that B-cell depletion does not lead to reduced tumor growth in the more relevant spontaneous KPC model, demonstrating B cells do not drive PDAC progression.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.