Abstract

   Pancreatic ductal adenocarcinoma (PDAC) is a devastating disease,
   necessitating approaches to improve prognosis. As the mediator of
   allergic process, mast cells have been found in various cancers and are
   associated with survival. However, the biological behaviors of
   tumor‐associated mast cells (TAMCs) remain unclear. Herein, an
   excessive infiltration of TAMCs in PDAC is demonstrated, which
   apparently associated with poor survival in PDAC patients. PDAC cells
   are found to recruit CXCR2^+ MCs into TME, and then inhibited MCs
   ferroptosis, and maintained their proliferation. Concomitantly, the
   tumor‐derived exosome miR‐188‐5p activated the PTEN/AKT/GSK3β
   signaling, further stabilized transcriptional factor ERG by inhibiting
   its ubiquitin degradation, and finally enhanced the transcription of
   cxcl10 within TAMCs. In reverse, TAMCs‐derived CXCL10 reversely
   promoted tumor epithelial‐mesenchymal transition and induced
   immunosuppressive tumor microenvironment by recruiting CXCR3^+ Tregs.
   Sodium cromoglycate (SCG) is a membrane stabilizer for MCs and
   confirmed as an effective and widely used agent to block TAMCs‐derived
   CXCL10 and further sensitize the therapeutic efficacy of anti‐PD‐1
   antibody plus gemcitabine for PDAC. These findings illuminate a
   critical and innovative crosstalk between TAMCs and PDAC cells that
   promote PDAC progression, and SCG sensitizes PDAC to the current
   immuno‐chemotherapy, which reveals its potential to be a valuable
   adjuvant for PDAC patients.

   Keywords: CXCL10, immune escape, pancreatic ductal adenocarcinoma,
   sodium cromoglycate, tumor‐associated mast cell
     __________________________________________________________________

   Pancreatic cancer promotes the accumulation of mast cells in the tumor
   microenvironment. Tumor‐derived exosomes deliver miR‐188‐5p to
   tumor‐associated mast cells (TAMCs), enhancing TAMCs‐derived CXCL10
   secretion by strengthening ERG transcription. TAMCs‐derived CXCL10
   promotes tumor migration, stemness, and immunosuppressive environment.
   Sodium cromoglycate disrupts TAMCs‐tumor crosstalk by inhibiting the
   secretion of CXCL10 from TAMCs, thereby sensitizing the therapeutic
   effects of immuno‐chemotherapy.

   graphic file with name ADVS-12-2417724-g002.jpg

1. Introduction

   Pancreatic ductal adenocarcinoma (PDAC) is a highly malignant type of
   cancer characterized by its resistance to chemotherapy and radiation
   therapy. Currently, it stands as the estimated third leading cause of
   cancer‐related death in the United States, with only a 13% 5‐year
   survival rate.^[ [64]^1 ^] Despite advancements in PDAC treatment,
   long‐term survival rates have shown limited improvements over the past
   few decades. Therefore, it is imperative to better understand the
   characteristics of PDAC to explore novel and effective strategies to
   improve the current therapeutic efficacy.

   There is a growing belief that the intricate interplay between tumor
   cells and the surrounding microenvironment significantly fuels tumor
   progression and affects patient outcomes.^[ [65]^2 ^] PDAC is notable
   for its distinct histological characteristics, which encompass intense
   stroma desmoplasia and an abundance of fibroblasts, immune cells, and
   aberrant vascular structures. Stroma and immune cells actually augment
   the malignant attributes of tumor cells.^[ [66]^3 ^] Among them,
   myeloid lineage cells make up the largest immune cell population in
   PDAC tumor, which have a powerful impact on PDAC patient outcomes. For
   instance, myeloid‐derived suppressor cells (MDSCs) or tumor‐associated
   macrophages (TAMs) can establish a tolerant tumor microenvironment
   (TME) by manipulating T cells phenotype and disrupting the
   immuno‐oncology cycle.^[ [67]^4 ^]

   Mast cells (MCs) constitute vital components of myeloid cells, and
   participate in allergy by secreting histamine and protease.
   Additionally, MCs modulate tissue repair, angiogenesis, and immune
   responses by interacting with other cells.^[ [68]^5 ^] Recently, MCs
   have gained attention for their role in malignancies.
   Tumor‐infiltrating MCs actively participate in tumor progression, and
   are associated with poor prognosis in various cancers, such as
   cholangiocarcinoma,^[ [69]^6 ^] breast cancer,^[ [70]^7 ^] and gastric
   cancer.^[ [71]^8 ^] Although the relationship between MCs and
   tumorigenesis remains controversial in PDAC mouse model,^[ [72]^9 ^] it
   is undisputed that high tumor MCs infiltration was correlated with poor
   survival, demonstrating the pro‐tumoral capacity of MCs in progressed
   tumor.^[ [73]^10 ^] However, the potential mechanism by which MCs
   promote tumor progression in PDAC remains unclear and requires further
   study.

   Transcriptional analysis of MCs’ originating from various tissues has
   revealed their transcriptional diversity^[ [74]^11 ^] Moreover, within
   the same tissue, the heterogeneous distribution of cytokines can give
   rise to distinct subpopulations, also shown the heterogeneity of mast
   cells.^[ [75]^12 ^] Thus, the unique transcriptional characteristics of
   tumor‐educated MCs, known as tumor‐associated mast cells (TAMCs), needs
   to be clear in PDAC. In this study, we have demonstrated that PDAC
   patients with a high density of TAMCs in the TME had poor outcomes.
   PDAC cells recruited CXCR2^+MCs, and orchestrated their biological
   behavior, resulting in the accumulation in TME. Furthermore,
   tumor‐derived exosome miR‐188‐5p promoted the transcription of CXCL10
   in TAMCs via the PTEN/AKT/GSK3β/ERG axis. Reversely, CXCL10 produced by
   TAMCs directly enhanced cancer cell migration and stemness through
   PPARγ signaling, and promoted immune escape via recruiting
   immunosuppressive Tregs. Finally, sodium cromoglycate (SCG), a membrane
   stabilizer, suppressed PDAC dissemination, and enhanced sensitivity to
   the combination therapy of αPD‐1 antibody and gemcitabine. Our findings
   demonstrate the intricate crosstalk between TAMCs and PDAC cells, and
   propose SCG as a potential modifier for the current immuno‐chemotherapy
   in PDAC.

2. Results

2.1. Mast Cell Density is Increased in PDAC Tissues and Correlates with Poor
Prognosis

   To investigate the TAMCs infiltration in PDAC, we reanalyzed previous
   RNA‐seq data from five PDAC tissues and their corresponding adjacent
   normal tissues using CIBERSORT analysis.^[ [76]^13 ^] The analysis
   revealed a significant accumulation of TAMCs in PDAC tissues compared
   to their adjacent normal tissues (Figure [77] 1a). Toluidine blue,
   c‐kit, and TBSAB1 were stained in PDAC for mast cell identification.
   Among them, TPSAB1 exhibited the most distinct and identifiable
   staining (Figure [78]S1, Supporting Information). The distribution of
   TAMCs was further validated in tissue microarrays (TMAs) from two
   separate tertiary medical centers. The results revealed that TAMCs were
   primarily distributed in the tumor interstitium (Figure [79]1b), and
   they were significantly enriched in tumor tissues compared to their
   adjacent normal tissues in both ZS and JHH cohorts (Figure [80]1c,d).
   PDAC patients with higher TAMCs infiltration were also found to have
   significantly shorter overall survival (OS) than those with lower
   infiltration in TCGA cohort (Figure [81]1e).

Figure 1.

   Figure 1
   [82]Open in a new tab

   Mast cells excessively accumulate in PDAC tissues and correlate with
   poor prognosis. a) Mast cell infiltration analysis from transcriptome
   data of five PDAC tissues and paired adjacent normal tissues via
   CIBERSORT analysis. b) Representative images of TPSAB1 staining in PDAC
   tissues and adjacent normal tissues. Scale bar: 200µm. c) and d)
   Quantification and comparison of mast cells in tumor tissues and
   adjacent normal tissues from JHH (Johns Hopkins Hospital) cohort (c)
   and ZS (Zhongshan Hospital) cohort (d) by two‐sample t test and paired
   t test analysis, respectively. e) The Kaplan‐Meier plot of overall
   survival in PDAC stratified by mast cell density using CIBERSORT
   algorithm from TCGA database. f) and g) Kaplan‐Meier plot of overall
   survival (f) and recurrence free survival (g) in PDAC from the ZS
   cohort stratified by mast cells density. h) The visualization of
   independent prognostic factor for overall survival and recurrence free
   survival in the ZS cohort via multivariate Cox regression analysis.

   The clinical relevance and prognosis of TAMCs were further validated in
   the ZS cohort. The entire cohort was divided into the TAMC^hi and
   TAMC^low group according to the calculated optimal cut‐off value (100.5
   cells). Under the correlation analysis, only tumor differentiation was
   significantly associated with TAMCs density in PDAC (P = 0.020), while
   no significant relationship was found with others (Table [83]S1,
   Supporting Information). The Kaplan‐Meier analysis revealed that the
   median OS (18.0 versus 29.0 months, P = 0.025, Figure [84]1f) and
   recurrence‐free survival (RFS, 12.0 versus 22.0 months, P = 0.007,
   Figure [85]1g) in the TAMC^hi group were significantly shorter than
   those in the TAMC^low group. After adjusted analysis, TAMCs density in
   PDAC was finally eventually identified as the only significant
   independent risk indicator for both OS (P = 0.027, HR = 1.891, 95%CI:
   1.077‐3.321) and RFS (P = 0.017, HR = 1.859, 95%CI: 1.115‐3.099)
   (Figure [86]1h; Table [87]S2, Supporting Information). Collectively,
   our results demonstrated that higher density of TAMCs within TME was
   correlated with poor outcome in PDAC patients.

2.2. Pancreatic Cancer Cells Promote Migration and Sustain Proliferation of
Mast Cells

   Stem cell factor (SCF) plays a crucial role in MC proliferation,
   survival, and chemotaxis.^[ [88]^14 ^] We found that PDAC cell lines
   constitutively expressed SCF (Figure [89]S2a, Supporting Information).
   Moreover, SCF expression in tumor positively correlated with TAMCs
   infiltration, as determined by xCell and CIBERSORT (Figure [90]S2b,
   Supporting Information), which suggested that SCF might participated in
   TAMCs recruitments. To ascertain this hypothesis, human peripheral mast
   cells (hMCs) were cultured in vitro (Figure [91]S2c, Supporting
   Information) and confirmed as c‐kit^+FcεR1^+ cells by flow cytometry
   (Figure [92]S2d, Supporting Information). Blocking SCF/CD117 axis on
   hMCs by CD117 antibody could partially reverse the recruitment ability
   induced by tumor‐conditioned medium (Figure [93] 2a), which is
   consistent with previous studies.^[ [94]^15 ^] This indicates that SCF
   is partially involved in the recruitment of MCs in PDAC.

Figure 2.

   Figure 2
   [95]Open in a new tab

   PDAC recruits mast cell infiltration and promotes its proliferation. a)
   Mast cell migration assay comparing the recruitment ability of
   tumor‐conditioned medium. Surgical resected PDAC tissues (n = 5) were
   digested into single cells and resuspended by DMEM. After culturing for
   24h, supernatant was harvested as tumor‐conditioned medium. Anti‐CD117
   was used at 10µg mL^−1 to block SCF/CD117 signaling. b) Schematic
   representation of mast cell migration assay and qRT‐PCR analysis. Mast
   cells were inoculated into the upper chamber for 12 h and collected
   from both upper and lower chambers. c) Relative expression of chemokine
   receptors in migrated mMCs via qRT‐PCR analysis. The expression of
   chemokine receptors was calculated by ΔΔCT^upper/ΔΔCT^upper and
   normalized to control groups. n = 4 per group. d) Representative
   histogram and statistical analysis of CCR5, CXCR1, and CXCR2 expression
   on migrated hMCs via flow cytometry. n = 4 per group. e) Representative
   histogram images and normalized heatmap of chemokine receptor
   expression in tumor‐associated mast cells in PDAC (n = 5). Mast cells
   were gated on CD45^+, CD117^+, and FcεRI^+. f) Representative
   immunofluorescence images of CXCR2 and TPSAB1 co‐staining in PDAC
   tissues and quantification for the ratio of CXCR2^+TAMCs to total
   TAMCs. TPSAB1, green; CXCR2, red; DAPI, blue. Red arrow:
   CXCR2^+TPSAB1^− cells; White arrow: CXCR2^+TPSAB1^+ TAMCs; Green arrow:
   CXCR2^−TPSAB1^+ MCs; Scale bars: 100µm. g) KEGG pathways were enriched
   based on DGEs from mast cells co‐cultured with tumor cells or not. h)
   Proliferated tumor‐associated mast cells were defined as Ki‐67^+
   TPSAB1^+ double positive cells by immunofluorescence staining. TPSAB1,
   green; Ki‐67, pink; cytokeratin 19, orange; DAPI, blue. Scale bars:
   50µm. i) Heatmap of proliferation‐related gene expressions from
   transcriptional RNA sequencing for mast cell following indicated
   treatments. j) Representative histogram of CFSE staining and the ratios
   of proliferative cells to total cells from indicated groups. n = 3 per
   group. mMCs were labeled by CFSE and stimulated by tumor‐conditioned
   medium or DMEM as control for 48 hours. k) Representative histogram of
   cell cycle staining and the statistical analysis of the percentage in
   each cell cycle phase from indicated groups. mMCs were stimulated by
   tumor‐conditioned medium or DMEM as control for 48 hours, and then
   stained by PI for cell cycle analysis. n = 4 per group. *P < 0.05, **P
   < 0.01, and ***P < 0.001, Data were displayed as mean ± SD.

   It has been reported that CXCR family also participates in MCs
   migration in gastric cancer.^[ [96]^16 ^] To further explore the
   potential mechanism, murine bone marrow‐derived mast cells (mMCs) were
   expanded in vitro (Figure [97]S3a, b, Supporting Information), and
   identified by immunostaining (Figure [98]S3c, Supporting Information)
   and flow cytometry (Figure [99]S3d, Supporting Information). Similarly,
   a significant chemotaxis of mMCs was observed when co‐cultured with KPC
   and Panc02 cell lines (Figure [100]S3e, Supporting Information).

   To identify the specific chemokine receptors involved in mast cell
   migration, a migration assay was performed as previously reported^[
   [101]^16 ^] (Figure [102]2b). The results showed that ccr5, cxcr1, and
   cxcr2 were highly expressed in the low‐chamber mMCs in both murine PDAC
   cell lines, suggesting their potential role in the chemotaxis process
   (Figure [103]2c), but only CXCR2^+hMCs were recruited by human PDAC
   cell lines (Figure [104]2d). Similarly, TAMCs had higher CXCR2
   expression when compared to MCs in adjacent normal pancreas
   (Figure [105]2e). CXCR2 is an essential chemotactic receptor in myeloid
   cells. The dual immunofluorescence analysis revealed that majority of
   TAMCs strongly expressed CXCR2 in tumor lesion (Figure [106]2f). CXCR2
   receptor blockade significantly reversed the chemotaxis effect of PDAC
   conditioned medium (Figure [107]S3f, Supporting Information). Six
   potential chemokines that could interact with CXCR2 were upregulated in
   tumor tissues in most public databases, particularly CXCL5 and CXCL8
   (Figure [108]S3g, h, Supporting Information). Thus, these findings
   collectively suggest that pancreatic cancer could recruit CXCR2^+MCs
   into TME and exert pro‐tumor biological behavior.

   The RNA‐seq revealed a significant enrichment of cell cycle pathway MCs
   co‐cultured with PDAC cells, indicating that tumor cells may influence
   the proliferation of MCs (Figure [109]2g). Furthermore, the dual
   immunofluorescence analysis showed that a majority of TAMCs exhibited
   abundant expression of Ki‐67, suggesting their increased proliferative
   ability in the TME (Figure [110]2h). Additionally, the cell cycle
   related genes, including CDK family and CCND1, were significantly
   upregulated in MCs after tumor education (Figure [111]2i). Then, the
   increased proliferative ability of MCs stimulated by tumor‐conditioned
   medium also verified by flow cytometry in vitro (Figure [112]2j,k).
   Traditionally, MCs can be activated by IgE and play a traditional role
   in allergic disorder by secreting inflammatory mediators, including
   arachidonic acid metabolic products, histamine, and inflammatory
   cytokines.^[ [113]^17 ^] However, we found that TAMCs downregulated the
   expression of the IgE receptor, Fcer2a. Additionally, the expression of
   phospholipase A2 initiation enzyme (pla2g5), leukotrienes‐related
   enzymes (ltc4s and lta4h), or prostaglandin‐related enzymes (ptgs2 and
   ptgds) were also decreased in TAMCs. Cytokines from the interleukin
   family, which are important in IgE‐related inflammation, were also
   decreased in TAMCs (Figure [114]S3i, Supporting Information).
   Degranulation assay also demonstrated that TAMCs had lower histamine
   release ability (Figure [115]S3j, Supporting Information). These
   results indicated that TAMCs had lost their traditional allergic
   response characteristics and exhibited unique biological behaviors
   instead.

2.3. Pancreatic Cancer Cell‐Derived Exosomes Restrain Mast Cell Ferroptosis
through PI3K/AKT Signaling

   To confirm whether pancreatic cancer cells could inhibit the death of
   MCs, we applied different cell death inducers as previously reported.^[
   [116]^18 ^] Our results showed that PDAC cell‐conditioned medium
   effectively rescued cell death induced by H[2]O[2] ^[ [117]^19 ^] and
   RSL3,^[ [118]^20 ^] suggesting that tumor‐conditioned medium might
   protect MCs from ferroptosis (Figure [119] 3a). Subsequently,
   KPC‐conditioned medium was shown to reduce Erasin‐induced lipid ROS
   accumulation (Figure [120]3b) and cell death (Figure [121]3c) in MCs,
   and this effect was similarly observed in Panc02‐conditioned medium
   (Figure [122]S4a, Supporting Information). In addition to lipid ROS
   accumulation and elevated cell death, intracellular ROS and Fe^2+ are
   also indicators of ferroptosis. Pretreatment of MCs with KPC
   supernatant significantly rescued the intracellular ROS
   (Figure [123]3d) and Fe^2+ ion (Figure [124]3e; Figure [125]S4b,
   Supporting Information) induced by Erastin in mMCs.

Figure 3.

   Figure 3
   [126]Open in a new tab

   Tumor‐conditioned medium inhibits mast cell ferroptosis via AKT pathway
   activation. a) The death rate of indicated cells following treatment
   with different cell death inducers. RSL3: ferroptosis activator;
   Staurosporine(STS): apoptosis activator; hydrogen peroxide(H[2]O[2]):
   ROS stress inducer; Rapamycine(Rapa): autophagy inducer). Cells were
   pre‐treated by KPC‐conditioned medium or DMEM as control for 12 h, then
   stimulated by death inducers for 24 h, and stained by PI for evaluation
   of cell death. n = 5 per group. b) and c) The lipid ROS levels (b) and
   death rates (c) were compared between mast cells treated by
   KPC‐conditioned medium or not for 24 h. d) and e) Flow cytometry
   analysis for ROS levels (d) and Fe^2+ levels (e) in mast cells
   following indicated treatments. n = 4 per group. f) Western blot
   analysis for ACSL4, SLC7A11, and GPX4 expression in mast cells after
   indicated treatment for 24 h. g) and h) Flow cytometry analysis for
   lipid ROS levels (n = 4). g) and cell death rates (n = 6). h) in mast
   cells after conditioned medium with or without exosomes. i) Histogram
   showing the diameter distribution of KPC‐derived exosomes. j)
   KPC‐derived exosomes were pre‐stained by PKH26 (membrane dye for
   cell‐tracing assay) and used for exosome intake assay by co‐culturing
   with mast cells. k) Flow cytometry analysis for mast cell proliferation
   by CFSE after indicated treatment for 48 hours. n = 4 per group. l)
   Representative image of three PDAC patient‐derived organoids. m) Mast
   cells co‐cultured with three patient‐derived organoids and then treated
   with Erastin to detect lipid ROS levels. n = 4 per group. n) The effect
   of AKT pathway inhibitors including MK‐2206 (AKT phosphorylation
   inhibitor, 3µM) and LY294002 (PI3K inhibitor, 10µM) on lipid ROS (left
   panel) and cell death rate (right panel) of mast cells after the
   treatment of Erastin and KPC‐conditioned medium. n = 4 per group. o)
   The expression of GPX4 and SLC7A11 in mast cells after they were
   pre‐treated with AKT inhibitors and stimulated by KPC‐conditioned
   medium (upper) or Panc02‐conditioned medium (lower). n.s. P > 0.05, *P
   < 0.05, **P < 0.01, and ***P < 0.001, Data were displayed as mean ± SD.

   Ferroptosis is governed by the process of membrane lipid peroxidation
   and inner defenses. We analyzed the expression of three core proteins
   as the representation for lipid hydroperoxides generation and
   elimination process in ferroptosis. Tumor‐conditioned medium
   significantly upregulated cystine/glutamate transporter (solute carrier
   family 7 member 11, SLC7A11) and the key enzyme for lipid hydroperoxide
   elimination (glutathione peroxidase 4, GPX4) expression in mMCs, both
   of which negatively regulate ferroptosis by importing L‐cystine and
   catalyzing the degradation of lipid peroxides (Figure [127]3f).
   However, Acyl‐CoA synthetase long chain family member 4 (ACSL4), which
   participates in polyunsaturated fatty acid‐containing phospholipids
   (PUFA‐PLs) synthesis, was also found to be aberrantly increased after
   tumor conditioned medium treatment.

   Given that MCs are primarily located in the stroma and distant from
   PDAC cells, we hypothesized that tumor cells may educate TAMCs in
   non‐contact dependent way. To determine which pathway mainly mediated
   this process, N‐Smase inhibitor GW4869 was utilized to block exosome
   secretion. We observed that the protective effect of tumor‐conditioned
   medium could be partially blocked by GW4869, suggesting the involvement
   of exosomes derived from KPC (Figure [128]3g,h) and Panc02 (Figure
   [129]S4c, d, Supporting Information) in this process. The KPC‐derived
   exosomes were separated by ultracentrifugation and further identified
   (Figure [130]3i; Figure [131]S4e, f, Supporting Information). PKH26
   labeled tracking assay showed the uptake of KPC‐derived exosomes by MCs
   (Figure [132]3j). Exosome depletion in tumor conditioned medium did not
   impair mMCs proliferation, revealing that PDAC cells promoting MCs
   proliferation are exosome‐independent (Figure [133]3k; Figure [134]S4g,
   Supporting Information). Moreover, three PDAC patient‐derived organoids
   were established (Figure [135]3l), and a significant decreased
   Erastin‐induced ferroptosis in hMCs was observed when co‐cultured with
   organoid hMCs (Figure [136]3m).

   AKT pathway was significantly enriched (Figure [137]2h) and confirmed
   (Figure [138]S4h, Supporting Information) in TAMCs, which have been
   reported as regulator for ferroptosis process.^[ [139]^21 ^] We
   suspected that tumor conditioned medium mediated resistance to
   ferroptosis in TAMCs was AKT pathway dependent. As expected, the
   inhibitors of AKT pathway eliminated the protective effect of
   KPC‐conditioned medium (Figure [140]3n) and Panc02‐conditioned medium
   (Figure [141]S4i, j, Supporting Information). The expression of GPX4
   and SLC7A11 was also restored after exposure to AKT inhibition,
   suggesting that AKT activation mediated by the conditioned medium
   participated in the ferroptosis process of MCs (Figure [142]3o).
   Collectively, our findings suggest that PDAC cell‐derived exosomes
   inhibit cell death by preventing the ferroptosis cascade.

2.4. Sodium Cromoglycate Abrogates PDAC Progression Fostered by
Tumor‐Associated Mast Cell Derived CXCL10

   RNA‐seq analysis identified twelve cytokines that exhibited notably
   changed in TAMCs compared with MCs (Figure [143] 4a) and validated by
   qRT‐PCRs (Figure [144]4b). Notably, cxcl10 mRNA showed the highest
   variation after co‐culturing and was the only prognostic indicator
   among them (Figure [145]S5a, Supporting Information), suggesting the
   potential role of TAMCs‐derived CXCL10 in promoting tumor progression.
   The ELISA analysis revealed a significant increase of CXCL10 secretion
   by mMCs upon culturing with PDAC conditioned medium (Figure [146]4c).
   Concomitantly, the expression of cxcl10 in PDAC was significantly
   higher compared to their adjacent normal tissues (Figure [147]S5b, c,
   Supporting Information). CXCL10 was significantly associated with
   shorter OS in PDAC patients from the TCGA cohort (1.09 years versus
   1.81 years, P < 0.001, Figure [148]S5d, Supporting Information) and
   CTPAC cohort (0.95 year versus 2.18 years, P < 0.001, Figure [149]S5e,
   Supporting Information). Among tumor‐infiltrating immune cells, TAMCs
   are identified as an important source of CXCL10 in PDAC tissues by flow
   cytometry (Figure [150]4d,e), and immunofluorescence staining
   (Figure [151]4f).

Figure 4.

   Figure 4
   [152]Open in a new tab

   Tumor associated mast cell‐derived CXCL10 promotes tumor progression.
   a) Heatmap showing the significant changes of cytokines in both KPC and
   Panc02 educated mast cells by RNA‐seq. Mast cells were co‐cultured with
   tumor cells or stimulated by tumor‐conditioned medium for 24 hours. b)
   qRT‐PCR analysis for indicated cytokine transcriptions in mast cells
   that co‐cultured with KPC cells or not for 24 h. n = 4 per group. c)
   ELISA assay analysis for CXCL10 levels in mMCs supernatant after mMCs
   stimulated by indicated tumor‐conditioned medium. n = 4 per group. d)
   Flow cytometry analysis for CXCL10 expression in tumor‐infiltrating
   immune cells. The cell suspension was first gated by single cells and
   then CD45^+CXCL10^+ double positive cells were further divided into
   four types of immune cells according to the corresponding
   representative markers. CD68, macrophages; CD11c, dendritic cells;
   CD66b, neutrophils; c‐kit, mast cells. e) Flow cytometry analysis for
   CXCL10 expression in tumor‐infiltrating macrophages, dendritic cells,
   neutrophils and mast cells in PDAC tissues. The representative
   histogram (left) and relative normalized MFI expression (right) in four
   types of tumor‐infiltrating immune cells were displayed. f)
   Representative immunofluorescence staining of CXCL10 and TPSAB1 in PDAC
   tissues. TPSAB1, green; CXCL10, red; DAPI, blue. Green arrow: TAMCs;
   white arrow: tumor cells; orange arrow: other cells with CXCL10
   positive expression. g) Schematic representation of mMC adoptive
   transfer in murine models. Blank: without MCs adoptive transfer;
   CXCL10^NC: adoptive transfer MCs with negative control plasmid
   transfection; CXCL10^sh1 and CXCL10^sh2: adoptive transfer MCs with
   CXCL10 knockdown plasmid 1 and plasmid 2 transfection. h)
   Representative H&E staining and immunofluorescence staining for
   tumor‐infiltrating engineered mMCs labeled by GFP. The tumor tissues
   were collected after mMCs adoptive transferred for 24 or 48 hours. i)
   Kaplan‐Meier analysis of survival for orthotopic PDAC mice received
   mMCs adoptive transfer or not. Survival curves were analyzed by
   log‐rank tests. n = 8 per group. j) Tumor growth curves in KPC (left)
   and Panc02 (right) subcutaneous PDAC models with SCG intraperitoneal
   injection or PBS as control in C57BL/6. Student's t test was used for
   comparison. n = 6 per group. n.s. P > 0.05, *P < 0.05, **P < 0.01, and
   ***P < 0.001, Data were displayed as mean ± SD.

   To investigate the impact of TAMC‐derived CXCL10 in vivo, we
   constructed CXCL10 knockdown mMCs with GFP overexpression (Figure
   [153]S6a, Supporting Information) and validated it (Figure [154]S6b,
   Supporting Information). The gene‐manipulated mMCs were then adoptively
   transferred to an orthotopic murine PDAC model (Figure [155]4g). After
   adoptive transfer, the mMCs primarily distributed to peripheral tumor
   tissues within 24 h and infiltrated into tumor lesions at 48 h
   (Figure [156]4h), confirming the recruitment of circulating MCs into
   the PDAC TME in vivo. Mice transplanted with CXCL10^sh mMCs had
   significantly reduced tumor burden (Figure [157]S6c, d, Supporting
   Information) and prolonged survival compared to those transplanted with
   CXCL10^NC mMCs (Figure [158]4i), highlighting the pro‐tumoral role of
   TAMCs‐derived CXCL10 in vivo.

   It has been reported that sodium cromoglicate (SCG) can inhibit the
   degranulation and cytokines release in MCs.^[ [159]^8 , [160]^22 ^] In
   this study, we found that SCG could also decrease CXCL10 secretion in
   vitro, suggesting a therapeutic effect by inhibiting mMCs function
   (Figure [161]S6e, Supporting Information). As expected, SCG
   administration significantly restrain tumor growth (Figure [162]4j;
   Figure [163]S6f, g, Supporting Information). Notably, SCG did not
   directly affect cancer cell proliferation (Figure [164]S6h, Supporting
   Information) and CXCL10 secretion (Figure [165]S6i, Supporting
   Information) in vitro, suggesting that the therapeutic effect of SCG
   dependent on inhibiting functions of TAMCs. Together, our observations
   support the idea that TAMCs promote PDAC progression through the
   secretion of CXCL10, and this process can be abrogated by SCG.

2.5. Tumor‐Derived Exosome miR‐188‐5p Promotes CXCL10 Transcription via
PTEN/AKT/ERG Axis

   A significant upregulation of cxcl10 mRNA in TAMCs, indicating that
   CXCL10 was regulated at the transcriptional level. We found that AKT
   inhibition effectively resorted CXCL10 transcription in TAMCs
   (Figure [166] 5a; Figure [167]S7a, Supporting Information),
   Furthermore, tumor‐conditioned medium without exosomes not only
   suppressed the transcriptional activity of CXCL10 (Figure [168]5b;
   Figure [169]S7b, c, Supporting Information), but also significantly
   decreased AKT activation (Figure [170]5c).

Figure 5.

   Figure 5
   [171]Open in a new tab

   Tumor‐derived exosome miR‐188‐5p promotes CXCL10 transcription via
   PTEN/AKT/ERG axis in tumor‐associated mast cells. a) ELISA assay
   analysis for supernatant CXCL10 levels from tumor‐associated mast cells
   with or without AKT inhibitor pretreatment. n = 4 per group. b) ELISA
   assay analysis for supernatant CXCL10 level from tumor‐associated mast
   cells stimulated by conditioned medium with or without exosomes. n = 4
   per group. c) Western blot analysis for AKT pathway activation in mast
   cells stimulated by normal or exosomes‐depleted KPC or Panc02
   conditioned medium for 24 h. d) Western blot analysis for AKT pathway
   activation in tumor‐associated mast cells after miR‐188‐5p transfected.
   e) Western blot analysis for AKT pathway activation in tumor‐associated
   mast cells. KPC were transfected with miR‐188‐5p or negative control
   for 48h, and then the conditioned medium was harvested and used for
   mast cells stimulation. f) ELISA assay analysis for CXCL10 levels in
   tumor‐associated mast cell supernatant after indicated conditioned
   medium stimulation. n = 4 per group. miR^i, miR‐188‐5p inhibitor. NC,
   negative control. g) The predicted binding site of miR‐188‐5p and PTEN
   3′UTR region. h) Integrated bioinformatics analysis for potential
   transcriptional factors (TF) that could bind to Cxcl10 promoter. TSPP
   ([172]https://www.signalingpathways.org/index.jsf) and UCSC
   ([173]https://genome.ucsc.edu/) database were used for predicting TF
   binding to cxcl10 promoter. KnockTF
   ([174]https://bio.liclab.net/KnockTF/search.php) were used for
   analyzing activated TF via different gene expressions. i) Western blot
   analysis for ERG nuclear translocation in tumor‐associated mast cells
   following AKT pathway inhibitor. j) ELISA assay analysis for CXCL10
   levels in tumor‐associated mast cell supernatant after ERG knockdown by
   si‐RNA. n = 4 per group. k) ChIP‐PCR assay for ERG binding sites to
   Cxcl10 promoter following indicated treatments. n = 3 per group. l)
   Schematic representation of ERG binding sites in JARSPAR database
   (upper) and dual‐luciferase reported assay (lower) for binding sites
   identification. Student's t test was used for comparison. n = 4 per
   group. The difference of three mutated plasmids is described in
   Supplementary Figure [175]9D. m) Western blot analysis for the effect
   of tumor‐conditioned medium on the half‐life of ERG in tumor‐associated
   mast cells treated with CHX (100µg mL^−1) for indicated incubation
   time. n) Western blot analysis for the effect of KPC and Panc02
   conditioned medium on the ubiquitination level of ERG in mMCs. o)
   Schematic of the domain structure of ERG and the mutated site. p) The
   ubiquitination level of ERG T187A mutated protein after
   tumor‐conditioned medium stimulation in the presence of MG132 in mMCs.
   q) Western blot analysis for IP enriched cell lysates from mMCs. mMCs
   were transfected with Flag‐ERG^WT, Flag‐ERG^T187A for 48 hours and
   lysed for co‐IP assay. ERG was enriched by Flag antibody or isotype IgG
   as control. n.s. P > 0.05, *P < 0.05, **P < 0.01, and ***P < 0.001,
   Data were displayed as mean ± SD.

   Non‐coding RNA can be packaged into exosomes and affect target cells
   behavior. We enriched the target genes of differentially expressed
   miRNA in exosomes from KPC and murine pancreatic ductal cell (MPDC) as
   previously reported ^[ [176]^23 ^] and found PI3K‐AKT pathway as
   prominently enriched pathway in this dataset (Figure [177]S7d,
   Supporting Information). Notably, we found a significant decrease
   expression of PTEN, an important negative regulator for AKT activation,
   after co‐culturing with PDAC cells, suggesting that it could be the
   mechanism underlying AKT pathway activation in TAMCs (Figure [178]5c).
   Subsequently, we screened out three miRNAs as potential candidates that
   were significantly upregulated in KPC exosomes and targeted PTEN. Among
   them, miR‐188‐5p was remarkably upregulated in TAMCs (Figure [179]S7e,
   Supporting Information), and overexpression of miR‐188‐5p in MCs
   activated AKT pathway (Figure [180]5d). Conversely, conditioned medium
   from tumor treated with a miR‐188‐5p inhibitor reversed the
   phosphorylation of AKT in mMCs (Figure [181]5e), and also restored
   CXCL10 secretion in TAMCs (Figure [182]5f; Figure [183]S7f, Supporting
   Information). Furthermore, we identified a highly conserved nucleotide
   binding site of miR‐188‐5p in PTEN (Figure [184]5g). The interaction of
   miR‐188‐5p and PTEN was confirmed in a luciferase assay, which has been
   reported in renal and vascular disease,^[ [185]^24 ^] further
   suggesting that miR‐188‐5p could be the key molecule responsible for
   AKT activation in TAMCs. Collectively, these results demonstrate that
   tumor‐derived exosome miR‐188‐5p promotes AKT activation in TAMCs by
   degrading PTEN.

   To further explore the transcriptional factors (TFs) that mediated this
   process, seven potential candidates were screened out by overlapping
   the TF predicted by UCSC, TSPP and knockTF database (Figure [186]5h).
   Subsequently, the correlations between TFs and CXCL10 in pan‐cancer
   datasets were analyzed. Top Five TFs with the closest correlation with
   CXCL10 were chosen as candidates and knocked down for further studies
   (Figure [187]S8a, b, Supporting Information). Among them, ERG and SP1
   knockdown was found to be significantly effective in downregulating
   CXCL10 expression in MCs (Figure [188]S9a, Supporting Information).

   ERG significantly translocated into nuclear after stimulation with
   tumor‐conditioned medium, which was reversed by AKT pathway inhibitors
   in mMCs (Figure [189]5i; Figure [190]S9b, Supporting Information).
   However, the nuclear translocation of SP1 showed no significant change
   (Figure [191]S9c, Supporting Information). Importantly, ERG knockdown
   effectively restored CXCL10 secretion in MCs (Figure [192]5j).
   Additionally, the ChIP assay showed that ERG could bind to the CXCL10
   promoter (Figure [193]5k). Mutated CXCL10 luciferase promoter plasmids
   were constructed to further validate the binding site between ERG and
   CXCL10 (Figure [194]5l; Figure [195]S9d, Supporting Information).
   Collectively, these results demonstrate that PDAC‐derived exosome
   miR‐188‐5p facilitates CXCL10 transcription through the PTEN/AKT/ERG
   axis.

   We found that tumor‐conditioned medium did not affect transcriptional
   expression (Figure [196]S9e, Supporting Information), but significantly
   increased ERG protein level (Figure [197]S9f, Supporting Information).
   This suggested that AKT may regulate ERG through post‐transcriptional
   modifications (PTMs). Cycloheximide inhibition assay revealed that
   tumor‐conditioned medium prolonged half‐life of ERG (Figure [198]5m),
   indicating that PTM‐mediated ERG accumulation in MCs is a key mechanism
   underlying the upregulation of CXCL10. As ubiquitination is an
   important mechanism in protein degradation, we analyzed the ubiquitin
   level of ERG and found that tumor supernatant decreased ERG
   ubiquitination (Figure [199]5n). A previous study reported that GSK3β
   could phosphorylate ERG^T187, thereby promoting its interaction with
   FBW7, an important E3 ubiquitin ligase.^[ [200]^25 ^] We demonstrated a
   significant increase in the phosphorylation of GSK3β following
   treatment with tumor‐conditioned medium (Figure [201]5c), indicating
   that GSK3β might play a similar role in inhibiting ERG ubiquitination
   in TAMCs. Furthermore, GSK3β knockdown in TAMCs could reserve the
   effect of tumor‐conditioned medium on ERG ubiquitination levels (Figure
   [202]S9g, Supporting Information).

   Next, the 3D structure of the mERG protein was analyzed (Figure
   [203]S9h, Supporting Information), and mut‐mERG^T187A plasmid were
   constructed as previously reported (Figure [204]5o).^[ [205]^25 ^]
   Notably, mut‐ERG^T187A plasmid abrogated the stabilizing effect of
   tumor‐conditioned medium (Figure [206]5p). Furthermore, the ERG
   mutation impaired the interaction between GSK3β‐ERG‐FBW7 complex
   (Figure [207]5q), indicating that FBW7 interacted with ERG and promoted
   its ubiquitination, depending on the phosphorylation of Thr^187. These
   results revealed that tumor‐conditioned medium increased ERG expression
   via PTMs, promoted ERG nuclear translocation, bind to the CXCL10
   promoter, and ultimately resulted in the upregulation of CXCL10
   transcription.

2.6. Tumor‐Associated Mast Cell‐Derived CXCL10 Promotes PDAC Cell Mobility
and Stemness via CXCR3‐PPARγ Axis

   CXCL10, also known as interferon‐inducible protein 10, interacts with
   CXCR3 and selectively attracts CXCR3^+T cells into the TME.^[ [208]^26
   ^] CXCL10 could also impact tumor growth, apoptosis and metastasis by
   targeting CXCR3 on tumor cells.^[ [209]^27 ^] To decipher the mechanism
   through which CXCL10 promotes malignant behavior in PDAC, we
   co‐cultured MCs with murine PDAC cells, and observed that TAMCs
   co‐culturing and rmCXCL10 stimulation could obviously promote tumor
   migration and invasion (Figure [210] 6a,b; Figure [211]S10a‐c,
   Supporting Information). However, no proliferation effect was observed
   after TAMCs treatment (Figure [212]S10d, Supporting Information),
   indicating that TAMCs indirectly promote tumor growth in vivo.

Figure 6.

   Figure 6
   [213]Open in a new tab

   CXCL10 activates PDAC tumor migration and stemness by activating PPARγ.
   a) and b) Wounding assay (a) and transwell assay (b) analysis for KPC
   and Panc02 migration ability by co‐culturing with mast cell for 24 h
   associated with CXCL10 antibody or isotype IgG as control, or
   stimulating with rmCXCL10 (100ng mL^−1) for 24 h. n = 4 per group. c)
   Quantification of migration ability by wounding scratch assay (n = 3)
   and transwell assay (n = 4) in CFPAC‐1 (left) and PANC‐1 (right)
   stimulated by rhCXCL10 (100ng mL^−1) or not. d) Schematic for KPC and
   Panc02 orthotopic tumor models that received tumor surgical resection
   after mMCs adoptive transfer or SCG treatment. Recurrence or metastasis
   after surgical resection were evaluated by IVIS imaging system. e)
   Representative optical image in KPC and Panc02 murine models (left)
   with different treatments, and quantification analysis (right) of
   luciferase optical density. n = 5 per group. f) Western blot analysis
   for EMT‐related proteins in cancer cells treated with rhCXCL10 for 24
   h. g) The representative image of immunofluorescence staining for ZEB1
   in PDAC patient‐derived organoid slices after indicated treatments. h)
   Western blot analysis for stemness markers in cancer cells treated with
   rhCXCL10 for 24 h. i) Sphere formation assay analysis for PANC‐1
   pre‐treated with rhCXCL10 or control for 24 h. n = 5 per group. j) KEGG
   functional enrichment analysis of DEGs between PANC‐1 stimulated by
   rhCXCL10 or not for 24 h. k) Western blot analysis for PPAR pathway in
   PANC‐1 and CFPAC‐1 stimulated by rhCXCL10 or not. l) Western blot
   analysis for PPARγ expression in the indicated cells. CXCR3 were
   knockdown by siRNA transfection in PANC‐1 and CFPAC‐1, and then
   stimulated by rhCXCL10 for 24 h. m) Dual‐luciferase assay analysis for
   PPAR transcriptional activity in PANC‐1 and CFPAC‐1 after rhCXCL10
   stimulation. n = 5 per group. n.s. P > 0.05, *P < 0.05, **P < 0.01, and
   ***P < 0.001, Data were displayed as mean ± SD.

   Similarly, CXCL10 does not directly affect tumor growth and apoptosis
   in vitro (Figure [214]S11a‐e, Supporting Information). However,
   rhCXCL10 stimulation significantly enhanced tumor migration
   (Figure [215]6c; Figure [216]S12a, b, Supporting Information) and
   invasion (Figure [217]S12c, Supporting Information). CXCL10
   overexpression in PDAC cells also increased liver metastasis tumor
   burden in vivo (Figure [218]S12d, Supporting Information). The
   association between CXCL10 and epithelial‐mesenchymal transition (EMT)
   or metastatic signatures were also identified in the TCGA database
   (Figure [219]S12e, Supporting Information). Furthermore, EMT‐related
   genes were significantly increased after CXCL10 stimulation (Figure
   [220]S12f, Supporting Information).

   Then, an orthotopic PDAC murine model was established and surgical
   resection for tumors was performed on Day 18 (Figure [221]6d). Adoptive
   transfer of mMCs significantly promoted distant metastasis and local
   recurrence after surgical resection, while SCG treatment significantly
   inhibited distant metastasis (Figure [222]6e; Figure [223]S12g,
   Supporting Information). Tumor cells invaded into adjacent pancreas in
   MCs‐AT group in resected tissues (Figure [224]S12h, Supporting
   Information). The RNA‐seq was also performed for PANC‐1 with or without
   CXCL10 stimulation, and showed that rhCXCL10 stimulation significantly
   upregulated N‐cadherin, Vimentin, and ZEB1 expression, coinciding with
   increasing mobility following rhCXCL10 stimulation (Figure [225]S13a,
   Supporting Information). Similar changes were observed in human PDAC
   cells (Figure [226]6f). In patient‐derived organoids upon rhCXCL10
   stimulation or coculturing with hMCs, ZEB1 was similarly obviously
   upregulated and underwent nuclear translocation (Figure [227]6g). The
   association between CXCL10 and the stemness signature was also observed
   in the TCGA database (Figure [228]S13b, Supporting Information).
   rhCXCL10 was found to upregulate stemness markers and spheres formation
   ability in vitro (Figure [229]6h,i; Figure [230]S13c, d, Supporting
   Information). Immunohistochemistry also validated the upregulated
   expression of EMT and stemness markers in murine orthotropic tumor
   tissues (Figure [231]S13e, Supporting Information). These results
   collectively indicate that CXCL10 can promote the mobility and stemness
   of PDAC cells.

   To further explore the underlying mechanism by which CXCL10 mediates
   mobility, KEGG pathway enrichment analysis was performed, and
   identified that the PPAR pathway was significantly enriched in the
   rhCXCL10 stimulated group (Figure [232]6j). PPARγ and its co‐activators
   PGC1α and PGC1β were significantly upregulated after CXCL10 stimulation
   (Figure [233]6k; Figure [234]S14a, Supporting Information). Intrinsic
   PPARγ knockdown reversed CXCL10‐mediated changes of EMT makers and
   migration ability (Figure [235]S14b‐d, Supporting Information).

   CXCR3 has also been found in tumor cells (Figure [236]S14e, Supporting
   Information), leading to the speculation that PPARγ may be activated
   following CXCL10‐CXCR3 stimulation. CXCR3 knockdown was found to
   significantly decrease the effect of CXCL10 on PPARγ (Figure [237]6l;
   Figure [238]S14f, Supporting Information). Additionally, CXCR3
   knockdown obviously reduced the translational activity of the PPAR
   luciferase report plasmid after CXCL10 stimulation (Figure [239]6m;
   Figure [240]S14g, Supporting Information). Finally, the PPARγ knockdown
   significantly reversed the stemness markers and the capacity of sphere
   formation induced by CXCL10 stimulation (Figure [241]S14h, i,
   Supporting Information). Overall, these results demonstrate that CXCL10
   can promote the mobility and stemness in PDAC through CXCR3/PPARγ axis.

2.7. Blockade of Tumor‐Associated Mast Cell‐Derived CXCL10 as a Strategy by
Reducing CXCR3^+Tregs Mediated Immune Escape

   To further investigate the effect of SCG on immune microenvironment, we
   examined tumor‐infiltrating immune cells and found a significant
   decrease in Tregs after SCG treatment by flow cytometry and
   immunofluorescence staining (Figure [242] 7a,b). It is widely accepted
   that Th1 cells preferentially express CXCR3 and CCR5, while CCR8 and
   CCR4 are highly expressed in Th2 and FoxP3^+Tregs. However, a specific
   subset of CXCR3^+Treg has been identified in cancers and known to
   promote tumor progression.^[ [243]^28 ^] This specific subtype may be
   attracted by CXCL10, which supports our hypothesis that SCG can
   influence TME by inhibiting the secretion of CXCL10 from TAMCs. We also
   found a higher proportion of CXCR3^+Tregs in tumor tissues compared to
   PBMCs, indicating that CXCR3 expression on Tregs plays a vital role in
   their chemotaxis (Figure [244]7c). The presence and percentage of
   CXCR3^+Tregs was further confirmed in human PDAC tissues through
   immunohistochemistry (Figure [245]7d). We found that MCs induced the
   migration of CXCR3^+ Tregs in a CXCL10‐dependent manner in chemotaxis
   assay (Figure [246]7e). Co‐culturing T cells with or without
   CXCR3^+/CXCR3^−Tregs isolated from peripheral PBMC revealed that
   CXCR3^+Tregs could also inhibit CD8^+T cell function (Figure [247]7f).
   Therefore, TAMCs can attract CXCR3^+Tregs into TME, and thereby induce
   a stronger immunosuppressive response in PDAC.

Figure 7.

   Figure 7
   [248]Open in a new tab

   CXCL10 recruits CXCR3^+ Tregs into microenvironment and SCG synergizes
   with gemcitabine plus αPD‐1. a) TSNE analysis for tumor‐infiltrating
   immune cells with or without SCG treatment and the proportions of these
   cells. KPC cell lines were orthotopically inoculated in C57BL/6 mice
   and received SCG treatment, tumor tissues were harvested at 20 days and
   analyzed by flow cytometry (n = 4). b) Dual‐fluorescence staining
   images and quantification of tumor‐infiltrating Treg in SCG treatment
   and control group. Foxp3, red; CD4, green; DAPI, blue. c) Flow
   cytometry analysis (left) and MFI quantitative analysis (right) of the
   CXCR3^+ level on Treg in paired PMBCs and PDAC tissues. d)
   Dual‐fluorescence staining of CXCR3^+ Treg in PDAC tissues (upper).
   CXCR3, red; FoxP3, green; DAPI, bule. The percentage of CXCR3^+Treg in
   total tumor infiltrated Treg were calculated in seven PDAC patients
   (lower). e) Quantification of Treg proportions in migrated CD4^+ cells
   recruited by mMCs‐conditioned medium with or without antibodies
   blockade in migration assay. Murine spleen was dissociated as single
   cells and added into upper chamber, indicated medium were then added
   into lower chamber for migration assay. n = 4 per group. f) Flow
   cytometry analysis for the level of exhausted markers in T cells when
   coculturing with CXCR3^+ or CXCR3^− Treg sorted by FACS. n = 3 per
   group. g) SCG was intraperitoneal injection at 75mg kg^−1 for three
   times per week. Gemcitabine (20mg kg^−1) plus αPD‐1 (100µg per mouse)
   were intraperitoneal injection once a week. 20 days after inoculation,
   luciferase activity was detected by IVIS imaging. Representative
   optical image (left) in KPC murine models and quantification (right) of
   luciferase optical density. n = 5 per group. n.s. P > 0.05, *P < 0.05,
   **P < 0.01, and ***P < 0.001, Data were displayed as mean ± SD.

   Chemo‐immunotherapy revolutionizes cancer treatment and holds promise
   for PDAC.^[ [249]^29 ^] As our preceding results, TAMCs‐derived CXCL10
   recruits CXCR3^+Treg into TME, resulting in increasing exhausted
   phenotypes in CD8^+T cells, notably PD‐1, in PDAC. This indicates that
   αPD‐1 may also restore cytotoxic T cells exhaustion induced by TAMCs.
   Additionally, stemness of cancer cells is a key factor in gemcitabine
   resistance, suggesting that blocking CXCL10 from TAMCs by SCG could
   also be a potential combination to enhance chemotherapeutic efficacy.
   Therefore, we hypothesized that SCG might provide a synergistic effect
   when combined with gemcitabine and αPD‐1.

   In orthotopic murine PDAC model, we found that SCG alone slightly
   inhibited tumor growth and the combination of SCG with
   gemcitabine/αPD‐1 was more effective than each agent alone
   (Figure [250]7g). These results indicate that SCG sensitize the
   therapeutic efficacy of gemcitabine plus αPD‐1 in PDAC and could be a
   promising adjuvant for chemo‐immunotherapy in PDAC.

3. Discussion

   Complex communication networks between tumor and immune cells play a
   critical role in tumor progression and drug resistance. This study
   demonstrates that a crosstalk between PDAC cells and TAMCs ultimately
   contributes to cancer metastasis and immuno‐chemotherapeutic
   resistance, as illustrated in Figure [251]S15 (Supporting Information).
   In light of these findings, we propose SCG as a promising therapeutic
   strategy by inhibiting MCs’ CXCL10 secretion, disrupting the crosstalk
   between MCs and tumor cells, and thereby alleviating tumor metastasis
   while enhancing sensitivity to immuno‐chemotherapy.

   CXCR2 is vital for the recruitment of MCs in physiological ^[ [252]^30
   ^] and pathological conditions.^[ [253]^31 ^] In this study, we have
   also demonstrated that CXCR2 plays a crucial role in recruiting MCs in
   PDAC, while CXCR4 is the main chemotaxis receptor for MC infiltration
   in gastric cancer.^[ [254]^16 ^] This highlights the diversity of mast
   cells in various cancers. Furthermore, TAMCs in gastric cancer rarely
   expressed Ki‐67, which is in stark contrast to PDAC, where the majority
   of TAMCs (>85%) were Ki‐67 positive in our study.

   Ferroptosis is a non‐apoptotic, iron‐dependent form of programmed cell
   death characterized by lipid peroxide accumulation. Our study
   demonstrated that upregulated GPX4 and SLC7A11 is consistent with
   decreased ferroptosis in TAMCs after tumor conditioned medium
   stimulation. However, ACLS4 was reversely upregulated after treatment
   with tumor‐conditioned medium, which was contradicted to ferroptosis
   resistance in TAMCs. The Hippo pathway, especially the YAP‐TEAD
   complex, has been reported to bind to the ACSL4 promoter.^[ [255]^32 ^]
   Our data also revealed a significant enrichment of the Hippo as the
   second enriched pathway in TAMCs, which may explain the aberrant high
   expression of ACSL4 in low ferroptosis level. Ferroptosis sensitivity
   is determined by the susceptibility to lipid peroxidation and the
   overall defense capacity of cells.^[ [256]^33 ^] TAMCs exhibited
   significantly high lipid ROS generation and elimination, while
   antioxidative process through GSH/GSSH had higher activity,
   counteracting the effect of ACSL4.

   However, it is worth to note that treatment with tumor‐conditioned
   medium restored RSL3‐induced ferroptosis in MCs, suggesting that the
   inhibition of ferroptosis in MCs is not solely mediated by System Xc‐
   and GPX4‐dependent mechanisms. It had been reported that purified GPX4
   alone cannot be inhibited by RSL3, indicating the involvement of other
   cellular components in this process.^[ [257]^34 ^] Tumor conditioned
   medium treatment could interrupt the interaction and rescue cells from
   ferroptosis. On the other hand, ferroptosis occurs after enough lipid
   peroxidation accumulation. Sebastian et.al reported that cells lacking
   GPX4 activity showed ferroptosis resistance if lipid ROS production is
   disrupted at the same time.^[ [258]^35 ^] Fe^2+ ion is crucial for
   lipid peroxide by Fenton reaction. Labile iron pool (LIP) can
   contribute to spontaneous oxidation and further drive ferroptosis.^[
   [259]^36 ^] We observed significant decrease of Fe^2+ in TAMCs, which
   could be other potential mechanism that contribute to GPX4‐independent
   ferroptosis. Thus, it is worth to further investigate the details of
   how PDAC inhibits ferroptosis in MCs and alleviates the accumulation of
   MCs by targeting potential mechanisms.

   MCs have transcriptional diversity depending on different surroundings.
   One classification for MCs is based on their produced proteases, by
   categorizing as tryptase only (MC[T]), chymase only (MC[C]), or a
   combination of both (MC[TC]).^[ [260]^37 ^] In this scenario, MCs could
   induce various biological function when releasing the proteases.
   However, we found an impaired degranulation capacity in TAMCs compared
   with MCs, highlighting chemokine/cytokine secretion is of great
   significance in TAMCs. Secreted histamine is the key molecular in
   anaphylactic reactions, and this could inhibit PDAC cells growth in
   vitro.^[ [261]^38 ^] In tumor microenvironment, the subtype of
   histamine receptors (HRs) differential expression and different
   intracellular signaling cause the diversity biological function. HR1‐2
   are both expressed endothelial cells, which participate in histamine
   mediated vasodilation and increase vascular permeability.^[ [262]^39 ^]
   H3R is expressed in the central nervous system. H4R has a stronger
   capacity in response to histamine, IgE, and compound 48/80 than that
   observed with H1R antagonists.^[ [263]^40 ^] In PDAC, TAMCs showed
   decreased secretion of histamine, so we paid attention to its cytokine
   secretion. In this study, we identified CXCL10 as crucial chemokine in
   TAMCs. CXCL10 is the ligand for CXCR3, which have three variants
   including CXCR3a, CXCR3b and CXCR3‐alt, with CXCR3a binding to CXCL9,
   CXCL10, and CXCL11, while CXCR3b preferably binds to CXCL4.^[ [264]^41
   ^] Our results showed that significant CXCL10 upregulation in TAMCs
   compared with MCs. There are still some controversies of CXCL10
   biological function in tumors. High CXCL10 expression in tumor tissue
   correlated with poor survival in PDAC, even though CXCL10 exhibited
   anti‐tumor effect by recruiting CXCR3 positive immune cells to remodel
   TME, including CXCR3^+CD4^+T cells and CXCR3^+CD8^+T cells.^[ [265]^42
   ^] In our study, we found notably CD8^+T cells increase after SCG
   treatment in vitro, contradictorily with decreasing TAMCs‐derived
   CXCL10 secretion. We suspected that this was caused by synchronous
   recruitment of CXCR3^+Tregs. CXCR3^+Treg is a strong immunosuppressive
   subtype, which could suppress T cells function by both
   contact‐dependent^[ [266]^28 ^] and independent^[ [267]^43 ^] way.
   After SCG treatment, decreased TAMCs‐derived CXCL10 recruited less
   CXCR3^+Treg into tumor, and this might be one possible reason that
   caused the increasing CD8^+T cells infiltration in our study. In
   addition, CXCR3^+Treg also significantly inhibit effective T cells
   proliferation,^[ [268]^44 ^] this might be another factor that less
   CXCR3^+Treg infiltration promote more CD8^+T cells proliferation and
   finally led to CD8^+T cells accumulation. Therefore, the balance
   between CXCR3^+CD8^+ cells and CXCR3^+Tregs decided the effect of
   CXCL10 on TME. Additionally, peripheral blood mononuclear cells of PDAC
   patients contained more Tregs, indicating that circulating Tregs could
   be preferentially recruited by CXCL10 compared to other cells.^[
   [269]^45 ^]

   The issues of recurrence and delayed distant metastasis following
   radical tumor resection cannot be overlooked in PDAC. Our findings
   indicated that TAMCs increased the recurrence rate in PDAC and led to
   shorter OS and RFS. We found that TAMCs could promote PDAC cells EMT
   transition by targeting tumoral CXCR3, which was responsible for tumor
   metastasis. Similar observation has been reported in thyroid cancer.^[
   [270]^46 ^] CXCL10 activated tumoral CXCR3 and its downstream in PDAC
   cells, finally promoted ZEB1 nuclear translocation. ZEB1 mediated EMT
   transition is closely associated with the acquisition of stemness in
   tumor cells, and implicated in resistance to gemcitabine‐based
   chemotherapy.^[ [271]^47 ^] Currently, gemcitabine is regarded as the
   first‐line treatment for advanced PDAC. One clinical trial in
   metastatic PDAC observed modest increases in OS with
   αPD‐1/GEM/Nab‐paclitaxel compared to chemotherapy alone.^[ [272]^29 ^]
   Moreover, in locally advanced PDAC, αPD‐1 antibody combined with
   chemotherapy significantly prolonged OS and PFS.^[ [273]^48 ^] Based on
   our findings, we conducted a preclinical trial using SCG as an adjuvant
   agent for immuno‐chemotherapy against PDAC, which yielded satisfactory
   outcomes.

   In summary, our study clarifies the crosstalk between PDAC and TAMCs,
   which leads to the abnormal accumulation of TAMCs in tumor lesions via
   CXCR2‐mediated chemotaxis. Tumor‐derived exosome miR‐188‐5p promotes
   CXCL10 expression via PTEN/AKT/ERG signaling in TAMCs. Additionally,
   TAMCs‐derived CXCL10 reversely enhanced the mobility and stemness of
   PDAC cells via CXCR3/PPARγ axis, and recruits CXCR3+ Tregs into TME,
   thereby promoting tumor progression and immune escape. In pre‐clinical
   trials, SCG is confirmed as an effective agent in decreasing recurrence
   and dissemination after radical resection of PDAC, and significantly
   enhances sensitivity to current immuno‐chemotherapy. Therefore, SCG
   shows promise as a therapeutic adjuvant in the current strategy for
   PDAC.

   However, this study had several limitations. First, we reported the
   ferroptosis protection of tumor conditioned medium for MCs, however,
   the underlying mechanism required further exploring. Second, the
   detailed mechanism of how CXCR3^+Tregs inhibits CD8^+T cell function in
   PDAC remains to be investigated. Third, even though negative
   correlation between TAMCs and survival has been reported in several
   studies, the clinical/pathological correlation with tumor‐infiltrating
   MCs level mains to be validated in another validation cohort.

   In summary, this study identified the crosstalk between TAMCs and PDAC
   cells, and demonstrated an innovative mechanism that promotes PDAC
   progression and immunosuppression. CXCL10 blockade with SCG sensitizes
   PDAC to the current immuno‐chemotherapy, which reveals its potential to
   be a valuable adjuvant for PDAC patients.

4. Experimental Section

Cell Cultures

   PDAC Cell Lines: Human PDAC cell lines (PANC‐1 and CFPAC‐1) were
   purchased from the National Collection of Authenticated Cell Cultures
   and authenticated using STR profiles. Mouse‐derived PDAC cell lines,
   Panc02 and KPC cells (LSL‐Kras^G12D/+; Trp53^R172H/+; Pdx‐1^Cre/+),
   were generously provided by Johns Hopkins Hospital. PANC‐1 and KPC
   cells were cultured in high glucose DMEM supplemented with 10% FBS;
   CFPAC‐1 cells were cultured in IMDM supplemented with 10% FBS; Panc02
   cells were cultured in RPMI‐1640 supplemented with 10% FBS. All cell
   lines were maintained in a humidified atmosphere with 5% CO[2] at 37
   °C. Mycoplasma was routinely tested during cells culturing.

Primary Immune Cells

   For murine mast cells, progenitor mMCs were isolated and expanded from
   C57BL/6 femur bone marrow. Fresh bone marrow was incubated with RBC
   lysis buffer to lyse red blood cell, then filtered through 70µm sieve
   to remove cell lumps. The residual cells were maintained and amplified
   in RPMI‐1640 supplemented with IL‐3 (10ng mL^−1), SCF (10ng mL^−1) and
   10% (v/v) FBS for 4 weeks. The purity was identified by flow cytometry
   and immunohistochemistry staining with indicated markers before further
   study.

   Human CD34^+ progenitor cells were enriched using an isolation kit from
   PBMC according to manufacturer's instructions, and differentiated into
   hMCs as previously reported.^[ [274]^49 ^] Human mast cells were
   stained with c‐kit and FcεR1a antibodies and identified the purity by
   flow cytometry.

   Human T cells were isolated and amplified as previously reported.^[
   [275]^50 ^] Briefly, PBMCs were isolated from peripheral blood and
   amplified in RPMI‐1640 containing anti‐CD3 (1µg mL^−1), anti‐CD28
   (1µg mL^−1), and rhIL‐2 (10ng mL^−1) and 10% (v/v) FBS.

   Human Tregs from PBMCs were labeled by CD3^+CD4^+CD25^+CD127^low and
   further sorted into CXCR3+ and CXCR3^low for CXCR3^posTreg or
   CXCR3^negTreg, respectively.

PDAC Patient‐Derived Organoids

   Three PDAC patient‐derived organoids were established for co‐culturing
   assay. In brief, surgical resected PDAC specimens from Zhongshan
   Hospital, Fudan University were digested by the Tumor Dissociation Kit,
   and embedded in Matrigel as dome. The domes were overlaid with organoid
   culture medium provided by Hangzhou Chexmed technology co., LTD. The
   medium was changed every 4 days. After the indicated treatment, the
   organoids were harvested using Organoid Harvest Solution and embedded
   in paraffin for organoid slice preparation.

DNA/RNA and Protein Experiments

   Design and Construction for siRNA/shRNA, Luciferase‐Reported Plasmid,
   or miRNA Analogue/ Antagonist: The siRNA targeting human SCF, PPARγ,
   and CXCR3 were purchased from Tsingke Biotechnology. Murine Creb1, Erg,
   Sp1, Max, and Nfyb si‐RNA were designed and synthesized by Yuanmin
   Biotech. miR‐188‐5p inhibitor and miR‐188‐5p mimic and negative control
   were purchased from RiboBio Co., Ltd. Lentivirus encoding murine
   cxcl10‐shRNA (CXCL10^sh1/sh2) or negative control (CXCL10^NC) and
   luciferase reported plasmid were designed and synthesized by GeneChem.
   Lentivirus was prepared by transfecting plasmid DNA and packaging
   plasmids in 293T cells, and virus supernatant was harvested after 48 h.
   The targeting sequences can be found in Table [276]S3 (Supporting
   Information).

Cytoplasm and Nuclear Separation, Western Blot, and Co‐Immunoprecipitation
(co‐IP)

   Cells suspensions lysed by RIPA for total protein analysis. Cytoplasmic
   and nuclear proteins were obtained using a cytoplasmic and nuclear
   separation kit. The expression of proteins in the cell lysate was
   assessed by western blot assays. Briefly, the proteins were transferred
   onto a PVDF membrane, and blocked with a 5% milk solution, and then
   incubated with indicated primary and secondary antibodies. The bands
   were visualized using an ECL imaging system.

   For co‐IP, primary antibodies or IgG as negative control were incubated
   with Protein A/G Magnetic Beads for 30 minutes at room temperature and
   washed five times to remove unconjugated antibody. Then, cell lysates
   were incubated with conjugated antibodies for an additional 2 h at room
   temperature. The beads were then washed five times with washing buffer.
   To elute proteins for western blots, loading buffer were added into the
   beads, and the mixture was heated to 99 °C for 10 min to dissociate
   targeting proteins from beads. Finally, the proteins were used for
   western blot analysis.

Enzyme‐Linked Immunosorbent Assay (ELISA)

   The secretion of CXCL10 from TAMCs was evaluated using the Mouse CXCL10
   Duoset ELISA kit, following the manufacturer's instructions. Mast cells
   were pretreated with tumor‐conditioned medium or DMEM as a control for
   24 h. Afterwards, the cells were washed by PBS buffer to remove the
   initial medium, subsequently sed into fresh medium at concentration of
   5×10^5 mL^−1 for 48 h, and finally indicated supernatants were
   collected for the ELISA analysis according to manufacturer's
   instruction. To eliminate evaporation effect on supernatant, the CXCL10
   concentration in indicated supernatants was normalized to the final
   mast cells density as CXCL10^normalized = CXCL10^measurement × [cell
   density/5×10^5]. The optical density of each group was measured at 450
   nm using a microplate reader (Molecular Devices, China), and the
   concentration of CXCL10 was calculated using a standard curve.

Immunohistochemistry and Immunofluorescence

   Specimens were fixed in 4% paraformaldehyde, embedded in paraffin and
   sliced to 3µm sections. After deparaffinization, the slides for
   immunohistochemistry underwent antigen retrieval by microwaving, and
   non‐specific sites were blocked with 5% goat serum.

   For immunohistochemistry, endogenous peroxidase was blocked using a 3%
   hydrogen peroxide solution. Slides were then incubated overnight at 4
   °C with appropriate diluted primary antibodies, followed by incubation
   with a 1:500 diluted HRP‐conjugated secondary antibodies for an
   additional 2 h. The slides were stained by the BAD reagent and
   counterstained with hematoxylin. Staining intensity and density of
   positive cells were evaluated by two pathologists. The representative
   slides were captured by microscope (Olympus, Japan) or ScanScope XT
   scanner (Aperio Technologies). The density of TAMCs in PDAC were
   calculated by TAMCs (numbers) /stroma area (mm3) for each case. The
   expression of proteins for each sample was defined as the intensity of
   positive staining (brown).

   For immunofluorescence, the slides were incubated overnight at 4 °C
   with appropriate diluted primary antibodies, followed by incubation
   with a fluorescence‐labeled secondary antibody for an additional 2 h.
   After washing out the secondary antibody, the slides were stained by
   DAPI. The representative images of indicated slices were captured by
   fluorescence microscope (Olympus, Japan). The evaluation methods of
   immunofluorescence staining were similar to immunohistochemistry.

RT‐qPCR Assay and Chromatin Immunoprecipitation (ChIP)

   The cells were lysed using RNAiso reagent, and total RNAs were
   extracted according to manufacturer's instructions. Then RNA was then
   reverse transcribed into cDNA using PrimeScript RT reagent kit. The
   indicated cDNA, primers, and SYBG Green Mix were mixed and analyzed
   using the QuantStudio™ 5 RT‐qPCR System (Applied Biosystems, USA). The
   expression of the target genes was normalized to GAPDH and quantified
   as ΔΔCt. The primer sequences used in this study are listed in Table
   [277]S4 (Supporting Information).

   The ChIP assay was performed using the Chromatin Immunoprecipitation
   Kit, following the manufacturer's instructions. Briefly, treated mast
   cells were cross‐linked by 1.5% formaldehyde and terminated by glycine.
   Nuclei were extracted from the collected cells and digested with
   micrococcal nuclease at 37 °C for 20 min. The resulting products were
   then incubated overnight at 4 °C with anti‐ERG antibody, IgG isotype as
   a control, and Histone H3 as a positive control. The lysates were
   subsequently incubated with Protein G beads at 4 °C for 2 h. The
   immunoprecipitation products were separated using a magnetic stand, and
   the DNA was extracted. The DNA was analyzed by qPCR, and the binding
   sites were evaluated based on the enrichment level. The primer
   sequences for the ChIP PCR assay can be found in Table [278]S5
   (Supporting Information).

Clinical Data, Tumor Tissues, and Follow‐Ups

   Tissue Microarray and Follow‐Ups: Tissue microarrays (TMA) were created
   using archival formalin‐fixed paraffin‐embedded tissue blocks obtained
   from surgically resected primary PDAC. This was done using a manual
   Tissue Puncher/Arrayer (Beecher Instruments, Silver Spring, MD) as
   previously described.^[ [279]^51 ^] We arrayed 102 cores with 1.4 mm
   diameters (containing 52 PDAC and 50 adjacent normal tissues) on
   recipient blocks at Johns Hopkins Hospital. Similarly, 172 cores
   (containing 90 PDAC and 82 adjacent normal tissues) from Zhongshan
   hospital were arrayed as TMAs for further analysis. The OS was
   calculated from the date of surgery to the time of death for any reason
   or to the time of censorship at the last follow‐up. The RFS was defined
   as the period between the date of surgery and the date of recurrence of
   disease or death, whichever occurred first or was censored at the last
   follow‐up. Informed consent was obtained from all patients, and this
   protocol was approved by the Ethics Committee of Zhongshan Hospital,
   Fudan University (No. B2023‐129).

Infiltrating Immune Cell Analysis in PDAC

   PDAC and its adjacent normal tissues were dissociated to single cell by
   the Tumor Dissociation Kit according to manufacturer's instructions.
   The single cells were collected and stained for surface markers at 4 °C
   for 30 min, and then fixed with IC Fixation Buffer. After washing three
   times by PBS buffer, the cell suspensions were permeabilized and
   stained for intracellular markers using Permeabilization Buffer.
   Finally, the data were acquired by the BD FACS Arial Flow Cytometer (BD
   Biosciences, USA), and analyzed using Flowjo V10 (BD Biosciences, USA).

RNA Sequencing and Bioinformatic Analyses

   Bulk RNA‐Sequencing: RNA extraction was completed by the Qiagen AllPrep
   DNA/RNA Mini kit (Qiagen, Hilden, Germany) according to manufacturer
   specifications. RNA‐sequencing was performed at the Johns Hopkins
   Experimental and Computational Genomics Core (ECGC). All RNA samples
   passed the quality control procedure using the Bioanalyzer Total RNA
   Pico kit (Agilent, Santa Clara CA). The libraries for RNA sequencing
   were constructed by TruSeq Stranded Total RNA with Ribo‐Zero (Illumina,
   San Diego CA), which was monitored using the Bioanalyzer High
   Sensitivity kit (Agilent, Santa Clara CA). The libraries were sequenced
   on an Illumina Novaseq 6000 (Illumina, San Diego CA) instrument using
   150bp paired‐end dual indexed reads to a target depth of ≈100000 000
   reads per sample with Q30 > 92%. Reads were aligned to the reference
   genome utilizing illumine package bcl2fastq. Differential expression
   analysis and statistical testing were performed using DESeq2 software.
   The biological functions were investigated via KEGG pathway analysis
   and Gene Set Enrichment Analysis (GSEA) comparing the expression of
   different genes (DGE).

TCGA Analysis

   Pancreatic cancer bulk RNA‐seq data were downloaded from the TCGA
   database, and the expression of genes were cleanout and normalized for
   subsequent analysis.

In Vitro Biological Function Assay

   Mast Cell and Treg Migration Assay: To perform the mast cell migration
   assay, we seeded the upper chamber of 3µm uncoated 12‐well Boyden
   chamber (BD Biosciences, USA) with 2 × 10^5 mMCs in 400µL of completed
   DMEM. The lower chamber was filled with 750µL of completed DMEM as a
   control, as well as the indicated tumor‐conditioned medium. After 12
   hours, the filters were removed and the number of migrated mast cells
   was counted. All migration assays were done in triplicate.

CCK‐8, Colony Formation, EDU, and Apoptosis Assay

   In the CCK‐8 assay, PDAC cell lines were seeded in 96‐well plates and
   cultured in 100µL of appropriate medium with or without 100ng mL^−1
   rhCXCL10. After that, 10µL CCK‐8 solution was added to each well, and
   the plates were incubated for one hour. The absorbance at 450nm was
   then measured at indicated time points after cell inoculation. The data
   represents the mean ± standard deviation (SD) from five independent
   experiments.

   For the colony formation assay, the cells were seeded in 6‐well plates
   and cultured in medium with or without 100ng mL^−1 rhCXCL10. Depending
   on the colony growth, the cells were cultured for one to two weeks.
   After that, the cells were fixed with 4% paraformaldehyde and stained
   with crystal violet to count the number of colonies.

   In the EDU assay, the cells were seeded in six‐well plates and cultured
   in an appropriate medium with or without 100ng/mL rhCXCL10 for 48 h.
   The proliferated cells were then detected using EDU assay according to
   the manufacturer's instructions and visualized by immunofluorescence
   microscope (Olympus, Japan).

   For the apoptosis assay, the cells were seeded in six‐well plates and
   cultured in an appropriate medium with or without 100ng mL^−1 rhCXCL10
   for 48 h. After that, the cells were digested into single cells and
   stained using an apoptosis kit following the manufacturer's
   instructions. The early and later apoptosis rate were analyzed by a
   flow cytometer Arial II (BD Biosciences, USA).

Cell Stemness and Mobility Evaluation

   After indicated treatment, tumor cells were resuspended in DMEM/F12
   containing 20ng mL^−1 rhEGF, 1×B27, and 100ng mL^−1 bFGF, and 200 cells
   per well were sed into 96‐well ultra‐low attachment plate (Corning,
   USA) for 14 days culturing. The number of sphere clone that diameter is
   over 75µm were calculated in each well, and the sphere formulation
   efficacy was calculated by sphere number/200.

   For migration assay, transwell assays were perform in an 8‐µm size.
   Tumor cells were seed in the uncoated upper chamber with 300µL FBS‐free
   medium, while 750µL of completed DEME medium was added to the lower
   chamber. After 24 h of incubation, the cells were fixed with 4%
   paraformaldehyde for 20 min and stained with crystal violet. The cells
   in the upper chamber were then removed. Invading cells were
   photographed using a microscope (Olympus, Japan) in three random views,
   and the average number of cells was calculated. The experiment was
   repeated three times for biological replication.

   For invasion assay, 8µm transwell chamber were coated by matrix gel
   with pre‐cold pipette tips and kept in cell incubator for
   solidification. Then indicated tumor cell suspensions without FBS‐free
   medium were added to the upper chamber and completed medium was added
   into the lower chamber. After incubation for 48 hours, the chambers
   were harvested for invasion ability evaluation similarly to migration
   assay.

Transfection and Luciferase Reporter Assay

   CFPAC‐1 and PANC‐1 cells were seed in 6‐well plates 24 h before
   transfection. They were transfected with PPARγ promoter luciferase
   reporter and Renilla‐luciferase plasmid (as a control) using
   lipofectamine 3000. Similarly, mast cells were co‐transfected with
   wt‐CXCL10/mut‐CXCL10 promoter luciferase reporter plasmid and ERG
   overexpression plasmid. Two days after transfection, PDAC cells were
   stimulated by CXCL10 for 48 h and lysed for luciferase activity assay
   according to the manufacturer's instructions. Mast cells were also
   lysed for dual‐luciferase reporter assay.

Mast Cell Proliferation and Cell Cycle Assay

   To perform the mast cell proliferation assay, mast cells were
   pre‐stained with CFSE in a concentration of 2.5uM for 10min at 37 °C
   and washed by PBS for three times. Then, the cells were seeded in a
   10cm dish, and cultured in DMEM or tumor‐conditioned medium
   supplemented with 10ng mL^−1 SCF, 10ng mL^−1 IL‐3, and 10% (v/v) FBS.
   The cells were cultured for 48 h and subsequently collected for
   analysis using the Arial II flow cytometer.

   For the cell cycle assay, mast cells were seeded in a 10cm dish and
   cultured for 48 h in DMEM or tumor‐conditioned medium supplemented with
   10ng mL^−1 SCF, 10ng mL^−1 IL‐3, and 10% (v/v) FBS. The cells were then
   collected and fixed with ethanol. After fixation, the cells were
   digested by RNase (100µg mL^−1) and washed with PBS. Finally, the cells
   were stained by PI (50µg mL^−1) for 30min and analyzed by the Arial II
   flow cytometer.

Cell Death and Ferroptosis Analysis

   Mast cells were cultured in tumor‐conditioned medium or DMEM as a
   control for 24 h. Afterward, they were stimulated with the indicated
   concentration of Erastin for 24 h. The cells were then collected and
   incubated with C11 BODIPY (1mg mL^−1) for 20 min to analyze lipid ROS,
   or incubated with PI (5µg mL^−1) / DAPI (10µg mL^−1) for 5 min to
   analyzed cell death, or incubated with DCFH‐DA (10uM) for 30 min to
   analyze total ROS.

   To evaluate ferroptosis, cells were cultured in tumor‐conditioned
   medium or DMEM as a control for 24 h. They were then stimulated with
   the indicated concentration of Erastin for 6 h. After collection, the
   cells were incubated with FerroOrange for 30 min, and analyzed using a
   flow cytometer or fluorescence microscope.

Exosome Identification and Intake Assay

   KPC‐derived exosomes were collected from 50 mL KPC conditioned medium
   and purified through ultra‐centrifugation. In brief, the medium was
   centrifuged at 2000g for 10 min and then at 10 000g for 10 min to
   remove cells and debris. The supernatant was further centrifuged at
   100 000g for 90 min to precipitate and collect the exosomes for
   observation under a transmission electron microscope (Hitachi, Japan)
   and analysis through western blot.

   The purified exosomes were labeled using PKH26 labeling kit. Briefly,
   the exosomes were mixed with PKH26, which was diluted in Reagent C, and
   washed to remove any unconjugated dye. The labeled exosomes were
   subsequently resuspended for the mast cells intake assay and observed
   using a fluorescence microscope.

Pre‐Clinical Murine Model Experiments

   Establishment of Subcutaneous/Orthotopic Murine PDAC Model: Female
   C57BL/6 mice, six weeks old, were purchased from JieSiJie Laboratory
   Animal Co, Ltd and kept in a specific pathogen‐free (SPF) atmosphere.
   To establish a subcutaneous tumor model, 100µL of KPC or Panc02 cell
   suspension (2 × 10^7 mL^−1) mixed with Matrigel was injected into the
   back subcutaneously. For the orthotopic tumor model, mice were
   anesthetized with 2% (m/v) pentobarbital sodium solution and 50µL KPC
   or Panc02 cell suspension (2 × 10^7 mL^−1) mixed with Matrigel was
   injected into the pancreas tail. The tumor volume was calculated by
   1/2× (long diameter) × (short diameter)2.

Mast Cells Adaptive Transfer Experiment

   Mast cells were isolated from C57BL/6 mice and maintained in vitro
   until purified, as mentioned above. Lentivirus was used to construct
   CXCL10 knockdown mast cells, and transfection efficiency was determined
   by qRT‐PCR analysis. For the mast cell transplantation assay, sh‐CXCL10
   knockdown (CXCL10^sh1/sh2) or negative control (CXCL10^NC) MCs (10^7
   mast cells per mouse) were injected into the tail vein of tumor burden
   mice at 3, 6, and 9 days after pancreatic cancer inoculation.

Monotherapy or Combination Therapeutics

   The indicated treatments were started three days after tumor
   inoculation. SCG was dissolved in PBS buffer and administrated
   intraperitoneally at dose of 75mg kg^−1 three times a week. Gemcitabine
   (hydrochloride formulation) was dissolved in PBS buffer and
   administrated intraperitoneally at dose of 25mg kg^−1 once a week.
   αPD‐1 antibody was dissolved in PBS buffer and administrated
   intraperitoneally at dose of 3mg kg^−1 once a week.

Optical Imaging for Tumor Burden Evaluation and Survival Analysis

   Optical imaging was used to detect tumor cells transfected with
   luciferase using an in vivo imaging system (CleVue™ Vieworks, Korean).
   D‐luciferin sodium salt was intraperitoneally injected at a dose of
   150mg kg^−1 per mouse, and luciferase activity was evaluated within
   30min after injection.

   For survival analysis, PDAC orthotopically inoculated mice received the
   indicated treatment and were monitored until they succumbed
   spontaneously or were humanely terminated via euthanasia upon the
   appearance of severe physical symptoms such as severe cachexia,
   ascites, hypothermia, and tumor size exceeding 2 cm.

Histological Pathology

   At the end of animal experiment, the mice were sacrificed
   humanistically. Tumor tissues were harvested and fixed in 4%
   paraformaldehyde. Tissues were routinely embedded in paraffin and
   sliced into sections for H&E or immunohistochemistry staining. The
   detailed manufacturer's information of the reagents above is listed in
   Table [280]S6 (Supporting Information).

Statistical Analysis

   All statistical analyses were performed by SPSS 22.0 and visualized by
   GraphPad Prism 9 software (GraphPad, San Diego, CA, USA). After
   Kolmogorov‐Smirnov normality test, two‐tailed Student's t‐test or
   one‐way ANOVA analysis was used for statistics. Post hoc LSD or Tamhane
   method were further used in subgroup analysis for one‐way ANOVA
   analysis. Non‐parametric test was applied for non‐normally
   distribution. All data were presented as mean with standard deviation.
   Kaplan‐Meier curves with log‐rank tests were used to analyze survival,
   and the univariate and multivariate Cox proportional hazards regression
   models were applied to analyze the independent prognostic indicators
   for OS and RFS. All the in vitro assays were repeated at least three
   times with biological replicates. The number of independent biological
   replicates was represented by n value in the figure legends or dots in
   the figures. Statistical significance was considered as P values <0.05.

Ethics Approval

   This study involves samples from human participants approved by the
   Ethics Committee of Zhongshan hospital, Fudan University (B2023‐129).
   Participants gave informed consent to participate in the study prior to
   enrolment and the consent obtained directly from patients. All animal
   experiments were approved by Animal Ethics Committee of Zhongshan
   Hospital, Fudan University (Y2022‐702).

Conflict of Interest

   The authors declare no conflict of interest.

Author Contributions

   H.Y., Q.C., and S.G. contributed equally as co‐first authors in this
   article. H.Y., Q.C., S.G., and N.P. performed conceptualization, data
   curation, and wrote the draft, formal analysis, and methodology. S.S.,
   Y.X., J.R.H., T.H., W.G., J.W., L.Z., H.X., C.S., J.H., W.W., Y.J., and
   M.G.G. performed formal analysis and methodology. N.P., S.G., W.W., and
   L.L. performed funding acquisition. N.P., J.Y., W.W., W.L., and L.L.
   performed supervision and reviewed the draft. All authors have read and
   approved the article.

Supporting information

   Supporting Information
   [281]ADVS-12-2417724-s001.docx^ (13.9MB, docx)

Acknowledgements