胰腺癌的serpin–髓系细胞轴与局部免疫抑制

Fig. 1 | Perturb-map identifies tumour-derived extracellular factors controlling PDAC growth and TME architecture. a, Schematic of the Perturb- map approach. KPC cells were transduced with a PC/CRISPR library, enabling spatially resolved analysis of clonal fitness and tumour–microenvironment interactions in vivo. gRNA, guide RNA. b, Representative images and digital reconstructions of orthotopic KPC tumours at days 7, 14 and 21 after transplantation. Colours indicate gene knockouts (35 Pro-Codes, 7 epitope tags (E)). Tumours from n = 5 (day 7), n = 7 (day 14) and n = 8 (day 21) mice were analysed; two images acquired per mouse. Individual mice represent biological replicates. Scale bars, 250 μm. c, Representative spatial heterogeneity maps of tumours at days 7 and 21. Colours indicate the percentage of non-matching neighbouring tumour cells (yellow, higher heterogeneity; black, lower heterogeneity). White boxes denote focal regions identified by mean-shift clustering. d, Representative Pro-Code (PC) composition in focal regions shown in c. Each colour represents a distinct knockout population. Quantification is shown in Extended Data Fig. 4c. e, log2-transformed fold change of normalized (norm.) fitness scores (in vivo/in vitro) relative to the F8-knockout (KO) control. Hatched cells indicate adjusted P > 0.05. Two-sided Mann–Whitney tests with false discovery rate (FDR) correction for multiple comparisons. NS, not significant. f, Neighbourhood enrichment analysis of immune cells within 25 μm of knockout tumour cells at days 7, 14 and 21. Heat map colours indicate interaction significance relative to a permuted null distribution. Red and blue annotation (left) indicates relative fitness enrichment and depletion, respectively, compared with control. g, Experimental design for analysis in Rag2−/− mice. KPCPC/CRISPR cells were orthotopically transplanted and tumours were collected 14 days later. One section from n = 8 mice was analysed. h, Abundance of knockout populations across Rag2−/− tumours. Colours below the axis show relative depletion or enrichment in immunocompetent mice at day 14. Serpinb2 and Serpine1 are highlighted in red. Statistical evaluation is shown in Extended Data Fig. 4f. Histology slides in a,g adapted from Servier Medical Art, microbiology and cell culture image kit, histology slide illustration, CC BY 4.0 (https://smart.servier. com/); mouse outlines in a,g adapted from ref. 47 (copyright © 2026 MyJoVE Corporation).

Fig. 2 | Loss of Serpinb2 and Serpine1 promotes T cell-mediated anti- tumour immunity. a, Differential expression of SERPINB2 (top) and SERPINE1 (bottom) in human tumours versus normal tissues across cancers. Pancreatic adenocarcinoma cohorts are labelled in red; non-significant comparisons are represented in grey. TCGA and GTEx data. ACC, adrenocortical carcinoma; BLCA, bladder urothelial carcinoma; BRCA, breast invasive carcinoma; CESC, cervical squamous cell carcinoma and endocervical adenocarcinoma; CHOL, cholangiocarcinoma; COAD, colon adenocarcinoma; DLBC, diffuse large B cell lymphoma; ESCA, oesophageal carcinoma; FC, fold change; GBM, glioblastoma multiforme; HNSC, head and neck squamous cell carcinoma; KICH, kidney chromophobe; KIRC, kidney renal clear cell carcinoma; KIRP, kidney renal papillary cell carcinoma; LAML, acute myeloid leukaemia; LGG, low-grade glioma; LIHC, liver hepatocellular carcinoma; LUAD, lung adenocarcinoma; LUSC, lung squamous cell carcinoma; OV, ovarian serous cystadenocarcinoma; PAAD, pancreatic adenocarcinoma; PCPG, pheochromocytoma and paraganglioma; PRAD, prostate adenocarcinoma; READ, rectum adenocarcinoma; SARC, sarcoma; SKCM, skin cutaneous melanoma; STAD, stomach adenocarcinoma; TGCT, testicular germ cell tumour; THCA, thyroid carcinoma; THYM, thymoma; UCEC, uterine corpus endometrial carcinoma; UCS, uterine carcinosarcoma. b, Meta-analysis of SERPINB2 (left) or SERPINE1 (right) expression and overall survival (data from refs. 48–52). Squares indicate hazard ratios (HRs; size proportional to sample size), lines represent 95% confidence intervals (CIs) and diamonds show pooled random-effects estimates (n = 1,176 patients from 8 PDAC cohorts). c, Experimental design for in vivo validation of Serpinb2 and Serpine1 knockout. Control cells carried an F8-targeting single guide RNA (sgRNA). d, Tumour mass of control, Serpinb2-KO, and Serpine1-KO orthotopic tumours (n = 7 mice per group). e, Intratumoural CD8+ T cells quantified by flow cytometry (control n = 5, Serpinb2-KO n = 6, Serpine1-KO n = 4). f, Relative abundance of T cell subsets identified by scRNA-seq of control, Serpinb2-KO and Serpine1-KO tumours. Teff, effector T cells; Tex, exhausted T cells; Tn, naive T cells; Treg, regulatory T cells. g, Representative pseudocoloured immunohistochemistry of control, Serpinb2-KO and Serpine1-KO tumours stained for CK19, CD8a and GZMB. Scale bars, 125 µm. h,i, Quantification of CD8+ (h) and GZMB+CD8+ (i) cells in tumours from g (n = 6 mice per group). j, Experimental design for analysis of serpin-knockout tumours in Rag2−/− and NSG mice. k, Tumour mass of control, Serpinb2-KO and Serpine1-KO tumours from NSG mice (n = 5 mice per group). l, Kaplan–Meier survival of mice bearing control, Serpinb2-KO or Serpine1-KO tumours treated with IgG (n = 5 per group) or anti-PD-1 (control n = 6, Serpinb2-KO n = 6, Serpine1-KO n = 7). m, Experimental design for evaluation of PAI-039 and anti-PD-1 combination therapy. n, Kaplan–Meier survival analysis of mice bearing KC tumours treated with vehicle (n = 6), PAI-039 (n = 6), anti-PD-1 (n = 6) or PAI-039 plus anti-PD-1 (n = 7). d,e,h,i,k, Two-sided unpaired Welch’s t-tests. Dots represent individual mice; bars show mean and s.d. l,n, Two-sided log-rank (Mantel–Cox) tests. Human body outlines in a,b adapted from Servier Medical Art, people image kit, human body outline illustration, CC BY 4.0 (https://smart.servier. com/); mouse outlines in c,j,m adapted from ref. 47 (copyright © 2026 MyJoVE Corporation).

PAI1 and PAI2 inhibit plasmin generation and thereby prevent fibrin degradation. Consistent with this role, fibrin(ogen)-dense regions were readily detected in orthotopic PDAC tumours (Fig. 3k). Both Serpinb2-KO and Serpine1-KO tumours showed reduced fibrin(ogen) deposition, indicating that each gene independently promotes fibrin accumulation in PDAC (Fig. 3k,l). These data indicate that cancer-produced Serpine1 and Serpinb2 remodel the ECM by promoting collagen and fibrin deposition. Fig. 3 | Serpinb2 and Serpine1 loss alters tumour transcriptional programmes and ECM deposition. a, Schematic overview of the Perturb-Map Multi-modal workflow. Histology slide adapted from Servier Medical Art, microbiology and cell culture image kit, histology slide illustration, CC BY 4.0 (https://smart. servier.com/). b, Representative tumour region of KPCPC/CRISPR tumours analysed by Xenium ST and tissue-based cyclic immunofluorescence (t-CyCIF). Left, Xenium-derived cell-type annotations; right, Pro-Code-defined tumour cell knockouts. c, Heat map of pathway activity derived from per-cell signature scores in tumour cells assigned to knockouts. Control-knockout (F8-KO) differences are shown as pathway activity scores (red, higher in knockout; blue, lower in knockout). Significance was defined at q < 0.01 after Benjamini– Hochberg correction. EMT, epithelial–mesenchymal transition; MHC I, major histocompatibility complex class I. d, Volcano plots showing differential gene expression in tumour epithelial cells from Serpinb2-KO (top) and Serpine1-KO (bottom) tumours compared with control (F8-KO). Differential expression was assessed using a two-sided Wilcoxon rank-sum test with Benjamini–Hochberg correction for multiple testing. Genes were considered significant at adjusted P ≤ 0.05 and are coloured by pathway membership as shown in c. e,f, Imaging- based quantification of PDPN+ (e) and αSMA+PDPN+ (f) cells, expressed as percentage per tumour (control n = 6, Serpinb2-KO n = 6, Serpine1-KO n = 6). Representative images are shown in Extended Data Fig. 7a. g,h, scRNA-seq analysis of stromal compartments showing CAF clusters (g) and relative abundance of CAF subsets (h). apCAFs, antigen-presenting CAFs; iCAFs, inflammatory CAFs; myCAFs, myofibroblastic CAFs; UMAP, uniform manifold approximation and projection. i, Representative haematoxylin and eosin (H&E) and Masson’s trichrome staining of control, Serpinb2-KO and Serpine1-KO tumours. Scale bars, 50 µm. j, Quantification of trichrome-positive collagen area (control n = 4, Serpinb2-KO n = 4, Serpine1-KO n = 4). k, Representative H&E and fibrin(ogen) immunohistochemistry of control, Serpinb2-KO, and Serpine1- KO tumours. Scale bars, 50 µm. l, Quantification of fibrin(ogen)-positive area in tumours isolated from control (n = 4), Serpinb2-KO (n = 4) and Serpine1-KO (n = 4). e,f,j,l Two-tailed, unpaired Welch’s t-test. Dots represent individual mice; bars indicate mean + s.d.

Supplementary Table 2). Although SERPINB2+ and SERPINE1+ tumour cells also exhibited unique transcriptional signatures, both states con- verged on pathways linked to ECM remodelling and immune signalling, processes associated with immunosuppression in cancer (Fig. 5c). These programmes mirrored the pathway alterations observed in KPC cancer cells upon Serpinb2 and Serpine1 deletion in our spatial tran- scriptomics analysis of mouse PDAC, including TGFβ signalling, ECM remodelling and myeloid chemotactic pathways (Fig. 3c), supporting Fig. 4 | Serpinb2 and Serpine1 promote fibrin-dependent macrophage retention and immunosuppressive polarization. a, Intratumoural macrophages quantified by flow cytometry in control, Serpinb2-KO, and Serpine1-KO tumours. Same cohort as Fig. 2e (control n = 5, Serpinb2-KO n = 6, Serpine1-KO n = 4). b, Representative human PDAC section stained for CK19, CD11b, CD68, fibrin(ogen), PAI1 and PAI2. Cohort details in Extended Data Fig. 8d. Scale bars, 100 μm (main image) and 50 μm (enlarged view (right)). c,d, Spatial proximity of fibrin(ogen)-colocalized CD11b+F4/80+ cells to CK19+, PAI1+CK19+ or PAI2+CK19+ tumour cells. Quantification in d used control tumours (n = 6). e, Tumour mass following macrophage depletion (IgG control n = 4, IgG Serpinb2-KO n = 4, IgG Serpine1-KO n = 4, anti-CSF1 plus clodronate control n = 4, anti-CSF1 plus clodronate Serpinb2-KO n = 3, anti-CSF1 plus clodronate Serpine1- KO n = 4). Lipo, liposomes. f, Tumour mass after IgG or anti-CD18 treatment in mice bearing control, Serpinb2-KO or Serpine1-KO tumours (n = 5 per group or treatment group). g, Abundance of macrophage subsets determined by scRNA-seq in Serpinb2-KO and Serpine1-KO tumours versus control. h, UMAP of Arg1 expression (left) and fibrin(ogen) binding score (right) in macrophages. i, Schematic of BMDM fibrin culture assay. j, Mean fluorescence intensity (MFI) of CD86, ARG1 and PD-L1 in BMDMs cultured on plastic or fibrin-coated surfaces. Representative of three independent experiments. k, Cell-type enrichment analysis comparing the local microenvironment surrounding SERPINB2+ versus SERPINB2− (left) and SERPINE1+ versus SERPINE1− tumour regions (right). Effect sizes for three human PDAC datasets10,33,34. Macrop., macrophage. l, Enrichment of immune and stromal populations in proximity to SERPINB2+ and SERPINE1+ regions in three human PDAC datasets10,33,34. Two-sided Mann–Whitney U tests with Benjamini–Hochberg FDR correction. a,e,f,j, Two-sided unpaired Welch’s t-tests; dots represent individual mice (a,e,f) or technical replicates (j). d, Two-sided paired t-test. Bars show mean + s.d. Control denotes F8-targeting sgRNA cells unless indicated. Petri dish in i adapted from Servier Medical Art, microbiology and cell culture image kit, Petri dish illustration elements, CC BY 4.0 (https://smart.servier.com/); human body outlines in k,l adapted from Servier Medical Art, people image kit, human body outline illustration, CC BY 4.0 (https://smart.servier.com/).

conservation of a serpin-associated epithelial programme linked to immune modulation in PDAC. Serpin-expressing cells dominate local immunity In human and mouse pancreatic tumours, SERPINB2 and SERPINE1 are expressed by only a subset of malignant cells (Fig. 5a) and are associated with distinct microenvironments compared to non-serpin-expressing cells (Fig. 4k). We therefore asked whether this minority popula- tion is sufficient to drive broader TME remodelling and exert domi- nant, non-cell-autonomous effects. To model the low frequency of serpin-positive tumour cells in patient tumours, we co-transplanted 95% Serpine1-KO KPC cells with 5% serpin-expressing control KPC cells (Fig. 5d). Introduction of this small population led to preferential Fig. 5 | Rare SERPINB2- and SERPINE1-expressing tumour cells define immunosuppressive dominant programmes. a, Fraction of SERPINE1+ and SERPINB2+ tumour cells per patient across five human PDAC scRNA-seq datasets10,14,16,34,35. Only patients with more than 100 tumour cells were included (n = 71). b, Venn diagram showing the overlapping upregulated genes in SERPINB2+ and SERPINE1+ tumour cells (log2FC ≥ 1; adjusted P value ≤ 0.05). c, Pathway enrichment analysis of SERPINB2+ and SERPINE1+ malignant epithelial cells across the integrated human PDAC scRNA-seq dataset. Selected enriched pathways are shown. Complete results are reported in Supplementary Table 2. d, Schematic of the clonal mixing experiment. Serpine1-KO and control cells were mixed at a 95:5 ratio and orthotopically transplanted into immunocompetent mice. Adapted from ref. 47 (copyright © 2026 MyJoVE Corporation). e, Tumour mass of control, Serpine1-KO, and mixed tumours (95% Serpine1-KO:5% control; n = 5 per condition). f–h, Imaging-based quantification of intratumoural macrophages (f), CD8+ T cells (g) and GZMB+CD8+ T cells (h). Five mice were analysed per condition. i, Fraction of knockout and control tumour cells at injection and end-point in Serpine1-KO mixing experiments. Injected fractions represent the input cell composition. End-point (recovered) fractions are shown as mean + s.d. (n = 5 mice). j, Representative multiplex immunofluorescence images from tumours generated by mixing 95% Serpine1-KO and 5% serpin-expressing cells. Scale bars, 250 μm (main image) and 50 μm (enlarged view (right)). k, CD8+ T cell fractions across low-mixing control-majority, low-mixing Serpine1-KO-majority, and high-mixing regions of interest (ROIs). Up to 20 spatially distributed ROIs per category were sampled from each of 5 independent tumours. ROIs from five mice are shown. l, ROI-level relationship between control tumour cell fraction and CD8+ T cell abundance in Serpine1 mixing experiments; each dot represents one ROI. e–h,k Two-tailed, unpaired Welch’s t-test. Bars indicate mean + s.d. Human body outline in a–c adapted from Servier Medical Art, people image kit, human body outline illustration, CC BY 4.0 (https://smart.servier.com/).