1 

SMAD4 and KRAS Status Shape Cancer Cell-Stromal Crosstalk and Therapeutic 

Response in Pancreatic Cancer  

Eloise G. Lloyd1, Muntadher Jihad1, Judhell S. Manansala1, Wenlong Li1, Priscilla S. W. Cheng1, 

Gianluca Mucciolo1, Marta Zaccaria1, Sara Pinto Teles1, Joaquín Araos Henríquez1, Sneha 

Harish1, Rebecca Brais2, Sally Ashworth1, Weike Luo1, Paul M. Johnson1, Lisa Veghini3, Mireia 

Vallespinos1, Vincenzo Corbo3,4, Giulia Biffi1#

1 University of Cambridge, Cancer Research UK Cambridge Institute, Li Ka Shing Centre, 

Robinson way, CB2 0RE, Cambridge, UK 

2 Histopathology, Cambridge University Hospitals NHS Foundation Trust, Addenbrooke’s 

Hospital, Cambridge, UK 

2 Department of Engineering for Innovation Medicine, University of Verona, Verona, Italy 

4 ARC-Net Research Centre, University of Verona, Verona, Italy 

# Correspondence: Giulia Biffi, Cancer Research UK Cambridge Institute, Li Ka Shing Centre, 

Robinson way, CB2 0RE, Cambridge, UK, Giulia.Biffi@cruk.cam.ac.uk, +44 (0)1223 769564. 

Running title: SMAD4 and KRAS status shape pancreatic cancer biology 

Conflict of interest disclosure statement 

The authors declare no potential conflicts of interest. 

mailto:Giulia.Biffi@cruk.cam.ac.uk


 2 

ABSTRACT 

Pancreatic ductal adenocarcinoma (PDAC) contains an extensive stroma that modulates 

response to therapy, contributing to the dismal prognosis associated with this cancer. Evidence 

suggests that PDAC stromal composition is shaped by mutations within malignant cells, but most 

previous work has focused on pre-clinical models driven by KrasG12D and mutant Trp53. 

Elucidation of the contribution of additional known oncogenic drivers, including KrasG12V mutation 

and Smad4 loss, is needed to increase understanding of malignant cell-stroma crosstalk in 

PDAC. Here, we used single-cell RNA-sequencing to analyze the cellular landscape of Trp53-

mutant mouse models driven by KrasG12D or KrasG12V in which Smad4 was wild-type or deleted. 

KrasG12D Smad4-deleted PDAC developed a fibro-inflammatory rich stroma with increased 

malignant JAK/STAT cell signaling and enhanced therapeutic response to JAK/STAT inhibition. 

SMAD4 loss in KrasG12V PDAC differently altered the tumor microenvironment compared to 

KrasG12D PDAC, and the malignant compartment lacked JAK/STAT signaling dependency. Thus, 

malignant cell genotype impacts cancer cell and stromal cell phenotypes in PDAC, directly 

affecting therapeutic efficacy. 

 

STATEMENT OF SIGNIFICANCE 

SMAD4 loss differentially impacts malignant-stromal cell signaling and treatment sensitivity of 

pancreatic tumors driven by KRASG12D or KRASG12V, highlighting the importance of 

understanding genotype-phenotype relationships for precision therapy. 

  



 3 

INTRODUCTION 

Pancreatic ductal adenocarcinoma (PDAC) is the fourth most common cause of cancer-related 

death and >85% of patients succumb to their disease within 5 years (1). PDAC is characterized 

by an extensive stroma that contributes to this dismal prognosis. Cancer-associated fibroblasts 

(CAFs) are abundant stromal cells that modulate PDAC progression and therapy response (2). 

Distinct populations of CAFs have been described in PDAC and other malignancies (2), including 

myofibroblastic (myCAFs), inflammatory (iCAFs) and antigen-presenting (apCAFs) CAFs (3–5). 

Additional subsets have been identified within or across these CAF states due to distinct cells of 

origin, reprogramming and/or functions (2–4,6–11). Moreover, other stromal cells, including 

macrophages and neutrophils, have been shown to be heterogenous (12,13). Thus, improving 

PDAC survival for all patients will likely require a better understanding of the heterogenous 

nature of the microenvironment across tumors (14). Determining malignant cell-stroma 

vulnerabilities in distinct patient-relevant contexts may guide the design of precision therapies 

for PDAC.  

 

PDAC can be classified in basal/squamous and classical/progenitor subtypes, which contain 

different tumor microenvironments (TMEs) (15–18). However, this classification has not yet 

translated into improved treatment design, perhaps because these subtypes can co-exist (19–

21). Identifying distinct groups of PDAC with specific malignant cell-stroma characteristics and 

therapeutic sensitivities thus remains a priority. Four main genetic drivers of PDAC have been 

described: KRAS mutations (in >90% patients) - with G12D (>40%) and G12V (>30%) mutations 

being the most abundant, TP53 mutations (>70%), SMAD4 loss (>30%) and CDKN2A loss 

(>30%) (22,23). However, most current knowledge of stroma composition comes from 

KRASG12D, p53 mutant KPC (KrasLSL-G12D/+; Trp53LSL-R172H/+; Pdx1-Cre) genetically engineered 

mouse models (GEMMs) (24), which recapitulate features of patient disease progression but 

may not capture stromal differences across human PDAC. Indeed, evidence indicates that 

mutations in PDAC malignant cells can differently shape the stroma (25–28). For example, 

increased fibrosis and reduced CD8+ T cell infiltration has been associated with TP53-mutated 

PDAC, while BRCA-mutated and BRCA-WT PDACs have different CAF composition (25,26). 

Moreover, the stroma affects PDAC progression and therapy response (2). Therefore, distinct 



 4 

PDAC groups may respond differently to therapies depending on their genotype and associated 

TME.  

 

Malignant cell transforming growth factor beta (TGF-) signaling is a key player in shaping the 

PDAC stroma since it drives myCAF formation (5,11). Loss of SMAD4, which is involved in TGF-

β signaling, is typically a late event in PDAC development and correlates with metastasis 

formation and worse prognosis (23,29). This effect may be in part dependent on the ability of 

TGF- signaling-deficient malignant cells to remodel epithelial and stromal compartments (28). 

To determine how SMAD4 deficiency tunes the PDAC TME and identify candidate dependencies 

of these aggressive tumors, we established Trp53-mutant mouse models driven by KrasG12D or 

KrasG12V in which Smad4 was wild-type (WT) or deleted.  

 

MATERIALS AND METHODS 

 

Human PDAC tissues 

Human PDAC tissues used in this study were obtained from surgical resections of patients (both 

males and females) treated at the University and Hospital Trust of Verona (Azienda Ospedaliera 

Universitaria Integrata, AOUI) for curative intent. Written informed consent was acquired from 

patients before specimens' acquisition. Specimens were acquired under protocols approved by 

the AOUI Ethics Committee (Comitato Etico Azienda Ospedaliera Universitaria Integrata): 

approval number 1911 (Prot. n 61413, Prog 1911 on 19/09/2018) and approval number 3456 

(Prot. n. 55859, Prog 3456 on 22/09/2021). Clinical information of the human PDAC tissues 

used, including sex, age, histological diagnosis, staging and genotype information for KRAS, 

TP53, CDKN2A, SMAD4 and TGFBR2, is provided in Supplementary Table S1. Genotype 

information for PDA-XX samples has been previously published and is based on targeted 

sequencing approach of organoids derived from those tissues (30). The genotype for all VR-Px 

samples was assessed in the same way. All tissues analyzed in this study were treatment naïve.  

 

Mouse models 

Male and female C57BL/6J (strain number 632, Charles River, RRID:IMSR_JAX:000664), 

athymic nude nu/nu (strain number 490, Charles River, RRID:IMSR_CRL:490) and NOD SCID 



 5 

gamma (NSG) mice (strain number 614, Charles River, RRID:IMSR_JAX:005557) were 

purchased from the Charles River Laboratory (7-9-week-old at arrival). All animals are housed 

in accordance with the UK Home Office “Code of Practice for the Housing and Care of 

Experimental Animals” guidelines. All animal procedures and studies were reviewed by the 

Cancer Research UK Cambridge Institute (CRUK-CI) Animal Welfare Ethics Review Board 

(AWERB), approved by the Home Office and conducted under project license number 

PP4778090 in accordance with relevant institutional and national guidelines and regulations.  

 

Orthotopic transplantation models of PDAC 

Pancreas injections were conducted as previously described (5,11). Briefly, single cells (10,000 

murine cells/mouse or 30,000-700,000 human cells/mouse (700,000 cells were injected for NSG 

cohorts 5 and 6) prepared from PDAC organoid cultures were resuspended as a 35 L 

suspension of 50% Matrigel in PBS and injected into the pancreas of 8-10-week-old mice. 

Pancreatic tumors were imaged using the Vevo 2100 Ultrasound at two different orientations 

with respect to the transducer. Tumor volumes were measured at two or three angles, whenever 

possible, using the Vevo LAB software program (version 5.7.0). Tumor volume analyses were 

performed blindly prior to plotting the data for visualization. Only mice with successful (i.e., non-

leaked) orthotopic injections were included for tumor volume and metastases analyses. For each 

experiment, SMAD4 WT and SMAD4-deficient/knockout (KO) cohorts were transplanted on the 

same day and later imaged by ultrasound on the same day. NSG mice were used for 

transplantation of human PDAC organoids, nu/nu mice were used for transplantation of KvPC 

(i.e. from the KrasFRT-LSL-G12V-FRT/+; Trp53LSL-R172H; Pdx1-Cre; Rosa26-FlpOERT2 GEMM) 

organoids since this GEMM had a mixed, not pure C57BL/6J, background (31). Nu/nu mice were 

also used for transplantation of KPC PDAC organoids that were generated by stable expression 

of CAS9, rather than CAS9 electroporation (see below for details about the two CRISPR 

technologies used to generate Smad4 WT and Smad4 KO KPC organoid lines), and to enable 

comparison with KvPC PDAC models. C57BL/6J mice were used for all other transplant models. 

Tumors were collected at similar sizes and weights for downstream analyses. 

 

In vivo AZD1480 treatment study 



 6 

The drug was prepared daily as a suspension in 0.1% Tween80, 0.5% hydroxyl propyl methyl 

cellulose in sterile water, and sonicated before administration. Palpable pancreatic tumors in 

C57BL/6J mice were imaged prior to enrolment (day -1) and at endpoint (day 14) using the Vevo 

2100 Ultrasound at two-three different orientations, whenever possible, with respect to the 

transducer. Mice with tumor diameters of 5 to 8 mm were randomized and enrolled 1 day after 

scanning (i.e. day 0). Mice were administered vehicle or 50-60 mg/kg of AZD1480 (S2162, CAS 

935666-88-9; Selleck or HY-10193, CAS 935666-88-9; MedChem Express) for 14 days, once a 

day (in the AM) via oral gavage. Tumor volumes were measured at two or three angles, 

whenever possible, using the Vevo LAB software program (version 5.7.0), and tumor growth 

based on these measurements was calculated by dividing the volume at day 14 for the volume 

at day -1. Tumor volume analyses were performed blindly prior to plotting the data for 

visualization. 

 

Cell lines and cell culture 

Murine pancreatic stellate cells (PSCs) (SV40-immortalized, C57BL/6J background), apart from 

PSC21, and murine PDAC KPC organoid lines (C57BL/6J background) were previously 

described (4,32). The PSC21 line was generated, immortalized and characterized as previously 

described (4,11). Briefly, to establish PSC21, we utilized two and a half pancreata from male 

C57BL/6J mice, and a density gradient centrifugation method with Histodenz (D2158; Sigma-

Aldrich) and Gey’s Balanced Salt Solution (G9779; Sigma-Aldrich). Murine KvPC (i.e. FPC 

organoids from the KrasFRT-LSL-G12V-FRT/+; Trp53LSL-R172H; Pdx1-Cre; Rosa26-FlpOERT2 GEMM, not 

pure C57BL/6J background) organoids were kindly obtained via Material Transfer Agreement 

(MTA) from Professor Tuveson (Cold Spring Harbor Laboratory, CSHL) and have been 

previously published (31). They were then cultured for three passages in complete organoid 

media with 10 M Nutlin-3a (SML0580; Sigma-Aldrich) to enrich for organoids that undergone 

p53 loss of heterozygosity (LOH) and generate T-LOH organoid lines, as previously done for 

KPC organoids (32). By inhibiting the interaction between p53 and the E3 ubiquitin protein ligase 

MDM2, Nutlin-3a leads to WT p53 activation, consequential depletion of cells that retain the WT 

Trp53 allele and upregulation of mutant p53 levels (32–34). Mouse PSCs were cultured in DMEM 

(41966029; Gibco) containing 5% FBS. All cells were typically cultured up to 40 passages, 

whenever possible, at 37C with 5% CO2. Cell line authentication was previously performed at 



 7 

the CRUK-CI for murine PSC4 and PSC5. Mycoplasma testing for murine PSCs was performed 

prior to each freezing. Human organoids were kindly obtained via MTA from Professor Tuveson 

(CSHL) and have been previously published (35). All tissue donations to generate these 

organoids had been reviewed and approved by the Institutional Review Board of CSHL and all 

clinical institutions. Written informed consent was obtained prior to acquisition of tissue from all 

patients. The studies were conducted in accordance with recognized ethical guidelines 

(Declaration of Helsinki). Clinical information of the human PDAC organoids used is provided in 

Supplementary Table S2. 

 

In vitro cell treatments 

PSCs were treated in Matrigel in 5% FBS DMEM with PDAC organoid conditioned media (CM) 

for as long as specified in the figure legends. KPC and KvPC PDAC organoids were treated with 

8 M AZD1480 for 48 hours prior to Western blot analysis of phospho-STAT3 (p-STAT3) and 

STAT3 levels or for 144 hours for proliferation assays. PDAC organoids were treated with 5-20 

ng/mL human TGF-β1 (T7039-2UG; Sigma), 2 μM A83-01 (SML0788; Sigma), 10 ng/mL 

interleukin 1 alpha (IL-1, 400-ML-005/CF; R&D Systems),  3-5 μg/mL IL-1-neutralizing 

antibody (MAB4001; R&D Systems; RRID:AB_2124216) or IgG control (400902; BioLegend) in 

reduced media (i.e. 5% FBS DMEM) for 48-72 hours prior to collecting protein or RNA, as 

indicated in the figure legends. Organoids were treated with 1-2 nM Trametinib (HY-10999, CAS 

871700-17-3; MedChemExpress) and 1 μM Pexmetinib (HY-16782, CAS 945614-12-0; 

MedChemExpress) in 5% FBS DMEM media for 72 hours prior to collecting protein or RNA. 

 

PCR-based genotyping of Trp53 1loxP 

PCR-based genotyping of Trp53 1loxP was previously described (32). Briefly, organoids were 

harvested and centrifuged at 1,000 rpm for 5 min at 4°C. Genomic DNA was extracted from 

organoids with DNEasy Blood & Tissue Kit (69504; Qiagen). Trp53 1loxP genotyping PCR 

reaction was performed in a 20 L reaction using AmpliTaq Gold 360 master mix (4398881; 

Thermo Fisher Scientific), 0.5 M each primer (p53loxF AGCCTGCCTAGCTTCCTCAGG and 

p53loxR CTTGGAGACATAGCCACACTG) and 100 ng of template DNA. The PCR cycling 

conditions were 95°C for 10 min, followed by 40 cycles at 95°C for 30 s, 56°C for 30 s, and 72°C 



 8 

for 30 s, then 72°C for 7 min (BioRad T100 Thermocycler). PCR products were separated on a 

2% agarose gel in TAE buffer. Gel imaging was performed with a Syngene U:Genius 3. 

 

Smad4 CRISPR/Cas9 knockout 

We utilized two different strategies to generate Smad4 KO organoids.  

 

To knock out Smad4 in organoids for KPC T6-LOH and T69A KO 1 and KO 2 clones/pools, lenti-

Cas9-Blast plasmids (52962; Addgene) for stable expression of CAS9 protein were used as 

previously described (5). Briefly, organoids were prepared as single cells and infected and 

selected using 2 g/mL blasticidin (A11139-03; Thermo Fisher Scientific). Single guide RNAs 

(sgRNAs) targeting exon 2 (for KO 1) or exon 3 (for KO 2) were designed using Benchling 

(RRID:SCR_013955) and cloned into the LRGN (LentisgRNA-EFS-GFP-neo) plasmid. Thus, 

organoids expressing these guides are green fluorescent protein (GFP)-positive and were used 

for the establishment of organoid-derived mouse models used for single-cell RNA-sequencing 

(scRNA-seq). Organoids were plated as single clones in the presence of geneticin (10131035; 

Thermo Fisher Scientific). Knockout was confirmed by western blot analysis (4-5 

clones/guide/organoid line). sgRNAs against the Rosa26 locus were included to generate control 

(i.e. WT) lines, and these Smad4 WT controls were kept as pools. 

 

To knock out Smad4 in organoids for KPC T6-LOH and T69A 263 and 264 clones/pools (i.e. KO 

3 and KO 4) and KvPC T93-LOH and T95-LOH 263 and 264 clones/pools (i.e. KO 3 and KO 4), 

CRISPR guides (phosphorothionate-modified sgRNA, 263: ATCAGGCCACCTCCACAGAC, 

264: AGACGGGCATAGATCACATG) were designed against exon 3 of the murine Smad4 gene 

(ENSMUST00000025393.14). A guide which targets the Rosa26 locus was also included to 

generate control (i.e. WT) lines (265: GAAGATGGGCGGGAGTCTTC), and these Smad4 WT 

controls were kept as pools. Mouse PDAC organoids were dissociated into single cells and 

100,000 cells were electroporated using an Amaxa 4D Nucleofector unit (Lonza) with 4 g 

TrueCut spCas9 protein V2 (A36498; Invitrogen) and 80 pmol guide RNA (Synthego), using 

program CM-137 and P3 nucleofector solution (V4XP-3032; Lonza). A cell pellet was taken 3- 

and 10- days post electroporation, and genomic DNA was extracted using the DNeasy blood 

and tissue kit (69506; Qiagen). Exon 3 of Smad4 was amplified by PCR using the Q5 High 



 9 

Fidelity DNA polymerase (M0491S; NEB). Smad4 primers used were FWD: 

TTCCCTTCAGCAGAAGCTGG, and REV: TGCTTCCCATACTGTTTGCA. Amplicons were 

subjected to Sanger sequencing and analyzed using Synthego ICE web tool to calculate the 

percent editing in a pool. Organoids were plated as single clones and knockout was confirmed 

by western blot analysis (6-7 clones/guide/ KPC organoid line; 5-7 clones/guide/ KPC organoid 

line). 

 

Western blot analyses 

PSCs and organoids were harvested in Cell Recovery Solution (354253; Corning) supplemented 

with complete, mini protease inhibitors (11836170001; Roche) and a phosphatase inhibitor 

cocktail (4906837001; Roche) and incubated for 30 min at 4C. Cells were pelleted at 1500 rcf 

for 5 min and lysed in 0.1% Triton X-100, 15 mmol/L NaCl, 0.5 mmol/L EDTA, 5 mmol/L Tris, pH 

7.5, supplemented with complete, mini protease inhibitors (11836170001; Roche) and a 

phosphatase inhibitor cocktail (4906837001; Roche). Cells were incubated on ice for 30 min 

briefly vortexed and pelleted at 13,200 rpm for 10 min at 4C. Concentration of protein collected 

in the supernatant was determined using DC protein assay (5000113-5; Bio-Rad). Standard 

procedures were used for western blotting. Primary antibodies used were ACTIN (8456; Cell 

Signaling Technology; RRID:AB_10998774), HSP90 (07-2174;Millipore; RRID:AB_10807022), 

SMAD4 (sc-7966; Santa Cruz; RRID:AB_627905), p-STAT3 (9145; Cell Signaling Technology; 

RRID:AB_2491009), STAT3 (9139; Cell Signaling Technology; RRID:AB_331757), p53 (P53-

CM5P-L; Leica; RRID:AB_2744683), HIF-1 (14179; Cell Signaling Technology; 

RRID:AB_2622225), p-p44/42 (p-Erk1/2) (4370; Cell Signaling Technology; 

RRID:AB_2315112), p44/42 (Erk1/2) (4695; Cell Signaling Technology; RRID:AB_390779), p-

p38 (4511; Cell Signaling Technology; RRID:AB_2139682), p38 (9212; Cell Signaling 

Technology; RRID:AB_330713), HSP60 (12165; Cell Signaling Technology; 

RRID:AB_2636980), SMAD2 (5339; Cell Signaling Technology; RRID:AB_10626777), 

p-SMAD2/SMAD3 (8828; Cell Signaling Technology; RRID:AB_2631089)  and GFP (ab6673; 

Abcam; RRID:AB_305643). Proteins were detected using appropriate HRP-conjugated 

secondary antibodies (Jackson ImmunoResearch Laboratories). All western blots are 

representative examples and have been repeated for at least two biological replicates. 

 



 10 

Proliferation assays 

For proliferation assays of PDAC organoids, 5,000 single cells were plated in 50 L of 100% 

Matrigel on 24-well plates (Corning/Nunc) and cultured in 500 L of reduced media (i.e. 5% FBS 

DMEM) or complete mouse organoid media (36). Organoid proliferation was followed for 96-144 

hours with an Incucyte organoid module (Sartorius) with measurement of the organoid area per 

well every 3 hours (with 4 technical replicates per measurement). Data were normalized to the 

first measurement (at 3 hours post-plating on day 0). For proliferation assays of PDAC organoids 

with 8 M of the JAK inhibitor (JAKi) AZD1480, 2,000-3,000 single cell organoids were plated in 

25 L of 100% Matrigel on 48-well plates (Corning/Nunc) and cultured in 250 L of reduced 

media. Organoid proliferation was followed as before, data were normalized to the first 

measurement (i.e. average of 6 technical replicates, at 3 hours post-plating on day 0) and used 

to calculate cell viability in comparison to the respective DMSO-treated controls. 

 

Immunohistochemical and histological analyses 

Human and murine organoid-derived Smad4 WT and Smad4 KO tumors were collected at 

comparable tumor sizes and weights. Standard procedures were used for immunohistochemistry 

(IHC). Primary antibodies for IHC were alpha smooth muscle actin (SMA, ab5694; Abcam; 

RRID:AB_2223021), SMAD4 (sc-7966; Santa Cruz; RRID:AB_627905), E-cadherin (ECAD, 

610182; BD Biosciences; RRID:AB_397581), Ki67 (12202; Cell Signaling Technology; 

RRID:AB_2620142), pH3 (9701S; Cell Signaling Technology; RRID:AB_331535) and p-STAT3 

(9145; Cell Signaling Technology; RRID:AB_2491009). Hematoxylin (H-3404-100, Vector Lab) 

was used as nuclear counterstain. Masson’s trichrome and Hematoxylin & Eosin stains were 

performed according to standard protocols by the Histology core at the CRUK-CI. FOXP3 and 

Ly6G stains were also performed by the Histology core. Briefly, after baking at 60C for 1 hour, 

sections were dewaxed and rehydrated on Leica’s automated ST5020. The staining was 

performed on Leica’s automated Bond-III platform in conjunction with their Polymer Refine 

Detection System (DS9800) and a modified version of their standard template, with protein block 

(X090930-2; Dako), anti-rat secondary antibody (A110-322A; Bethyl Laboratories; 

RRID:AB_10681533) and DAB Enhancer (AR9432; Leica). Primary antibodies used were Ly6G 

(127601; BioLegend; RRID:AB_1089179) and FOXP3 (14-5773; eBioscience; 

RRID:AB_467576). Antigen retrieval was performed at 100 C for 20 min in citrate or Tris EDTA 



 11 

for Ly6G or FOXP3, respectively. De-hydration and clearing were performed on Leica’s 

automated ST5020 before sections were mounted on Leica’s coverslipper, CV5030. Stained 

sections were scanned with Aperio ScanScope CS and analyzed using ImageScope software 

(RRID:SCR_014311) Positive Pixel Count algorithms or a Nuclear v9 algorithm, depending on 

the marker quantified.  

 

Images of tissue slides were obtained with an Axio Vert.A1 (ZEISS) apart for human PDAC 

tumors, which were snapshotted from the ImageScope software. The percentage of collagen 

area was determined by calculating the percentage of blue pixels relative to the entire stained 

area. To quantify SMA stain, the percentage of positive pixels was calculated relative to the 

entire section. To quantify p-STAT3, pH3 and Ki67 stains, the percentage of positive nuclei was 

calculated relative to the total number of nuclei. For Masson’s trichrome and SMA quantification 

of human PDAC tissues, only the PDAC area was included for analysis, following annotation 

from a pathologist. Human PDAC organoids hT1 and hT108 did not generate tumors within one 

year, following transplantation, and could not be analyzed. Tumor differentiation analysis was 

performed blindly by a pathologist by scoring the percentage of differentiated or undifferentiated 

area per tumor. We then defined tumors as ‘differentiated’ if > 60% was differentiated, 

‘undifferentiated’ if < 40% was differentiated and ‘mixed’ if 40-60% was differentiated. Analysis 

of epithelial/stroma proportion was done by calculating the percentage of ECAD+ area relative 

to the entire section. Stains and quantifications were performed blindly prior to plotting the data 

for visualization.  

 

Flow cytometry analyses 

Tumors were collected at comparable tumor sizes and weights, and processed as previously 

described (5). Cells were blocked for 15 min on ice with CD16/CD32 Pure 2.4G2 (553142, BD 

Bioscience; RRID:AB_394657). For flow cytometric analysis of endothelial cells, immune cells, 

epithelial cells and iCAFs, myCAFs, CD90+ CAFs, CD90- CAFs and apCAFs, cells were stained 

for 30 min on ice with anti-mouse CD31-PE/Cy7 (102418; BioLegend; RRID:AB_830757), 

CD45-PerCP/Cy5.5 (103132; BioLegend; RRID:AB_893344), CD326 (EpCAM)-AlexaFluor 488 

(118210; BioLegend; RRID:AB_1134099), PDPN-APC/Cy7 (127418; BioLegend; 



 12 

RRID:AB_2629804), MHCII-BV785 (107645; BioLegend; RRID:AB_2565977), Ly6C-APC 

(128015; BioLegend; RRID:AB_1732087) and CD90-PE (ab24904; Abcam; RRID:AB_448474).  

 

For flow cytometric analysis of CD56+ and CD49E+ CAFs of KPC cohorts 1 and 2 in nu/nu mice, 

cells were stained for 30 min on ice with anti-mouse CD56-APC (FAB7820A; Bio-Techne), 

CD49e-PE (557447; BD Biosciences; RRID:AB_396710), CD326 (EpCAM)-AlexaFluor 488 

(118210; BioLegend; RRID:AB_1134099), PDPN-APC/Cy7 (127418; BioLegend; 

RRID:AB_2629804), CD45-BV785 (103149; BioLegend; RRID:AB_2564590), CD26 

PerCP/Cy5.5 (45-0261-82; Thermo Fisher Scientific; , RRID:AB_1548738) and CD34 PE/Cy7 

(25-0349-41; Thermo Fisher Scientific; RRID:AB_1963577). For flow cytometric analysis of 

CD105+ CAFs of KPC cohorts 1 and 2 in nu/nu mice, cells were stained for 30 min on ice with 

anti-mouse CD105-PE/Cy7 (120409; BioLegend; RRID:AB_1027702), CD45-PerCP/Cy5.5 

(103132; BioLegend; RRID:AB_893344), CD326 (EpCAM)-AlexaFluor 488 (118210; 

BioLegend; RRID:AB_1134099), PDPN-APC/Cy7 (127418; BioLegend; RRID:AB_2629804), 

Ly6C BV785 (128041;BioLegend; RRID:AB_2565852) and CD140a PE (135906; BioLegend; 

RRID:AB_1953269). 

 

For flow cytometric analysis of CD56+, CD49E+ and CD105+ CAFs of KPC cohort 3 in nu/nu 

mice and all other KPC and KvPC cohorts, cells were stained for 30 min on ice with anti-mouse 

CD56-APC (FAB7820A; Bio-Techne), CD49e-PE (557447; BD Biosciences; RRID:AB_396710), 

CD105-PE/Cy7 (120409; BioLegend; RRID:AB_1027702), CD326 (EpCAM)-AlexaFluor 488 

(118210; BioLegend; RRID:AB_1134099), PDPN-APC/Cy7 (127418; BioLegend; 

RRID:AB_2629804), CD45-BV785 (103149; BioLegend; RRID:AB_2564590) and CD26 

PerCP/Cy5.5 (45-0261-82; Thermo Fisher Scientific;  RRID:AB_1548738). 

 

For flow-cytometric analysis of macrophages and neutrophils of all KPC and KvPC cohorts, cells 

were stained for 30 min on ice with anti-mouse CD45-PerCP/Cy5.5 (103132; BioLegend; 

RRID:AB_893344), CD11b-PE/Cy7 (101215; BioLegend; RRID:AB_312798), Ly6C-Alexa488 

(128021; BioLegend; RRID:AB_10640820), F4/80-BV785 (123141; BioLegend; 

RRID:AB_2563667), MHCII-APC/Cy7 (107627; BioLegend; RRID:AB_1659252), CD11c-APC 

(117309; BioLegend; RRID:AB_313779) and Gr1-PE (108407; BioLegend; RRID:AB_313372).  



 13 

 

For flow-cytometric analysis of B cells, T cells and NK cells of KPC cohorts in C57BL/6J mice, 

cells were stained for 30 min on ice with anti-mouse CD45-PerCP/Cy5.5 (103132; BioLegend; 

RRID:AB_893344), CD19-PE/Cy7 (115520; Biolegend; RRID:AB_313655), NK1.1-BV785 

(108749; Biolegend; RRID:AB_2564304), TCR-Alexa488 (109215; Biolegend; 

RRID:AB_493344); CD3-Alexa488 (100321; Biolegend; RRID:AB_389300), CD4-APC 

(100516; Biolegend; RRID:AB_312719) and CD8a-APC/Cy7 (100713; Biologend; 

RRID:AB_312752).  

 

Cells were resuspended in PBS with DAPI and analyzed on a BD FACSymphony cell analyzer. 

Flow analyses were performed blindly using FlowJo 10.8.2 (RRID:SCR_008520) prior to plotting 

the data for visualization.  

 

Cell sorting of PDAC organoid/PSC co-cultures for RNA-sequencing 

Sorting of PDAC organoid/PSC co-cultures was performed following 3.5 days culture in reduced 

media (i.e. 5% FBS DMEM). Following single cell digestion of co-cultures, cells were stained for 

30 min on ice with anti-mouse CD326 (EpCAM)-PE (118205; BioLegend; RRID:AB_1134176) 

and PDPN-AlexaFluor 488 (156208; BioLegend; RRID:AB_2814080). Cells were resuspended 

in PBS with DAPI and sorted with a BD FACSMelody cell sorter. 

 

RNA-sequencing analyses of PDAC organoids and PSCs flow-sorted from co-cultures 

RNA-seq data of KPC monocultures and co-cultures are available at the Gene Expression 

Omnibus (GEO) under the accession number GSE263080. RNA-seq data of KvPC 

monocultures and co-cultures are available at the GEO under the accession number 

GSE263081. Samples were collected in 1 mL of TRIzol Reagent (15596018; Invitrogen). RNA 

was extracted using the PureLink RNA mini kit (12183018A; Invitrogen). RNA concentration was 

measured using a Qubit and RNA quality was assessed on a TapeStation 4200 (Agilent) using 

the Agilent RNA ScreenTape kit. mRNA library preparations were performed using 55 L of 10 

ng/mL per sample (RNA integrity number, RIN > 8, except for one sample with RIN 7.2). Illumina 

libraries were then sequenced on 1 lane of SP flowcell on NovaSeq6000. FASTQ files were 

aligned, and the expression levels of each transcript were quantified using Salmon (v1.4.0) (37) 



 14 

with the annotation from ENSEMBL (GRCH39, release 109) with recommended settings. 

Transcript-level expression was loaded and summarized to the gene level by using tximport (38). 

Differential gene expression analysis was performed using DESeq2 (v2) (39) by applying 

lfcshrink function (40). The principal components for variance-stabilized data were estimated 

using plotPCA function, available in DESeq, and ggplot2 (https://ggplot2.tidyverse.org). Genes 

with adjusted p-value < 0.05 were selected as significantly differentiated between conditions. 

Following differential gene expression, genes were pre-ranked based on the negative 

logarithmic p-value and the sign of the log2 fold change. GSEA was performed using 

clusterprofiler against the Hallmark, Reactome, and C2 canonical pathway collection 

(C2.cp.v5.1) downloaded from the Molecular Signatures Database (MSigDB) (41).  

 

NicheNet: NichenetR (42) was used to infer the ligands activity and their regulation potential of 

a sender group by considering the expression of downstream genes in the receiver group (i.e. 

target genes). Ligands were ranked based on their area under the precision-recall curve (AUPR) 

to prioritize ligands inferred to regulate the target genes in the receiver population. NicheNet was 

applied to infer the interaction between PDAC organoids and PSCs in KPC Smad4 WT and KPC 

Smad4 KO conditions. “Target genes” were defined by using significant differentially expressed 

genes (DEGs, p.adj < 0.05, log2FC > 1) in each condition of the receiver group, and the “ligands” 

were defined the same way in the sender group. All expressed genes in the dataset were 

considered as “background expressed genes”.  

 

Transcription Factor (TF) activity analysis: TF activity analysis was performed with CollecTRI 

network (43) using a univariate linear model as implemented in decouplerR package (44). The 

stat column from the DESeq2 output was used as input. TFs with p-value < 0.05 were considered 

to have significant activity.  

 

RNA-sequencing analyses of Smad4 KO and Smad4 WT KPC PDAC tumors  

Tumors from C57BL/6J mice were collected at comparable tumor sizes and weights. RNA-seq 

data of Smad4 KO and Smad4 WT KPC PDAC tumors are available at the GEO under the 

accession number GSE274684. Tumor pieces were snap frozen in liquid nitrogen and stored at 

-80 C until processing. Samples were thawed and transferred to low bind tubes with 1 mL TRIzol 

https://ggplot2.tidyverse.org/


 15 

reagent (15596018; Invitrogen) and 1 metal bead (69989; Qiagen). Samples were then 

homogenized at frequency 30/s for up to 6 min using a TissueLyser II (Qiagen). RNA was 

extracted as above. mRNA library preparations were performed using 1 mg of RNA per sample 

(RIN 5-9). Illumina libraries were sequenced on 1 lane of SP flowcell on a NovaSeq6000. RNA-

seq analysis was performed as above although the transcript annotation used was GRCH39 

(release 109) from ENSEMBLE. 

 

Single-cell RNA-sequencing analyses of PDAC tumors 

Tumors from nu/nu mice were collected at comparable tumor sizes and weights. scRNA-seq 

data of KPC tumors are available at the GEO under the accession number GSE262879. scRNA-

seq data of KvPC tumors are available at the GEO under the accession number GSE262878. 

Cell Ranger (10x Genomics) workflow (45) was used to align FASTQ files to GRCh38 (mm10) 

mouse transcriptome reference to generate the raw counts of gene expression quantities. For 

the KPC dataset, the reference genome was modified by adding Gfp to it. From the raw counts 

of each sample, the SOLO tool that is implemented in scvi-tools was used to estimate doublets 

to be removed (46). The samples were integrated, and batch effect was removed using scvi-

tools (47). Scanpy workflow (48) was used for dimension reductions, clustering and defining 

markers of each cluster. Cells that have the percentage of mitochondrial genes more than 5% 

were filtered out. Unique molecular identifiers (UMI) were normalized to 10,000 counts and 

Leiden graph clustering were used for unsupervised clustering to identify the cell populations 

with similar transcriptomic profile. For dimension reduction we built the top 30 principal 

components (PC) and nearest neighbours’ graph (k = 10) on 2,000 highly variable genes. 

Uniform manifold approximation and projection (UMAP) were used to visualize the datasets in 

2-dimentional space. Markers of each cluster were defined using “rank_genes_groups” function 

from Scanpy. 

 

Copy Number Variation (CNV) analysis: A python implementation of inferCNV of the Trinity 

CTAT Project (https://github.com/broadinstitute/inferCNV) was used to estimate the copy 

number status in each cell type. We used the fibroblast cells as a reference key and a 250-genes 

window size. 

 

https://github.com/broadinstitute/inferCNV


 16 

Abundance test: We used Milo (49) python framework to compute differentially abundant 

neighbourhoods within the cell type clusters across conditions (i.e. Smad4 KO vs Smad4 WT). 

The following parameters were used k = 10, min proportion (p = 0.1) and dimensionality (d = 20). 

SpatialFDR < 0.1 was used for significance.  

 

Pseudobulk: For DE analysis we pooled all cells within a specific cell type by summing the gene 

expression of each gene to create a pseudo-bulk expression profile of each sample. 

DESeq2 uses the pseudo-bulk data to detect the differences between two sample groups. To 

generate the pseudo-bulk profile and perform DEA, we used the python implementation of 

decoupler R package (44) with the default options. GSEA was performed as with bulk RNA-seq 

data. 

 

CellChat: CellChat R package (50) was used with the recommended setting to infer and 

visualize the cell-cell communication between specific cell types in scRNA-seq data. 

 

NicheNet: NicheNet (42) was used to infer the ligands activity and their regulation potential of a 

sender group by considering the expression of downstream genes in the receiver group (i.e. 

target genes). “Target genes” were defined based on the DEGs using findmarkers function from 

Seurat in the receiver cell clusters between two conditions (e.g. KO vs WT) using MAST (51) (P 

value < 0.05, logFC > 0.25). Genes that were expressed by 5% of the receiver cell cluster were 

considered as background genes. “Ligands” were defined based on the DEGs using findmarkers 

function in the sender cell clusters between the same two conditions (e.g. KO vs WT) using 

MAST (P value < 0.05, logFC > 0.25). 

 

Quantitative PCR analyses 

Media was removed and cells were resuspended in TRIzol. RNA was prepared using a standard 

TRIzol/chloroform extraction method. Briefly, samples in TRIzol were transferred to 5PRIME 

phase lock gel tubes (2302830, VWR International) and mixed with chloroform before 

centrifuging for 15 min. The supernatant was then centrifuged for 10 min with isopropanol and 

glycogen (R0551, Fisher Scientific). Pellets were washed in ice-cold 75% ethanol and 

centrifuged again. Samples were resuspended in RNAse-free water and RNA concentration was 



 17 

quantified using a NanoDrop (ThermoFisher). Reverse transcription was performed with 100 ng 

- 1 µg of RNA using TaqMan reverse transcription reagents (N808-0234; Applied Biosystems) 

and following the manufacturer’s instructions. qPCR was performed using TaqMan master mix 

(4440040; Applied Biosystems) and TaqMan gene-specific probes (Applied Biosystems) on a 

QuantStudio 6 Flex Real-Time PCR system. Gene expression was normalised to Hprt 

housekeeping gene.  

 

Enzyme Linked Immunosorbent Assay (ELISA) 

Organoid-conditioned media samples were collected from organoids grown in 5% FBS DMEM 

for 3-5 days. ELISA assays were used to detect murine IL-1 (MLA00; R&D Systems) and 

murine TGF- (BMS608-4, Thermo Fisher Scientific). Media was assayed according to the 

manufacturer’s instructions. 

 

Statistical analysis 

GraphPad Prism software (RRID:SCR_002798), customized R and python scripts were used for 

graphical representation of data. Statistical analysis was performed using non-parametric Mann-

Whitney test, unpaired or paired Student’s t test or chi-square test. All statistical details of 

experiments are specified in the figure legends and/or panel figures, including the number of 

technical and biological replicates, and how significance was defined. 

 

Data availability 

For RNA-seq and scRNA-seq datasets, the data generated in this study are publicly available at 

the GEO under the accession numbers GSE263080, GSE263081, GSE262879, GSE262878 

and GSE274684. 

 

For ultrasound-based tumor volume, western blot, qPCR, organoid proliferation assay, flow 

cytometry, histology and immunochemistry data, the raw data are available on request from the 

corresponding author. No publicly available data was reused. This paper does not report original 

code. 

 

RESULTS 



 18 

 

Smad4 loss impacts the immune TME and malignant-stromal crosstalk in KPC PDAC 

Human PDAC tumors display remarkable inter-tumoral clinical, histological, and genetic 

heterogeneity, rendering it difficult to understand how mutations cooperate to drive 

tumorigenesis and remodel the TME (Supplementary Fig. 1A-B; Supplementary Table S1). 

Therefore, to start to dissect the relationship between malignant cell genotype and stromal 

phenotype, we engrafted patient-derived PDAC organoids harboring KRASG12D (K) or KRASG12V 

(Kv) with TP53-mutant (P) or TP53-loss (PP), and SMAD4 WT or SMAD4 deficient status into 

the pancreata of NOD SCID gamma (NSG) mice (35) (Fig. 1A; Supplementary Fig. 1C-D; 

Supplementary Table S2). SMAD4 loss accelerated PDAC progression in KRASG12D and 

KRASG12V TP53-mutant PDAC (i.e. KP and KvP) (Fig. 1B-C; Supplementary Fig. 1E-F). 

SMAD4 loss also increased metastasis formation in KP and KvP PDAC compared to SMAD4 

WT KP and KvPP tumors of similar size (Supplementary Fig. 1G-H). Together, these results 

agree with patient data showing that SMAD4 loss typically drives an aggressive PDAC 

phenotype, although analysis of isogenic human organoid-derived models would better address 

this (29,52). Finally, SMAD4 loss decreased the desmoplasia – measured as collagen deposition 

and alpha smooth muscle actin (SMA) levels (Fig. 1D-F; Supplementary Fig. 1I). To further 

explore the impact of SMAD4 loss on PDAC progression, overcome inter-patient heterogeneity 

and study bi-directional malignant-stromal crosstalk, we generated murine isogenic organoids 

with Smad4 WT or deleted. 

 

KRASG12D is the most frequent KRAS mutant allele in PDAC patients (23). Thus, to further study 

the role of KRAS and SMAD4 status in PDAC, we deleted Smad4 from KPC GEMM-derived 

organoids (24,32,36) (Supplementary Fig. 2A-D). Isogenic Smad4 knockout (KO) KPC 

(hereafter, KPCSmad4-KO) organoids grew faster than Smad4 WT KPC (hereafter, KPCSmad4-WT) 

organoids both in vitro and as orthotopically-grafted tumors in immunocompromised nu/nu or 

immunocompetent C57BL/6J mice (Fig. 2A-D; Supplementary Fig. 2E-F).  

 

To understand how SMAD4 loss in KPC malignant cells might impact the TME, we generated 

single-cell RNA-sequencing (scRNA-seq) profiles of similar size KPCSmad4-WT and KPCSmad4-KO 

tumors in which the malignant cells were engineered to express green fluorescent protein (GFP) 



 19 

as part of one of our CRISPR strategies (Fig. 2E-F; Supplementary Fig. 2G-K;  

Supplementary Table S3). Similar proportions of cell populations were previously found in 

similarly processed human PDAC tumors (3). scRNA-seq profiles of malignant cells could be 

readily distinguished by the expression of Gfp transcripts and inferred copy number variations 

(CNVs; Supplementary Fig. 2L-M). Smad4 deletion markedly altered the transcriptomes of 

KPCSmad4-KO relative to KPCSmad4-WT malignant cells (6,121 differentially expressed genes, DEGs, 

false discovery rate, FDR < 0.05) (Fig. 2G). Remarkably, scRNA-seq profiles also revealed 

significant and selective changes in the transcriptomes and proportions of CAFs, macrophages 

and neutrophils in KPCSmad4-KO relative to KPCSmad4-WT tumors (Fig. 2G-I). Thus, SMAD4 loss in 

murine PDAC malignant cells impacts both CAFs and innate immune cells.  

 

Additional analyses confirmed an increase in neutrophils and a decrease in macrophages in 

KPCSmad4-KO relative to KPCSmad4-WT tumors of similar size and showed a significant decrease in 

natural killer (NK) cells (Fig. 2J-M; Supplementary Fig. 2G and 2N-U). Of note, C57BL/6J mice 

with KPCSmad4-KO PDAC showed more metastases than KPCSmad4-WT models with comparable 

tumor size (Supplementary Fig. 2O and 2V). Analysis of neutrophils identified previously 

described immature T1, mature T2 and tumor-promoting T3 populations (13), as well as a small 

subset expressing T1 markers and Mpo, Ly6g and Ly6c1 transcripts (Fig. 2N-P; Supplementary 

Fig. 3A). The gene signature of T3 neutrophils was upregulated in KPCSmad4-KO tumors (Fig. 2Q; 

Supplementary Fig. 3B). Additionally, analysis of macrophages identified recently described 

sub-types of tumor-associated macrophages (TAMs) (12), as well as a small cluster, which we 

named Insyn2b-TAMs based on marker expression (Fig. 2R-T; Supplementary Fig. 3C). The 

expression of genes associated with interferon-primed TAMs (IFN-TAMs), lipid-associated 

TAMs (LA-TAMs) and inflammatory cytokine-enriched TAMs (Inflam-TAMs) were upregulated in 

KPCSmad4-KO relative to KPCSmad4-WT PDAC, while proliferating TAMs (Prolif-TAMs) were reduced 

(Fig. 2U-V).  

 

CellChat analysis (50), which infers ligand-receptor interactions and patterns of cell-cell 

communication, identified potential changes in communication among malignant cells, 

neutrophils, macrophages and CAFs in KPCSmad4-KO relative to KPCSmad4-WT PDAC (Fig. 3A). 

These included a role for malignant cell-derived tumor necrosis factor (TNF) in dictating 



 20 

macrophage composition, as well as macrophage- and malignant cell- derived (C-X-C motif) 

ligands 1 and 2 (CXCL1 and CXCL2) in neutrophil recruitment in KPCSmad4-KO PDAC (Fig. 3B-

D). Fibroblasts in KPCSmad4-KO tumors also appeared to be more involved in neutrophil 

recruitment via the CXCR2 pathway, as previously found in iCAF-rich PDAC (18). Of note, the 

most affected interaction in KPCSmad4-KO PDAC appeared to be between malignant cells and 

fibroblasts, in line with CAFs being the most impacted stromal cell population upon SMAD4 loss 

(2,294 DEGs, FDR < 0.05; Fig. 3A and 2G).  

 

To further investigate the crosstalk of these four cell populations, we applied NicheNet, which 

infers ligand-target relationships among cell types (42). This analysis suggested predominant 

TGF- signaling in KPCSmad4-WT tumors and interleukin 1 (IL-1) signaling in KPCSmad4-KO tumors 

from malignant cells and fibroblasts to macrophages (Fig. 3E; Supplementary Fig. 3D-F; 

Supplementary Table S4). This analysis also confirmed the role of KPCSmad4-KO malignant cell-

produced TNF in shaping macrophage composition and inferred a role for macrophage-

produced IL-1 and IL-6 in shaping malignant cells and fibroblasts (Fig. 3E-G; Supplementary 

Fig. 3G). Finally, it suggested that IL-1 from neutrophils contributes to the increase in iCAFs 

and highlighted a role of TNF and IL-1 in reciprocal malignant cell-fibroblast crosstalk in 

KPCSmad4-KO tumors (Fig. 3H-L; Supplementary Fig. 3H-L). 

 

These analyses show that loss of SMAD4 in murine KRASG12D p53-mutant PDAC malignant 

cells shapes the immune TME, and they suggest that CAFs are also profoundly impacted. 

 

Smad4 loss drives a fibro-inflammatory stroma in KPC PDAC 

While fibrosis and epithelial/stroma proportion were not affected, collagen deposition was 

reduced in KPCSmad4-KO tumors, further suggesting that malignant cell SMAD4 status impacts 

CAF composition in murine PDAC (Supplementary Fig. 4A-E). Indeed, proportions of iCAFs, 

myCAFs and apCAFs defined by scRNA-seq profiles significantly differ between KPCSmad4-KO 

and KPCSmad4-WT tumors (Fig. 4A-C; Supplementary Fig. 4F). Of note, scRNA-seq analysis 

inferred a decrease in the myCAF/iCAF ratio in KPCSmad4-KO tumors, which was confirmed by 

flow cytometry (Fig. 4C-F; Supplementary Fig. 4G-H). In keeping with these findings, iCAF-

associated pathways were upregulated while proliferation-associated pathways, which were 



 21 

shown to be enriched in myCAFs (5), were downregulated in KPCSmad4-KO tumors (Fig. 4G). 

Additionally, abundance of MHCII-expressing apCAFs was lower in KPCSmad4-KO tumors (Fig. 

4D-E; Supplementary Fig. 4G). PDAC apCAFs have been identified by both transcriptome 

signature and the expression of MHCII protein (3,10). Interestingly, SMAD4 loss in malignant 

cells resulted in the presence of CAFs that upregulate the iCAF signature but retain an apCAF 

transcriptome, while being negative for MHCII protein (Supplementary Fig. 4I).  

 

We further interrogated the heterogeneity of myCAFs by using CD90 (encoded by Thy1), which 

marks a myCAF subset (11), as well as CD105 (Eng, and a CAF-lineage marker (6)), CD49E 

(Itga5) and CD56 (Ncam1). CD105, CD49E and CD56 have not been described previously as 

myCAF markers but were enriched in myCAFs by scRNA-seq (Supplementary Fig. 4F). Each 

of these markers were significantly downregulated in CAFs in KPCSmad4-KO tumors, while tumor-

promoting CD90- myCAFs were increased (Fig. 4H-I; Supplementary Fig. 4J-N). Furthermore, 

malignant cell profiles in KPCSmad-KO tumors expressed high levels of Il1a and Il1b, which direct 

iCAF formation (5) (Fig. 4J-L). Thus, SMAD4 loss in KPC cells potentially shapes PDAC CAF 

composition through IL-1 signaling. 

 

To further deconvolute changes in malignant cell-fibroblast crosstalk upon SMAD4 loss, we 

leveraged a co-culture model of PDAC organoids and pancreatic stellate cells (PSCs) (4). PSCs 

are precursors of CAFs and model iCAFs and myCAFs in vitro (4,5,9,11). KPCSmad4-KO and 

KPCSmad4-WT organoids cultured either alone or with PSCs were flow-sorted and analyzed by 

RNA-seq (Fig. 4M; Supplementary Fig. 4O; Supplementary Tables S5-S6). In keeping with 

our in vivo observations, iCAF markers and associated pathways were increased, while myCAF 

markers and associated pathways were decreased in KPCSmad4-KO/PSC co-cultures (Fig. 4N; 

Supplementary Fig. 4P). Furthermore, the JAK/STAT pathway and hypoxia signature, which 

are iCAF features (5,53), were enriched in PSCs co-cultured with KPCSmad4-KO organoids (Fig. 

4N). These results were validated by upregulation of phospho-STAT3 (p-STAT3) and hypoxia-

inducible factor 1-alpha (HIF-1) levels in PSCs cultured with conditioned media (CM), which 

induces the iCAF phenotype (4,5), from KPCSmad4-KO relative to KPCSmad4-WT organoids (Fig. 4O).  

 



 22 

Thus, Smad4 deletion in KRASG12D p53-mutant PDAC malignant cells induces an inflammatory 

phenotype in CAFs. 

 

Smad4 loss upregulates IL-1 and JAK/STAT signaling in KPC PDAC 

To identify mediators of malignant cell-fibroblast crosstalk in Smad4-deleted KRASG12D PDAC, 

we evaluated bi-directional signaling in PDAC organoid/PSC co-cultures. NicheNet analysis 

pinpointed Il1a as the top malignant cell-produced mediator of malignant cell-CAF crosstalk in 

KPCSmad4-KO/PSC co-cultures (Fig. 5A; Supplementary Fig. 5A; Supplementary Table S7) 

(42). Indeed, IL-1 levels were upregulated in KPCSmad4-KO organoids compared to KPCSmad4-WT 

controls (Fig. 5B; Supplementary Fig. 5B). Corroborating previous work showing that IL-1 

signaling drives the iCAF phenotype (5), IL-1 neutralization impaired the iCAF phenotype 

(Supplementary Fig. 5C). Thus, these data strongly suggest that increased IL-1 expression 

following Smad4 loss in PDAC malignant cells enhances iCAF generation.  

 

Of note, Il1a levels in organoids were further upregulated in KPCSmad4-KO/PSC co-cultures 

compared to monocultures, suggesting a positive feedback loop in the presence of an iCAF-rich 

environment (Fig. 5B). Indeed, Il1a also appeared to be the top mediator of both CAF-CAF 

crosstalk and CAF-malignant cell crosstalk in KPCSmad4-KO/PSC co-cultures (Fig. 5C-D; 

Supplementary Fig. 5D-E). Furthermore, Tnf scored as the top mediator of malignant cell-

malignant cell crosstalk in KPCSmad4-KO/PSC co-cultures and monocultures (Supplementary 

Fig. 5F-I). In agreement with these results, Tnf expression was upregulated upon SMAD4 loss 

in PDAC cells and was not further upregulated in co-culture with PSCs (Supplementary Fig. 

5J-K). 

 

In line with these findings, IL-1 and NF-B signaling were enriched in KPCSmad4-KO relative to 

KPCSmad4-WT monocultures and further enhanced when co-cultured with PSCs (Fig. 5E). 

Additionally, proliferation-associated pathways were upregulated in KPCSmad4-KO organoids, in 

line with their higher proliferation rate compared to KPCSmad4-WT controls (Fig. 5E and 2B; 

Supplementary Fig. 2F). JAK/STAT signaling and p-STAT3 levels were also upregulated in 

KPCSmad4-KO (Fig. 5E-F). Finally, gene signatures associated with NF-B and JAK/STAT 



 23 

signaling were upregulated in KPCSmad4-KO malignant cells and whole tumors in PDAC mouse 

models (Fig. 5G-I; Supplementary Fig. 5L-M; Supplementary Table S8).  

 

Our analyses show that SMAD4 loss in murine KRASG12D p53-mutant PDAC malignant cells 

enhances JAK/STAT activation and upregulates IL-1, promoting iCAF formation.  

 

SMAD4-independent TGF- signaling pathways are upregulated in Smad4-deleted PDAC 

malignant cells 

We previously found that when TGF- signaling is reduced in PDAC CAFs, IL-1 signaling 

upregulates Lif and JAK/STAT activation, which drive iCAF formation (5). Moreover, JAK/STAT 

activation boosts IL-1 signaling in iCAFs, while JAK/STAT inhibition downregulates Il1a 

expression. On the contrary, addition of TGF- to iCAFs downregulates the expression of IL1R1 

and IL-1, as well as JAK/STAT activation, which are instead boosted by TGFBR1 inhibition (5). 

We thus investigated whether similar mechanisms occur in PDAC malignant cells upon SMAD4 

loss and impairment of TGF- signaling. Contrary to what was seen in iCAFs, treatment with the 

JAK inhibitor (JAKi) AZD1480 did not downregulate Il1a expression in KPC organoids 

(Supplementary Fig. 5N). Moreover, addition and neutralization of IL-1 had no impact on 

JAK/STAT activation in KPC organoids (Fig. 5J-K). These results suggest that IL-1 and 

JAK/STAT signaling are activated by different mechanisms in iCAFs and KPCSmad4-KO malignant 

cells. In line with this, treatment with TGF- upregulated the expression of Il1a, Il1r1 and Lif in 

KPC organoids (Fig. 5L). Moreover, Tgfb1 and phospho-SMAD2/3 (p-SMAD2/3) levels were 

upregulated in KPCSmad4-KO organoids compared to KPCSmad4-WT controls (Fig. 5J-M; 

Supplementary Fig. 5O). Altogether, these data indicated that TGF- upregulation may boost 

SMAD4-independent TGF- signaling in Smad4-deleted PDAC malignant cells, leading to Il1a 

upregulation. 

 

A recent study showed that activation of the p38 MAPK pathway (54) can be upstream Il1a 

expression in PDAC (55). However, while phospho-p38 (p-p38) levels were higher in KPCSmad4-

KO organoids compared to KPCSmad4-WT controls, p38 inhibition did not impair STAT3 activation 

or Il1a expression in either organoid line (Supplementary Fig. 5P-R). Thus, to identify candidate 

pathways activated in KPCSmad4-KO organoids, we performed transcription factor (TF) analysis, 



 24 

which confirmed JAK/STAT and NF-B signaling as main pathways active in KPCSmad4-KO 

organoids (Supplementary Fig. 5S; Supplementary Table S9). This analysis also suggested 

increased activation of MEK/ERK MAPK signaling, and its downstream target Jun (56), in 

KPCSmad4-KO organoids (Supplementary Fig. 5S). Since JAK/STAT and MAPK signaling have 

been described as SMAD4-independent TGF- pathways (57), we hypothesized that the 

increased TGF- production by KPCSmad4-KO organoids may be responsible for their activation. 

In line with this, addition of TGF- increased both Erk1/2 (i.e. p44/42) and STAT3 activation in 

KPCSmad4-WT organoids (Fig. 5N). Finally, while MEK inhibition did not affect STAT3 activation, 

it significantly downregulated Il1a levels (Fig. 5O; Supplementary Fig. 5T). This suggests that 

increased MEK/ERK signaling in KPCSmad4-KO organoids boosts IL-1 production.  

 

Therefore, SMAD4 loss in murine KRASG12D p53-mutant PDAC cells leads to activation of 

SMAD4-independent TGF- pathways that impact malignant signaling and the TME. 

 

Smad4 loss tunes TME crosstalk and signaling dependencies in PDAC with distinct KRAS 

status  

The KRASG12V mutation also occurs frequently in human PDAC (23). Therefore, we assessed 

whether Smad4 loss would impact KRASG12V PDAC similarly to that observed in KRASG12D 

PDAC. To do this, we used PDAC organoids generated from the KrasFRT-LSL-G12V-FRT/+; Trp53LSL-

R172H; Pdx1-Cre; Rosa26-FlpOERT2 (hereafter, KvPC) mouse model (31) and generated isogenic 

Smad4 KO (KvPCSmad4-KO) and Smad4 WT (KvPCSmad4-WT) organoids (Supplementary Fig.  

S6A-D).  

 

Allografted KvPCSmad4-KO organoids formed tumors that grew significantly faster (Fig. 6A-B; 

Supplementary Fig. 6E). KvPCSmad4-KO PDAC also contained more neutrophils and fewer 

macrophages than KvPCSmad4-WT tumors of similar size (Fig. 6C-F; Supplementary Fig. 6F-N; 

Supplementary Table S10). Furthermore, certain malignant cell-stromal interactions were 

similarly altered in both KRAS mutant-driven models upon SMAD4 loss. These included the 

potential role of malignant cell-produced TNF in shaping macrophage and neutrophil 

composition and of macrophage-produced CXCL1 and CXCL2 in recruiting neutrophils in 

KvPCSmad4-KO PDAC (Fig. 6G-K; Supplementary Fig. 6O-S; Supplementary Table S11). IL-1 



 25 

signaling mediated malignant cell-fibroblast crosstalk also in KvPCSmad4-KO PDAC (Fig. 7A-C; 

Supplementary Fig. 7A-H). Accordingly, Il1a and Il1b expression were upregulated in 

KvPCSmad4-KO malignant cells (Fig. 7D-E). Furthermore, scRNA-seq profiles of KvPCSmad4-KO 

PDAC-derived CAFs revealed a decrease in myCAF-associated pathways and an increase in 

iCAF-associated pathways (Fig. 7F-H; Supplementary Fig. 7I). Additionally, flow cytometry 

analysis confirmed loss of myCAFs, apCAFs and CD90+, CD49E+, CD56+ and CD105+ myCAF 

populations, as well as significant downregulation of the myCAF/iCAF ratio, in KvPCSmad4-KO 

PDAC tumors compared to WT controls (Fig. 7I-J; Supplementary Fig. 7J-K).  

 

Despite these similarities with KPC tumors, abundance of iCAFs, total CAFs and tumor-

promoting CD90- myCAFs was reduced in KvPCSmad4-KO PDAC (Fig. 7I; Supplementary Fig. 

7L-M). Moreover, the T3 neutrophil signature was not significantly upregulated in KvPCSmad4-KO 

PDAC (Fig. 7K-L; Supplementary Fig. 7N-O). Furthermore, while the expression of genes 

associated with Inflam-TAMs was upregulated in KvPCSmad4-KO PDAC, as observed in KPCSmad4-

KO tumors, genes associated with Prolif-TAMs were not clearly downregulated, and markers of 

IFN-TAMs or LA-TAMs were not clearly increased in KvPCSmad4-KO PDAC compared to WT 

controls (Fig. 7M-N; Supplementary Fig. 7P-Q). Thus, SMAD4 loss has different effects on the 

stroma composition of murine PDAC tumors with distinct KRAS mutations. Moreover, SMAD4 

loss was not sufficient to drive an increase in KvPC organoid proliferation in vitro, suggesting 

that the increase in tumor growth in vivo is mediated by changes in the TME and their crosstalk 

with malignant cells (Supplementary Fig. 7R). Finally, most strikingly, the JAK/STAT and NF-

B pathways were not upregulated in KvPCSmad4-KO malignant cells in vivo (Fig. 8A).  

 

To further explore how SMAD4 loss directly impacts KvPC malignant cells, we established co-

cultures of PSCs and KvPC organoids and analyzed both flow-sorted populations by RNA-seq 

(Fig. 8B; Supplementary Fig. 7S; Supplementary Tables S12-S13). Validating our in vivo 

observations, iCAF-associated pathways were upregulated, and myCAF-associated pathways 

were downregulated, when PSCs were co-cultured with KvPCSmad4-KO organoids, albeit with 

some differences compared to KPCSmad4-KO co-cultures (Fig. 8C; Supplementary Fig. 8A). Of 

note, as observed in vivo, JAK/STAT signaling and p-STAT3 levels were not upregulated in 

KvPC malignant cells following SMAD4 loss (Fig. 8D-E). Additionally, neither Tgfb1 nor p-



 26 

SMAD2/3 levels were upregulated in KvPCSmad4-KO organoids compared to KvPCSmad4-WT 

controls (Fig. 8E; Supplementary Fig. 8B). 

 

Since JAK/STAT signaling was upregulated upon Smad4 deletion in KPC, but not KvPC, 

malignant cells, we evaluated whether inhibition of this pathway led to an increased therapeutic 

sensitivity only in KPCSmad4-KO PDAC. No significant difference was observed in the proliferation 

of KvPCSmad4-KO organoids compared to KvPCSmad4-WT controls when exposed to the JAKi (5,58) 

(Fig. 8F-G). In contrast, KPCSmad4-KO organoids were more sensitive to JAK inhibition than 

KPCSmad4-WT controls (Fig. 8H-I). To test if this difference in JAK signaling dependency also 

exists in vivo, we treated mice harboring KPCSmad4-KO or KPCSmad4-WT PDACs for two weeks with 

the JAKi (Fig. 8J; Supplementary Fig. 8C-E). Remarkably, JAK inhibition impaired tumor 

growth and diaphragm metastases of KPCSmad4-KO, but not KPCSmad4-WT, models (Fig. 8K-L; 

Supplementary Fig. 8F-G). Moreover, levels of the apoptotic marker cleaved caspase 3 (CC3) 

were upregulated in JAKi-treated KPCSmad-KO tumors (Fig. 8M-N; Supplementary Fig. 8H-K). 

This suggests a selective increase in cell death upon JAK inhibition in KRASG12D p53-mutant 

PDAC with SMAD4 loss. In line with this, live singlets were reduced in JAKi-treated KPCSmad-KO 

tumors (Supplementary Fig. 8L). Moreover, although the myCAF/iCAF ratio was not 

significantly altered, collagen deposition was increased in KPCSmad4-KO PDAC following 

JAK/STAT inhibition, suggesting a change in CAF composition towards a more myofibroblastic 

phenotype (Supplementary Fig. 8M-P). Additionally, in JAKi-treated KPCSmad4-KO PDAC, 

macrophages, apCAFs and FOXP3+ T regulatory T cells were reduced, and non-exhausted total 

T cells and CD8+ T cells were increased (59), which may contribute to impairing tumor growth 

(Supplementary Fig. 8N and 8Q-U). 

 

Together, these analyses suggest that SMAD4 loss differently impacts malignant-stromal 

crosstalk and therapeutic sensitivities in murine PDAC with distinct KRAS status. 

 

DISCUSSION 

Cancers are rarely driven by single mutations. Distinct combinations of mutations can have both 

malignant cell intrinsic and extrinsic effects. Understanding how different combinations of 

mutations cooperate to drive the malignant phenotype is key if we aim to develop more effective 



 27 

cancer therapies. Here, we show how SMAD4 loss in PDAC malignant cells shapes tumor 

biology differently in the presence of two distinct KRAS mutations (Fig. 8O). While loss of 

SMAD4 generated a more fibro-inflammatory TME in both KPC and KvPC murine PDAC, it also 

led to distinct differences depending on the KRAS status. In addition, SMAD4 loss was 

associated with an increase in JAK/STAT dependency in KPC, but not KvPC, PDAC compared 

to Smad4 WT controls. Impaired tumor growth of JAK inhibitor-treated KPCSmad4-KO tumors was 

associated with an increase in apoptosis, as well as CD8+ T cell abundance. Considering recent 

studies demonstrating enhanced efficacy of JAK/STAT inhibition with immunotherapy in cancer 

(60,61), Smad4-deleted PDAC may show increased sensitivity to this combination regimen. 

SMAD4 and KRAS mutation-specific differences in JAK/STAT dependency may start to explain 

why JAK inhibitors showed promise in pre-clinical studies but failed to display benefit when used 

to treat patients with genetically heterogeneous PDAC (5,62–64). Although these clinical trials 

were done in the presence of chemotherapy, and such combinations could be explored in the 

future, our data suggest that better understanding of how PDAC genetics impact malignant-

stromal crosstalk is needed if we are to deploy more effective targeted therapies in the clinic. Of 

note, MAPK signaling was also upregulated in KPCSmad4-KO organoids, and Smad4-deficient 

PDAC cells have been shown to be more susceptible than Smad4 WT cells to MEK inhibition 

(65).  

 

Our study also highlights the power of comparing isogenic murine organoid-derived models to 

begin to deconvolute the complexity of human PDAC and pinpoint the impact of combinations 

of mutations on malignant cells and TME. Further modeling of the genetic complexity observed 

in patients will enable to better understand its impact on bi-directional malignant-stromal 

crosstalk, PDAC progression and therapeutic vulnerabilities. Due to the heterogeneous nature 

of the TME, scRNA-seq studies coupled with genetic information will be key in future efforts to 

more effectively target the biology of distinct groups of PDAC.  

 

Upregulation of STAT3 signaling in the epithelial compartment of KPC GEMMs was also 

observed following TGFBR2 loss and was associated with increased tissue tension and collagen 

fiber changes (28,52,66). Here, we found increased STAT3 activation in both malignant cells 

and the iCAF-rich microenvironment associated with KPCSmad4-KO, but not KvPCSmad4-KO, PDAC 



 28 

compared to Smad4 WT tumors. Furthermore, contrary to TGFBR2 loss (28), SMAD4 deletion 

in KPC and KvPC PDAC drove primary tumor growth in vivo. Accordingly, among mutations 

associated with impaired TGF- signaling, only SMAD4 loss was associated with worse survival 

of PDAC patients (52). Of note, our pre-clinical study suggested that growth kinetics of 

established KPCSmad4-KO and KPCSmad4-WT tumors is comparable.  

 

Evidence suggests that inflammatory CAFs play tumor-promoting and immunosuppressive roles 

in cancer (2,5,67,68). Therefore, since iCAFs increase in Smad4-deleted PDAC, it is tempting 

to speculate that they may be involved in the aggressive phenotype of these tumors. However, 

this remains to be assessed. Similarly, whether the decrease in apCAFs or increase in 

metastasis-promoting CD90- EGFR-activated myCAFs (11) participate in driving the aggressive 

phenotype of KPCSmad4-KO PDAC remains to be determined. Fibroblast-specific GEMMs will be 

required to address these questions. Finally, the reduction in NK cells, which typically play anti-

tumor roles (69,70), as well as changes in macrophage and neutrophil phenotype and 

abundance may also contribute to the faster progression of Smad4-deleted PDAC. Indeed, 

macrophages contribute to PDAC development and therapies targeting neutrophils in PDAC 

mouse models hindered tumor progression (71,72).  

 

As KRAS mutant-specific and pan-RAS inhibitors are being developed (73–75), understanding 

differences between KRAS mutations, and their associated TMEs, could be key to design 

effective combinatorial strategies. While our analyses revealed similar effects caused by SMAD4 

loss in the context of KPC and KvPC tumors, they also highlighted important differences in 

malignant cell features and stroma composition. Whether these are at least partially dependent 

on intrinsic differences of distinct KRAS mutations remains to be determined. Similarly, further 

investigation will be pivotal to determine the roles of other mutations in shaping the stroma and 

therapy response of PDAC.  

 

AUTHORS’ CONTRIBUTIONS 

E.G.L.: conceptualization, investigation, writing-review and editing. S.P.T, J.S.M., M.J., J.A.H, 

W.L., W.K.L, S.H., P.S.W.C., G.M., M.Z., P.M.J., S.A., L.V. and R.B.: investigation. R.B. V.C. 



 29 

shared human PDAC tissues. M.V.: supervision. G.B.: conceptualization, supervision, designed 

and conducted the experiments, writing-original draft. 

 

ACKNOWLEDGEMENTS  

The authors thank the core facilities at the Cancer Research UK Cambridge Institute (CRUK-

CI). This work was supported by a UKRI Future Leaders Fellowship of which G.B. is recipient 

and that supports W.L. This work was supported by a CRUK institutional grant, which supported 

J.S.M., S.M. and S.P.T., PCRF and US DoD PCARP grants, which supported G.M. and J.S.M., 

a NCI-CRUK grant, which supported M.J., W.K.L. and S.H., and a PCUK grant, which supported 

P.S.W.C. E.G.L. and J.A.H. were supported by a MRC-DTP and a Cambridge Trust studentship, 

respectively. The authors thank Professor Tuveson for sharing PDAC organoids. 

 

REFERENCES 

1.  Siegel RL, Giaquinto AN, Jemal A. Cancer statistics, 2024. CA: A Cancer Journal for 
Clinicians. 2024;74:12–49.  

2.  Biffi G, Tuveson DA. Diversity and Biology of Cancer-Associated Fibroblasts. Physiol Rev. 
2021;101:147–76.  

3.  Elyada E, Bolisetty M, Laise P, Flynn WF, Courtois ET, Burkhart RA, et al. Cross-species 
single-cell analysis of pancreatic ductal adenocarcinoma reveals antigen-presenting 
cancer-associated fibroblasts. Cancer Discov. 2019;  

4.  Ohlund D, Handly-Santana A, Biffi G, Elyada E, Almeida AS, Ponz-Sarvise M, et al. Distinct 
populations of inflammatory fibroblasts and myofibroblasts in pancreatic cancer. J Exp Med. 
2017;214:579–96.  

5.  Biffi G, Oni TE, Spielman B, Hao Y, Elyada E, Park Y, et al. IL1-Induced JAK/STAT Signaling 
Is Antagonized by TGFbeta to Shape CAF Heterogeneity in Pancreatic Ductal 
Adenocarcinoma. Cancer Discov. 2019;9:282–301.  

6.  Hutton C, Heider F, Blanco-Gomez A, Banyard A, Kononov A, Zhang X, et al. Single-cell 
analysis defines a pancreatic fibroblast lineage that supports anti-tumor immunity. Cancer 
Cell. 2021;  

7.  Dominguez CX, Muller S, Keerthivasan S, Koeppen H, Hung J, Gierke S, et al. Single-Cell 
RNA Sequencing Reveals Stromal Evolution into LRRC15(+) Myofibroblasts as a 
Determinant of Patient Response to Cancer Immunotherapy. Cancer Discov. 2020;10:232–
53.  



 30 

8.  McAndrews KM, Chen Y, Darpolor JK, Zheng X, Yang S, Carstens JL, et al. Identification of 
Functional Heterogeneity of Carcinoma-Associated Fibroblasts with Distinct IL6-Mediated 
Therapy Resistance in Pancreatic Cancer. Cancer Discov. 2022;12:1580–97.  

9.  Helms EJ, Berry MW, Chaw RC, DuFort CC, Sun D, Onate MK, et al. Mesenchymal Lineage 
Heterogeneity Underlies Nonredundant Functions of Pancreatic Cancer-Associated 
Fibroblasts. Cancer Discov. 2022;12:484–501.  

10.  Huang H, Wang Z, Zhang Y, Pradhan RN, Ganguly D, Chandra R, et al. Mesothelial cell-
derived antigen-presenting cancer-associated fibroblasts induce expansion of regulatory T 
cells in pancreatic cancer. Cancer Cell. 2022;40:656-673.e7.  

11.  Mucciolo G, Araos Henríquez J, Jihad M, Pinto Teles S, Manansala JS, Li W, et al. EGFR-
activated myofibroblasts promote metastasis of pancreatic cancer. Cancer Cell [Internet]. 
2023 [cited 2023 Dec 28]; Available from: 
https://www.sciencedirect.com/science/article/pii/S1535610823004300 

12.  Ma R-Y, Black A, Qian B-Z. Macrophage diversity in cancer revisited in the era of single-
cell omics. Trends Immunol. 2022;43:546–63.  

13.  Ng MSF, Kwok I, Tan L, Shi C, Cerezo-Wallis D, Tan Y, et al. Deterministic reprogramming 
of neutrophils within tumors. Science. American Association for the Advancement of 
Science; 2024;383:eadf6493.  

14.  Lloyd EG, Henríquez JA, Biffi G. Modelling the micro- and macro- environment of pancreatic 
cancer: from patients to pre-clinical models and back. Disease Models & Mechanisms. 
2024;17:dmm050624.  

15.  Collisson EA, Sadanandam A, Olson P, Gibb WJ, Truitt M, Gu S, et al. Subtypes of 
pancreatic ductal adenocarcinoma and their differing responses to therapy. Nature 
Medicine. Nature Publishing Group; 2011;17:500–3.  

16.  Moffitt RA, Marayati R, Flate EL, Volmar KE, Loeza SGH, Hoadley KA, et al. Virtual 
microdissection identifies distinct tumor- and stroma-specific subtypes of pancreatic ductal 
adenocarcinoma. Nature Genetics. Nature Publishing Group; 2015;47:1168–78.  

17.  Bailey P, Chang DK, Nones K, Johns AL, Patch AM, Gingras MC, et al. Genomic analyses 
identify molecular subtypes of pancreatic cancer. Nature. Nature Publishing Group; 
2016;531:47–52.  

18.  Somerville TD, Biffi G, Daßler-Plenker J, Hur SK, He X-Y, Vance KE, et al. Squamous trans-
differentiation of pancreatic cancer cells promotes stromal inflammation. Sawyers CL, Ojala 
PM, Oliver T, editors. eLife. eLife Sciences Publications, Ltd; 2020;9:e53381.  

19.  Raghavan S, Winter PS, Navia AW, Williams HL, DenAdel A, Lowder KE, et al. 
Microenvironment drives cell state, plasticity, and drug response in pancreatic cancer. Cell. 
2021;184:6119-6137.e26.  



 31 

20.  Chan-Seng-Yue M, Kim JC, Wilson GW, Ng K, Figueroa EF, O’Kane GM, et al. 
Transcription phenotypes of pancreatic cancer are driven by genomic events during tumor 
evolution. Nat Genet. Nature Publishing Group; 2020;52:231–40.  

21.  Williams HL, Dias Costa A, Zhang J, Raghavan S, Winter PS, Kapner KS, et al. Spatially-
resolved single-cell assessment of pancreatic cancer expression subtypes reveals co-
expressor phenotypes and extensive intra-tumoral heterogeneity. Cancer Research. 
2022;CAN-22-3050.  

22.  Aguirre AJ, Nowak JA, Camarda ND, Moffitt RA, Ghazani AA, Hazar-Rethinam M, et al. 
Real-time genomic characterization of advanced pancreatic cancer to enable precision 
medicine. Cancer Discovery. American Association for Cancer Research Inc.; 
2018;8:1096–111.  

23.  Raphael BJ, Hruban RH, Aguirre AJ, Moffitt RA, Yeh JJ, Stewart C, et al. Integrated 
Genomic Characterization of Pancreatic Ductal Adenocarcinoma. Cancer Cell. Cell Press; 
2017;32:185-203.e13.  

24.  Hingorani SR, Wang L, Multani AS, Combs C, Deramaudt TB, Hruban RH, et al. 
Trp53R172H and KrasG12D cooperate to promote chromosomal instability and widely 
metastatic pancreatic ductal adenocarcinoma in mice. Cancer Cell. 2005;7:469–83.  

25.  Maddalena M, Mallel G, Nataraj NB, Shreberk-Shaked M, Hassin O, Mukherjee S, et al. 
TP53 missense mutations in PDAC are associated with enhanced fibrosis and an 
immunosuppressive microenvironment. Proc Natl Acad Sci U S A. 2021;118:e2025631118.  

26.  Shaashua L, Ben-Shmuel A, Pevsner-Fischer M, Friedman G, Levi-Galibov O, Nandakumar 
S, et al. BRCA mutational status shapes the stromal microenvironment of pancreatic cancer 
linking clusterin expression in cancer associated fibroblasts with HSF1 signaling. Nat 
Commun. Nature Publishing Group; 2022;13:6513.  

27.  Vennin C, Mélénec P, Rouet R, Nobis M, Cazet AS, Murphy KJ, et al. CAF hierarchy driven 
by pancreatic cancer cell p53-status creates a pro-metastatic and chemoresistant 
environment via perlecan. Nat Commun. Nature Publishing Group; 2019;10:3637.  

28.  Laklai H, Miroshnikova YA, Pickup MW, Collisson EA, Kim GE, Barrett AS, et al. Genotype 
tunes pancreatic ductal adenocarcinoma tissue tension to induce matricellular fibrosis and 
tumor progression. Nature Medicine. Nature Publishing Group; 2016;22:497–505.  

29.  Iacobuzio-Donahue CA, Fu B, Yachida S, Luo M, Abe H, Henderson CM, et al. DPC4 Gene 
Status of the Primary Carcinoma Correlates With Patterns of Failure in Patients With 
Pancreatic Cancer. JCO. Wolters Kluwer; 2009;27:1806–13.  

30.  Malinova A, Schreyer D, Fiorini E, Pasini D, Bevere M, D’Agosto S, et al. ecDNA 
amplification of MYC drives intratumor copy-number heterogeneity and adaptation to stress 
in PDAC. BioRxiv. 2023;  



 32 

31.  Humpton TJ, Alagesan B, DeNicola GM, Lu D, Yordanov GN, Leonhardt CS, et al. 
Oncogenic KRAS Induces NIX-Mediated Mitophagy to Promote Pancreatic Cancer. Cancer 
Discovery. 2019;9:1268–87.  

32.  Oni TE, Biffi G, Baker LA, Hao Y, Tonelli C, Somerville TDD, et al. SOAT1 promotes 
mevalonate pathway dependency in pancreatic cancer. J Exp Med. 2020;217.  

33.  Filippini D, Agosto SD, Delfino P, Simbolo M, Piro G, Rusev B, et al. Immunoevolution of 
mouse pancreatic organoid isografts from preinvasive to metastatic disease. Sci Rep. 
Nature Publishing Group; 2019;9:12286.  

34.  Vassilev LT, Vu BT, Graves B, Carvajal D, Podlaski F, Filipovic Z, et al. In Vivo Activation 
of the p53 Pathway by Small-Molecule Antagonists of MDM2. Science. American 
Association for the Advancement of Science; 2004;303:844–8.  

35.  Tiriac H, Belleau P, Engle DD, Plenker D, Deschênes A, Somerville TDD, et al. Organoid 
profiling identifies common responders to chemotherapy in pancreatic cancer. Cancer 
Discov. 2018;8:1112–29.  

36.  Boj SF, Hwang C-I, Baker LA, Chio IIC, Engle DD, Corbo V, et al. Organoid models of 
human and mouse ductal pancreatic cancer. Cell. 2015;160:324–38.  

37.  Patro R, Duggal G, Love MI, Irizarry RA, Kingsford C. Salmon provides fast and bias-aware 
quantification of transcript expression. Nat Methods. Nature Publishing Group; 
2017;14:417–9.  

38.  Soneson C, Love MI, Robinson MD. Differential analyses for RNA-seq: transcript-level 
estimates improve gene-level inferences. F1000Res. 2015;4:1521.  

39.  Love MI, Huber W, Anders S. Moderated estimation of fold change and dispersion for RNA-
seq data with DESeq2. Genome Biology. 2014;15:550.  

40.  Zhu A, Ibrahim JG, Love MI. Heavy-tailed prior distributions for sequence count data: 
removing the noise and preserving large differences. Bioinformatics. Oxford University 
Press; 2019;35:2084.  

41.  Subramanian A, Tamayo P, Mootha VK, Mukherjee S, Ebert BL, Gillette MA, et al. Gene 
set enrichment analysis: A knowledge-based approach for interpreting genome-wide 
expression profiles. Proceedings of the National Academy of Sciences. Proceedings of the 
National Academy of Sciences; 2005;102:15545–50.  

42.  Browaeys R, Saelens W, Saeys Y. NicheNet: modeling intercellular communication by 
linking ligands to target genes. Nat Methods. Nature Publishing Group; 2020;17:159–62.  

43.  Müller-Dott S, Tsirvouli E, Vazquez M, Ramirez Flores RO, Badia-I-Mompel P, Fallegger R, 
et al. Expanding the coverage of regulons from high-confidence prior knowledge for 
accurate estimation of transcription factor activities. Nucleic Acids Res. 2023;51:10934–49.  



 33 

44.  Badia-I-Mompel P, Vélez Santiago J, Braunger J, Geiss C, Dimitrov D, Müller-Dott S, et al. 
decoupleR: ensemble of computational methods to infer biological activities from omics 
data. Bioinform Adv. 2022;2:vbac016.  

45.  Zheng GXY, Terry JM, Belgrader P, Ryvkin P, Bent ZW, Wilson R, et al. Massively parallel 
digital transcriptional profiling of single cells. Nat Commun. 2017;8:14049.  

46.  Bernstein NJ, Fong NL, Lam I, Roy MA, Hendrickson DG, Kelley DR. Solo: Doublet 
Identification in Single-Cell RNA-Seq via Semi-Supervised Deep Learning. Cell Syst. 
2020;11:95-101.e5.  

47.  Gayoso A, Lopez R, Xing G, Boyeau P, Valiollah Pour Amiri V, Hong J, et al. A Python 
library for probabilistic analysis of single-cell omics data. Nat Biotechnol. 2022;40:163–6.  

48.  Wolf FA, Angerer P, Theis FJ. SCANPY: large-scale single-cell gene expression data 
analysis. Genome Biology. 2018;19:15.  

49.  Dann E, Henderson NC, Teichmann SA, Morgan MD, Marioni JC. Differential abundance 
testing on single-cell data using k-nearest neighbor graphs. Nat Biotechnol. 2022;40:245–
53.  

50.  Jin S, Guerrero-Juarez CF, Zhang L, Chang I, Ramos R, Kuan C-H, et al. Inference and 
analysis of cell-cell communication using CellChat. Nat Commun. Nature Publishing Group; 
2021;12:1088.  

51.  Finak G, McDavid A, Yajima M, Deng J, Gersuk V, Shalek AK, et al. MAST: a flexible 
statistical framework for assessing transcriptional changes and characterizing 
heterogeneity in single-cell RNA sequencing data. Genome Biology. 2015;16:278.  

52.  Blackford A, Serrano OK, Wolfgang CL, Parmigiani G, Jones S, Zhang X, et al. SMAD4 
gene mutations are associated with poor prognosis in pancreatic cancer. Clin Cancer Res. 
2009;15:4674–9.  

53.  Schwoerer S, Cimino FV, Ros M, Tsanov KM, Ng C, Lowe SW, et al. Hypoxia potentiates 
the inflammatory fibroblast phenotype promoted by pancreatic cancer cell-derived 
cytokines. Cancer Research. 2023;CAN-22-2316.  

54.  Luo K. Signaling Cross Talk between TGF-β/Smad and Other Signaling Pathways. Cold 
Spring Harb Perspect Biol. 2017;9:a022137.  

55.  Singh SP, Dosch AR, Mehra S, De Castro Silva I, Bianchi A, Garrido VT, et al. Tumor Cell-
Intrinsic p38 MAPK Signaling Promotes IL1α-Mediated Stromal Inflammation and 
Therapeutic Resistance in Pancreatic Cancer. Cancer Res. 2024;84:1320–32.  

56.  Whitmarsh AJ, Davis RJ. Transcription factor AP-1 regulation by mitogen-activated protein 
kinase signal transduction pathways. J Mol Med (Berl). 1996;74:589–607.  



 34 

57.  Tang L-Y, Heller M, Meng Z, Yu L-R, Tang Y, Zhou M, et al. Transforming Growth Factor-
β (TGF-β) Directly Activates the JAK1-STAT3 Axis to Induce Hepatic Fibrosis in 
Coordination with the SMAD Pathway. J Biol Chem. 2017;292:4302–12.  

58.  Hedvat M, Huszar D, Herrmann A, Gozgit JM, Schroeder A, Sheehy A, et al. The JAK2 
Inhibitor, AZD1480, Potently Blocks Stat3 Signaling and Oncogenesis in Solid Tumors. 
Cancer Cell. 2009;16:487–97.  

59.  Steele NG, Carpenter ES, Kemp SB, Sirihorachai VR, The S, Delrosario L, et al. Multimodal 
Mapping of the Tumor and Peripheral Blood Immune Landscape in Human Pancreatic 
Cancer. Nat Cancer. 2020;1:1097–112.  

60.  Zak J, Pratumchai I, Marro BS, Marquardt KL, Zavareh RB, Lairson LL, et al. JAK inhibition 
enhances checkpoint blockade immunotherapy in patients with Hodgkin lymphoma. 
Science. 2024;384:eade8520.  

61.  Mathew D, Marmarelis ME, Foley C, Bauml JM, Ye D, Ghinnagow R, et al. Combined JAK 
inhibition and PD-1 immunotherapy for non-small cell lung cancer patients. Science. 
2024;384:eadf1329.  

62.  Hurwitz H, Van Cutsem E, Bendell J, Hidalgo M, Li C-P, Salvo MG, et al. Ruxolitinib + 
capecitabine in advanced/metastatic pancreatic cancer after disease 
progression/intolerance to first-line therapy: JANUS 1 and 2 randomized phase III studies. 
Invest New Drugs. 2018;36:683–95.  

63.  Wörmann SM, Song L, Ai J, Diakopoulos KN, Kurkowski MU, Görgülü K, et al. Loss of P53 
Function Activates JAK2-STAT3 Signaling to Promote Pancreatic Tumor Growth, Stroma 
Modification, and Gemcitabine Resistance in Mice and Is Associated With Patient Survival. 
Gastroenterology. 2016;151:180-193.e12.  

64.  Nagathihalli NS, Castellanos JA, Shi C, Beesetty Y, Reyzer ML, Caprioli R, et al. Signal 
Transducer and Activator of Transcription 3, Mediated Remodeling of the Tumor 
Microenvironment Results in Enhanced Tumor Drug Delivery in a Mouse Model of 
Pancreatic Cancer. Gastroenterology. 2015;149:1932-1943.e9.  

65.  Hasselluhn MC, Schlösser D, Versemann L, Schmidt GE, Ulisse M, Oschwald J, et al. An 
NFATc1/SMAD3/cJUN Complex Restricted to SMAD4-Deficient Pancreatic Cancer Guides 
Rational Therapies. Gastroenterology. 2024;166:298-312.e14.  

66.  Yachida S, White CM, Naito Y, Zhong Y, Brosnan JA, Macgregor-Das AM, et al. Clinical 
significance of the genetic landscape of pancreatic cancer and implications for identification 
of potential long-term survivors. Clin Cancer Res. 2012;18:6339–47.  

67.  Nicolas AM, Pesic M, Engel E, Ziegler PK, Diefenhardt M, Kennel KB, et al. Inflammatory 
fibroblasts mediate resistance to neoadjuvant therapy in rectal cancer. Cancer Cell. 
Elsevier; 2022;40:168-184.e13.  



 35 

68.  Erez N, Truitt M, Olson P, Arron ST, Hanahan D. Cancer-Associated Fibroblasts Are 
Activated in Incipient Neoplasia to Orchestrate Tumor-Promoting Inflammation in an NF-
kappaB-Dependent Manner. Cancer Cell. 2010;17:135–47.  

69.  Koh E-K, Lee H-R, Son W-C, Park G-Y, Kim J, Bae J-H, et al. Combinatorial immunotherapy 
with gemcitabine and ex vivo-expanded NK cells induces anti-tumor effects in pancreatic 
cancer. Sci Rep. 2023;13:7656.  

70.  Gürlevik E, Fleischmann-Mundt B, Brooks J, Demir IE, Steiger K, Ribback S, et al. 
Administration of Gemcitabine After Pancreatic Tumor Resection in Mice Induces an 
Antitumor Immune Response Mediated by Natural Killer Cells. Gastroenterology. 
2016;151:338-350.e7.  

71.  Nielsen SR, Strøbech JE, Horton ER, Jackstadt R, Laitala A, Bravo MC, et al. Suppression 
of tumor-associated neutrophils by lorlatinib attenuates pancreatic cancer growth and 
improves treatment with immune checkpoint blockade. Nat Commun. Nature Publishing 
Group; 2021;12:3414.  

72.  Caronni N, La Terza F, Vittoria FM, Barbiera G, Mezzanzanica L, Cuzzola V, et al. IL-1β+ 
macrophages fuel pathogenic inflammation in pancreatic cancer. Nature. Nature Publishing 
Group; 2023;623:415–22.  

73.  Kemp SB, Cheng N, Markosyan N, Sor R, Kim I-K, Hallin J, et al. Efficacy of a Small-
Molecule Inhibitor of KrasG12D in Immunocompetent Models of Pancreatic Cancer. Cancer 
Discovery. 2023;13:298–311.  

74.  Mahadevan KK, McAndrews KM, LeBleu VS, Yang S, Lyu H, Li B, et al. KRASG12D 
inhibition reprograms the microenvironment of early and advanced pancreatic cancer to 
promote FAS-mediated killing by CD8+ T cells. Cancer Cell. Elsevier; 2023;41:1606-
1620.e8.  

75.  Wasko UN, Jiang J, Dalton TC, Curiel-Garcia A, Edwards AC, Wang Y, et al. Tumor-
selective activity of RAS-GTP inhibition in pancreatic cancer. Nature. Nature Publishing 
Group; 2024;1–3.  

  

  



 36 

Figure 1. SMAD4-deficient human organoid-derived PDAC tumors have less fibrosis than 

SMAD4 WT tumors. (A) Schematic of analyses of orthotopically-grafted human organoid-

derived pancreatic ductal adenocarcinoma (PDAC) models in NOD SCID gamma (NSG) mice. 

U/S, ultrasound-based imaging. (B-C) Volumes, measured by U/S, of tumors derived from the 

transplantation of KP (58 days post-transplant) (B) or KvP (64 days post-transplant) (C) SMAD4 

wild-type (WT) or SMAD4-deficient human PDAC organoids with KRASG12D or KRASG12V 

mutation, respectively. (D) Representative SMAD4, Masson’s trichrome and alpha smooth 

muscle actin (SMA) stains in SMAD4 WT or SMAD4-deficient human organoid-derived KP 

PDAC. Scale bars, 50 m. (E-F) Quantification of Masson’s trichrome (E) and SMA (F) stains 

in SMAD4 WT or SMAD4-deficient human organoid-derived PDAC.  

 

For C, E and F, results show mean  SEM. *, P < 0.05, **, P < 0.01; ***, P < 0.001, Mann-

Whitney test. 

 

Figure 2. Smad4 loss impacts the immune TME in KPC PDAC. (A) Schematic of analyses of 

KPC (i.e. KrasG12D p53-mutant) orthotopically-grafted organoid-derived PDAC models in nu/nu 

(left) or C57BL/6J (right) mice. (B) Proliferation curves of KPCSmad4-WT or KPCSmad4-KO (KO3 and 

KO4) PDAC organoids cultured for 120 hours in reduced media (i.e., 5% FBS DMEM). Data 

were normalized to the first measurement (3 hours post-plating). Results show mean  SEM of 

n=2 biological replicates (with n=4 technical replicates each). ***, P < 0.001, Mann-Whitney test 

calculated for the last time point. (C) Volumes of tumors derived from transplantation of 

KPCSmad4-WT or KPCSmad4-KO organoids in C57BL/6J mice measured by U/S. Results show mean 

 SEM from 2 separate experiments, each with 1 WT group and 2 groups of KO pools from 2 

different guides (32 days (n= 4 WT and n= 8 KO) or 21 days (n= 4 WT and n= 6 KO) post-

transplant). ***, P < 0.001, Mann-Whitney test. (D) Volumes of tumors derived from 

transplantation of KPCSmad4-WT or KPCSmad4-KO PDAC organoids in nu/nu mice measured by U/S. 

Results show mean  SEM from 2 separate experiments, each with 1 WT group and 2 groups 

of KO pools from 2 different guides (25 days (n= 4 WT and n= 7 KO) or day 21 days (n= 4 WT 

and n= 6 KO) post-transplant). *, P < 0.05, Mann-Whitney test. (E-F) Uniform manifold 

approximation and projection (UMAP) plot shows cell clusters from KPCSmad4-WT (n=3) or 

KPCSmad4-KO (n=6) tumors analyzed by single-cell RNA-sequencing (scRNA-seq), color-coded 



 37 

by (E) genotype or (F) cell type. (G) Upregulated and downregulated differentially expressed 

genes (DEGs) in each cell type identified by pseudobulk analysis from scRNA-seq of KPCSmad4-

WT or KPCSmad4-KO tumors. False discovery rate (FDR) < 0.05. (H) Cell type contribution in 

KPCSmad4-WT or KPCSmad4-KO tumors. (I) Violin plots showing the distribution of groups of nearest 

neighbor cells from different cell type clusters upon the log-fold change between KPCSmad4-KO vs 

KPCSmad4-WT conditions computed with Milo. The malignant cell cluster was divided by cells from 

Smad4 WT or Smad4 KO tumors to clarify the directionality of abundance. (J-K) Representative 

flow plots of (J) neutrophils (CD45+CD11b+Gr1+) and (K) macrophages (CD45+Gr1-

CD11b+F4/80+) from KPCSmad4-WT or KPCSmad4-KO tumors in C57BL/6J mice. (L-M) Flow 

cytometric analysis of neutrophils (CD45+CD11b+Gr1+) and macrophages (CD45+Gr1-

CD11b+F4/80+) from live singlets in KPCSmad4-WT or KPCSmad4-KO tumors in (L) nu/nu or (M) 

C57BL/6J mice. Results show mean  SEM from 3 separate experiments, each with 1 WT group 

and 2 groups of KO pools from 2 different guides. **, P < 0.01; ***, P < 0.001, Mann-Whitney 

test. (N-O) UMAP plot of neutrophils from KPCSmad4-WT or KPCSmad4-KO tumors analyzed by 

scRNA-seq, color-coded by (N) genotype or (O) sub-cluster. (P) Sub-cluster contribution in 

neutrophils of KPCSmad4-WT or KPCSmad4-KO tumors. (Q) Gene set enrichment analysis (GSEA) of 

T3 neutrophil signature in neutrophils from KPCSmad4-KO PDAC compared to KPCSmad4-WT PDAC. 

The signature from Ng et al (13) is significantly upregulated. NES, normalized enrichment score. 

(R-S) UMAP plot of macrophages from KPCSmad4-WT or KPCSmad4-KO tumors analyzed by scRNA-

seq, color-coded by (R) genotype or (S) sub-cluster. RTM-TAM, resident-tissue macrophage-

like tumor-associated macrophage (TAM); LA-TAM, lipid-associated TAM; Angio-TAM, pro-

angiogenic TAM; Inflam-TAM, inflammatory cytokine-enriched TAM; Prolif-TAM, proliferating 

TAM; IFN-TAM, interferon-primed TAM. Macrophage subtypes are from Ma et al (12). (T) Sub-

cluster contribution in macrophages of KPCSmad4-WT or KPCSmad4-KO tumors. (U) Dot plot 

visualization of scaled average expression of macrophage subtype markers in macrophages 

from KPCSmad4-WT or KPCSmad4-KO tumors analyzed by scRNA-seq. Color intensity represents 

expression level and dot size represents the percentage of expressing cells. (V) Selected 

significantly upregulated (i.e. NES > 1.50 and FDR < 0.25) and downregulated (i.e. NES < -1.50 

and FDR < 0.25) pathways identified by GSEA of macrophages from KPCSmad4-KO compared to 

KPCSmad4-WT tumors, assessed by pseudobulk analysis from the scRNA-seq dataset.  

 



 38 

Figure 3. Smad4 loss impacts malignant-stromal crosstalk in KPC PDAC. (A) Cell-cell 

communication analysis using CellChat showing the number of connections lost (blue) or gained 

(red) between malignant cells, fibroblasts, macrophages and neutrophils in KPCSmad4-KO (n=6) 

compared to KPCSmad4-WT (n=3) tumors, as assessed by scRNA-seq. (B) Selected ligand-

receptor interactions and their strength based on CellChat analysis between malignant cells, 

fibroblasts, macrophages and neutrophils in KPCSmad4-KO tumors compared to KPCSmad4-WT 

tumors. Commun. Prob., communication probability. (C-D) Selected pathways with significantly 

different connections between malignant cells, fibroblasts, macrophages, and neutrophils in 

KPCSmad4-KO tumors compared to KPCSmad4-WT tumors. (E-F) Ligand-target heatmaps show top 

selected ligands of (E) malignant cells inferred to regulate target genes in macrophages and (F) 

macrophages inferred to regulate target genes in malignant cells in KPCSmad4-KO PDAC, as 

assessed by NicheNet analysis of scRNA-seq. (G-H) Ligand activity plots show the top ligands 

of (G) macrophages or (H) neutrophils inferred to regulate target genes in fibroblasts in 

KPCSmad4-KO PDAC, as assessed by NicheNet analysis. AUPR, area under the precision-recall 

curve. (I) Ligand-target heatmap shows top selected ligands of malignant cells inferred to 

regulate target genes in fibroblasts in KPCSmad4-KO PDAC, as assessed by NicheNet analysis. 

(J) Ligand activity plot shows the top ligands of malignant cells inferred to regulate target genes 

in fibroblasts in KPCSmad4-KO PDAC, as assessed by NicheNet analysis. (K) Ligand activity plot 

shows the top ligands of fibroblasts inferred to regulate target genes in malignant cells in 

KPCSmad4-KO PDAC, as assessed by NicheNet analysis. (L) Ligand-target heatmap shows top 

selected ligands of fibroblasts inferred to regulate target genes in malignant cells in KPCSmad4-KO 

PDAC, as assessed by NicheNet analysis.  

 

Figure 4 Smad4 loss drives a fibro-inflammatory stroma in KPC PDAC. (A-B) UMAP plots 

showing the cell cluster of cancer-associated fibroblasts (CAFs) from KPCSmad4-WT (n=3) or 

KPCSmad4-KO (n=6) PDAC analyzed by scRNA-seq, color-coded by sub-cluster (A) or genotype 

(B). (C) Pie charts showing proportions of different CAF clusters from KPCSmad4-WT or KPCSmad4-

KO tumors. P < 0.01, chi-square test. (D) Representative flow plots of Ly6C-MHCII- myCAFs, 

Ly6C+MHCII- iCAFs and Ly6C-MHCII+ apCAFs from KPCSmad4-WT or KPCSmad4-KO tumors in 

C57BL/6J mice. (E-F) Flow cytometric analyses of (E) myCAFs (Ly6C-MHCII- CAFs), iCAFs 

(Ly6C+MHCII- CAFs) and apCAFs (Ly6C-MHCII+ CAFs) and (F) myCAF/iCAF ratio from live 



 39 

singlets in KPCSmad4-WT or KPCSmad4-KO tumors in C57BL/6J mice. Results show mean  SEM 

from 3 separate experiments, each with 1 WT group and 2 groups of KO pools from 2 different 

guides. *, P < 0.05, **, P < 0.01, ***, P < 0.001, Mann-Whitney test. (G) Selected significantly 

upregulated (i.e. NES > 1.50, FDR < 0.25; apart for the IL6 JAK STAT3 signaling HALLMARK 

with NES=1.47) and downregulated (i.e. NES < -1.50, FDR < 0.25) pathways identified by GSEA 

of CAFs from KPCSmad4-KO compared to KPCSmad4-WT tumors, as assessed by pseudobulk 

analysis from scRNA-seq. The in vivo iCAF signature is from Elyada et al (3). (H) Dot plot 

visualization of the scaled average expression of myCAF-enriched markers in CAFs from 

KPCSmad4-WT or KPCSmad4-KO tumors, as analyzed by scRNA-seq. Color intensity represents 

expression level and dot size represents the percentage of expressing cells. (I) Flow cytometric 

analyses of CD90+, CD49E+, CD56+ and CD105+ CAFs from live singlets in KPCSmad4-WT or 

KPCSmad4-KO tumors in C57BL/6J mice. Results show mean  SEM from 3 separate experiments, 

each with 1 WT group and 2 groups of KO pools from 2 different guides. ***, P < 0.001, Mann-

Whitney test. (J-K) UMAP plots showing the malignant cell cluster from KPCSmad4-WT or 

KPCSmad4-KO tumors analyzed by scRNA-seq (J). Different genotypes are color-coded (K). (L) 

Dot plot visualization of the scaled average expression of Smad4, Il1a and Il1b in malignant cells 

from KPCSmad4-WT or KPCSmad4-KO tumors, as analyzed by scRNA-seq. Color intensity represents 

expression level and dot size represents the percentage of expressing cells. (M) Schematic of 

flow-sorting strategy of pancreatic stellate cells (PSCs) and KPCSmad4-WT or KPCSmad4-KO PDAC 

organoids from monocultures or co-cultures for RNA-sequencing (RNA-seq). (N) Selected 

significantly upregulated and downregulated pathways identified by GSEA of PSCs cultured with 

KPCSmad4-KO organoids (n=10) compared to PSCs cultured with KPCSmad4-WT organoids (n=5). 

The in vitro iCAF and myCAF signatures are from Öhlund et al (4). The TGF--induced myCAF 

in vitro signature is from Mucciolo and Araos Henríquez et al (11). The LRRC15+ CAF signature 

is from Dominguez et al (7). (O) Western blot analysis of phospho-STAT3 (p-STAT3), STAT3 

and hypoxia inducible factor 1 alpha (HIF-1) in PSCs cultured for 4 days in PDAC organoid 

conditioned media (CM) from KPCSmad4-WT or KPCSmad4-KO organoids. ACTIN, loading controls.  

 

Figure 5. Smad4 loss upregulates IL-1 and JAK/STAT signaling in KPC PDAC. (A) Ligand 

activity plot shows the top ligands of KPCSmad4-KO PDAC organoids inferred to regulate target 

genes in co-cultured PSCs, as assessed by NicheNet analysis of RNA-seq. (B) RNA-seq 



 40 

expression of Il1a in KPCSmad4-WT or KPCSmad4-KO organoids flow-sorted from monocultures or 

co-cultures with PSCs. Results show mean  SEM. *, P < 0.05, **, P < 0.01, Mann-Whitney test. 

(C-D) Ligand activity plots show the top ligands of PSCs in co-culture with KPCSmad4-KO organoids 

inferred to regulate target genes in (C) PSCs or (D) co-cultured organoids, as assessed by 

NicheNet analysis. (E) Selected GSEA pathways significantly enriched or depleted in KPCSmad4-

KO compared to KPCSmad4-WT malignant cells flow-sorted from monocultures or co-cultures with 

PSCs. The smooth pattern indicates the monocultures. Inflammatory pathways are in orange. 

Proliferation-associated pathways are in brown. Cocx, co-culture; monocx, monoculture. (F) 

Western blot analysis of p-STAT3 and STAT3 in KPCSmad4-WT or KPCSmad4-KO organoids cultured 

in reduced media for 2 days. ACTIN, loading control. (G) Selected significantly upregulated or 

downregulated pathways identified by GSEA of malignant cells from KPCSmad4-KO tumors (n=6) 

compared to malignant cells from KPCSmad4-WT tumors (n=3), as assessed by pseudobulk 

analysis of scRNA-seq. Inflammatory pathways are in orange. (H-I) UMAP plots of malignant 

cells from KPCSmad4-WT or KPCSmad4-KO tumors colored by the NES of the HALLMARK IL6 JAK 

STAT3 signaling (H) or the KEGG NF-B signaling (I) pathways, as analyzed by scRNA-seq. (J-

K) Western blot analysis of p-STAT3, STAT3, phospho-SMAD2/3 (p-SMAD2/3) and SMAD2 in 

KPCSmad4-WT or KPCSmad4-KO organoids cultured for 3 days in reduced media with (J) 5 g/mL 

anti-IL1 or isotype control, or (K) 10 ng/mL IL1-. ACTIN, loading controls. (L) qPCR analysis 

of Il1a, Lif, Il1r1, and Tgfb1 in KPCSmad4-WT or KPCSmad4-KO organoids cultured for 2 days in 

reduced media with or without 5-20 ng/mL TGF- in the presence or absence of 2 M TGFBR1 

inhibitor A83-01 (TGFBRi). Results show mean  SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001, 

paired and unpaired Student’s t test. (M) RNA-seq expression of Tgfb1 in KPCSmad4-WT or 

KPCSmad4-KO organoids flow-sorted from monocultures or co-cultures with PSCs. Results show 

mean  SEM. *, P < 0.05; **, P < 0.01, Mann-Whitney test. (N) Western blot analysis of p-STAT3, 

STAT3, phospho-p44/42 (p-p44/42) and p44/42 in KPCSmad4-WT or KPCSmad4-KO organoids 

cultured for 2 days in reduced media with 5 ng/mL TGF-. HSP60, loading control. (O) qPCR 

analysis of Il1a, Tgfb1, Lif, and Dusp6 (i.e. a MAPK MEK/ERK target) in KPCSmad4-WT or 

KPCSmad4-KO organoids cultured for 3 days in reduced media with DMSO or 1-2 nM MEK inhibitor 

trametinib (MEKi). Results show mean  SEM. *, P < 0.05; **, P < 0.01; ***, P < 0.001, paired 

and unpaired Student’s t test.  

 



 41 

Figure 6. Smad4 loss impacts the immune TME and malignant-stromal crosstalk in KvPC 

PDAC. (A) Schematic of analyses of KvPC (i.e. KrasG12V p53-mutant) orthotopically-grafted 

organoid-derived PDAC models in nu/nu mice. (B) Volumes of tumors derived from the 

transplantation of Smad4 WT or Smad4 KO KvPC (i.e. KrasG12V p53 mutant) PDAC organoids, 

measured by U/S. Results show mean  SEM from 3 separate experiments, each with 1 WT 

group and 2 groups of KO pools from 2 different guides (29 days (cohort 1), 30 days (cohort 2) 

or 65 days (cohort 3) post-transplant). ***, P < 0.001, Mann-Whitney test. (C-D) UMAP plot of all 

cell types from KvPCSmad4-WT (n=4) or KvPCSmad4-KO (n=4) PDAC tumors analyzed by scRNA-

seq, color-coded by genotype (C) or cell type clusters (D). (E) Cell type contribution in KvPCSmad4-

WT or KvPCSmad4-KO tumors. (F) Flow cytometric analysis of neutrophils (CD45+CD11b+Gr1+) and 

macrophages (CD45+Gr1-CD11b+F4/80+) from live singlets in KvPCSmad4-WT or KvPCSmad4-KO 

tumors. Results show mean  SEM from 3 separate experiments, each with 1 WT group and 2 

groups of KO pools from 2 different guides. ***, P < 0.001, Mann-Whitney test. (G) Selected 

ligand-receptor interactions and their strength based on CellChat analysis between malignant 

cells, fibroblasts, macrophages, and neutrophils in KvPCSmad4-KO compared to KvPCSmad4-WT 

tumors. (H-I) Selected pathways with significantly different connections between malignant cells, 

fibroblasts, macrophages, and neutrophils in KvPCSmad4-KO compared to KvPCSmad4-WT tumors. 

(J-K) Ligand-target heatmaps show top selected ligands of malignant cells inferred to target 

genes in macrophages (J) and neutrophils (K) in KvPCSmad4-KO PDAC, as assessed by NicheNet 

analysis of scRNA-seq.  

 

Figure 7. Smad4 loss drives a fibro-inflammatory stroma in KvPC PDAC. (A) Ligand activity 

plot shows the top ligands of malignant cells inferred to regulate target genes in fibroblasts in 

KvPCSmad4-KO PDAC, as assessed by NicheNet analysis of scRNA-seq. (B-C) Ligand-target 

heatmaps show top selected ligands of (B) malignant cells inferred to regulate target genes in 

fibroblasts and (C) fibroblasts inferred to regulate target genes in malignant cells in KPCSmad4-KO 

PDAC, as assessed by NicheNet analysis. (D) UMAP plot of malignant cells from KvPCSmad4-WT 

(n=4) or KvPCSmad4-KO (n=4) tumors analyzed by scRNA-seq, color-coded by genotype. (E) Dot 

plot visualization of the scaled average expression of Smad4, Il1a and Il1b in malignant cells of 

KvPCSmad4-WT or KvPCSmad4-KO tumors, as analyzed by scRNA-seq. Color intensity represents 

expression level and dot size represents the percentage of expressing cells. (F-G) UMAP plots 



 42 

of CAFs from KvPCSmad4-WT or KvPCSmad4-KO tumors analyzed by scRNA-seq, color-coded by 

genotype (F) or sub-cluster (G). (H) Selected significantly upregulated (i.e. NES > 1.50 and FDR 

< 0.25; apart from the NF-kappa B signaling pathway with NES = 1.49) and downregulated (i.e. 

NES < -1.50 and FDR < 0.25; apart from the fatty acid metabolism with NES = -1.48) pathways 

identified by GSEA of CAFs from KvPCSmad4-KO compared to KvPCSmad4-WT tumors, as assessed 

by pseudobulk analysis from scRNA-seq. The in vivo iCAF signature is from Elyada et al (3). 

The in vitro iCAF signature is from Öhlund et al (4). The TGF--induced myCAF in vitro signature 

is from Mucciolo and Araos Henríquez et al (11). The cCAF3 signature is from McAndrews et al 

(8). (I-J) Flow cytometric analyses of (I) myCAFs (Ly6C-MHCII- CAFs), iCAFs (Ly6C+MHCII- 

CAFs) and apCAFs (Ly6C-MHCII+ CAFs), and (J) CD90+, CD49E+, CD56+ and CD105+ CAFs 

from live singlets in KvPCSmad4-WT or KvPCSmad4-KO tumors. Results show mean  SEM from 3 

separate experiments, each with 1 WT group and 2 groups of KO pools from 2 different guides. 

*, P < 0.05; **, P < 0.01; ***, P < 0.001, Mann-Whitney test. (K) UMAP plot of neutrophils from 

KvPCSmad4-WT or KvPCSmad4-KO PDAC analyzed by scRNA-seq. Different sub-clusters are color-

coded. (L) GSEA of T3 neutrophil signature in neutrophils from KvPCSmad4-KO compared to 

KvPCSmad4-WT PDAC. The signature from Ng et al (13) is not significantly altered. (M) UMAP plot 

of macrophages from KvPCSmad4-WT or KvPCSmad4-KO tumors analyzed by scRNA-seq. Different 

sub-clusters are color-coded. (N) Dot plot visualization of the scaled average expression of 

macrophage markers in macrophages from KvPCSmad4-WT or KvPCSmad4-KO tumors analyzed by 

scRNA-seq. Color intensity represents expression level and dot size represents the percentage 

of expressing cells. 

 

Figure 8. Smad4 loss tunes signaling dependencies in PDAC with distinct KRAS status. 

(A) Selected significantly upregulated or downregulated pathways identified by GSEA of 

malignant cells from KvPCSmad4-KO PDAC (n=4) compared to malignant cells from KvPCSmad4-WT 

tumors (n=4), as assessed by pseudobulk analysis of the scRNA-seq dataset. (B) Schematic of 

flow-sorting strategy of PSCs and KvPCSmad4-WT or KvPCSmad4-KO PDAC organoids from 

monocultures or co-cultures for RNA-seq analysis. (C) Selected significantly upregulated and 

downregulated pathways identified by GSEA of PSCs cultured with KvPCSmad4-KO organoids 

(n=8) compared to PSCs cultured with KvPCSmad4-WT organoids (n=4). The in vivo iCAF and 

myCAF signatures are from Elyada et al (3). The in vitro iCAF and myCAF signatures are from 



 43 

Öhlund et al (4). The TGF--induced myCAF in vitro signature is from Mucciolo and Araos 

Henríquez et al (11). The LRRC15+ CAF and cCAF3 signatures are from Dominguez et al (7) 

and McAndrews et al (8), respectively. (D) Selected pathways found significantly enriched or 

depleted by GSEA in KvPCSmad4-KO malignant cells compared to KvPCSmad4-WT malignant cells 

flow-sorted from monocultures or co-cultures with PSCs. The smooth pattern indicates the 

monocultures. Inflammatory pathways are highlighted in orange. (E) Western blot analysis of p-

STAT3, STAT3, p-SMAD2/3 and SMAD2 in KvPCSmad4-WT or KvPCSmad4-KO organoids cultured in 

reduced media for 2 days. ACTIN, loading control. (F) Western blot analysis of p-STAT3 and 

STAT3 in KvPCSmad4-WT or KvPCSmad4-KO organoids cultured for 2 days in reduced media with or 

without 8 M of the JAK inhibitor (JAKi, AZD1480). ACTIN, loading control. (G) Proliferation 

curves of KvPCSmad4-WT and KvPCSmad4-KO organoids cultured for 144 hours in reduced media 

with or without 8 M JAKi. Data were normalized to the first measurement (3 hours post-plating) 

and to the DMSO control. Results show mean  SEM of n=3 biological replicates (with n=6 

technical replicates each). No statistical difference was found by Mann-Whitney test. (H) 

Western blot analysis of p-STAT3 and STAT3 in KPCSmad4-WT or KPCSmad4-KO organoids cultured 

for 2 days in reduced media with or without 8 M of the JAKi. ACTIN, loading control. (I) 

Proliferation curves of KPCSmad4-WT and KPCSmad4-KO organoids cultured for 144 hours in reduced 

media with or without 8 M JAKi. Data were normalized to the first measurement (3 hours post-

plating) and to the DMSO control. Results show mean  SEM of n=3 biological replicates (with 

n=6 technical replicates each). ***, P < 0.001, Mann-Whitney test. (J) Schematic of experimental 

design and downstream analyses of a 2-week JAKi treatment of KPCSmad4-WT and KPCSmad4-KO 

organoid-derived PDAC models in C57BL/6J mice. (K) Tumor growth (i.e. ratio of tumor volume 

at day 14 and tumor volume at day -1), measured by U/S, of 2-week vehicle- and JAKi- treated 

KPCSmad4-WT or KPCSmad4-KO PDAC tumors. Results show mean  SEM from 3 separate 

experiments. ***, P < 0.001, Mann-Whitney test. (L) Percentage of KPCSmad4-WT and KPCSmad4-

KO tumor-bearing mice with metastases in the liver, lungs and diaphragm following 2 weeks of 

treatment with vehicle or JAKi. Results show data from 3 experiments. **, P < 0.01, ***, P < 

0.001, chi-square test. (M) Representative cleaved caspase 3 (CC3) immunohistochemistry 

(IHC) stains in 2-week vehicle- or JAKi- treated KPCSmad4-WT and KPCSmad4-KO PDAC tumors. 

Scale bars, 50 m. (N) Quantification of CC3 stain in 2-week vehicle- or JAKi- treated KPCSmad4-

WT and KPCSmad4-KO PDAC tumors. Results show mean  SEM from 3 separate experiments. **, 



 44 

P < 0.01; ***, P < 0.001, Mann-Whitney test. (O) Model illustrating how Smad4 loss differently 

shapes murine KRASG12D (a) or KRASG12V (b) p53 mutant PDAC tumors, including their 

progression, tumor microenvironment and therapeutic sensitivity. 



Figure 1

A

NSG

Tumor weight
Histology

B
human PDAC organoids

PDAC tumor

Orthotopic surgery

C

0

100

200

300

400

500

600

Vo
lu

m
e 

(m
m

3 )

Tumor volume 

SMAD4-deficient KP hT105 (n=5)
SMAD4 WT KP hT48 (n=5)

0

100

200

300

400

Vo
lu

m
e 

(m
m

3 )

Tumor volume

SMAD4-deficient KvP hF24 (n=8)
SMAD4 WT KvP hT1 (n=8)

F
D Masson’s trichromeαSMA IHC SMAD4 IHC

Human organoid-derived PDAC tumors

E

KP and KvP SMAD4-deficient (n=25)

αSMA stain quantification

KP and KvPP SMAD4 WT (n=10)

Human organoid-derived PDAC tumors

Human organoid-derived PDAC tumors

**
***

Tumor volume
by U/S

KP and KvP SMAD4-deficient (n=25)

Masson’s trichrome quantification

KP and KvPP SMAD4 WT (n=10)

***

S
M

A
D

4 
W

T 
(h

T4
8)

S
M

A
D

4-
de

fic
ie

nt
 (h

T1
05

)

0

20

40

60

80

%
 C

ol
la

ge
n 

ar
ea

 / 
to

ta
l a

re
a

0

5

10

15

20

25

%
 P

os
iti

ve
 a

re
a 

/ t
ot

al
 a

re
a *



C

G

A KPCSmad4-WT or KPCSmad4-KO organoids

nu/nu

Histology & immunohistochemistry
Tumor weight

Flow cytometry
Single-cell RNA-sequencing

PDAC tumor

Orthotopic surgery

C57BL/6J

Tumor weight
Flow cytometry

RNA-sequencing
Immunohistochemistry

PDAC tumor

Orthotopic surgery

Cell type clusters

S

M
al

ig
na

nt
 c

el
l

Fi
br

ob
la

st
M

ac
ro

ph
ag

e
N

eu
tro

ph
il

E
nd

ot
he

lia
l c

el
l

D
uc

ta
l c

el
l 1

Pe
ric

yt
e

M
as

t c
el

l
A

ci
na

r c
el

l
D

en
dr

iti
c 

ce
ll

D
uc

ta
l c

el
l 2

DownregulatedUpregulated

N
um

be
r o

f D
E

G
s

B
 c

el
l

2000

1000

1000

2000

0

V

-3 -2 -1 0 1 2 3

G2M checkpoints HALLMARK
E2F targets HALLMARK

Mitotic spindle HALLMARK
ROS pathway HALLMARK

TNFA signaling via NFκB HALLMARK
Adipogenesis HALLMARK

Fatty acid metabolism HALLMARK
Angiogenesis HALLMARK

Interferon gamma response HALLMARK
Interferon alpha response HALLMARK

Hypoxia HALLMARK
Oxidative phosphorylation HALLMARK

NES

KPCSmad4-KO vs KPCSmad4-WT - macrophage cluster

Figure 2

scRNA-seq - KPCSmad4-WT (n=3) and KPCSmad4-KO (n=6) PDAC tumors

0

200

400

600

Vo
lu

m
e 

(m
m

3 )

Tumor Volume

*

KPCSmad4-WT (n=8)
KPCSmad4-KO (n=13)

nu/nu

Macrophage cluster

Macrophage cluster

Ly
ve

1
Fo

lr2
Ap

oe
Ac

p5
M

sr
1

M
rc

1
Ar

g1
C

cl
3

C
xc

l1
C

xc
l2

Sp
p1

Ve
gf

a
Vc

an
Th

bs
1

M
ki

67
C

dk
1

C
d2

74
H

2-
Aa

H
2-

Ab
1

Irf
1

St
at

1
N

be
a

In
sy

n2
b

B3
ga

lt2

U

Macrophage clusterT

Tumor volume
by U/S

Tumor volume
by U/S

KPCSmad4-WT or KPCSmad4-KO organoids

E Cell type clusters Cell type clustersF

Tumor volume

0

200

400

600

Vo
lu

m
e 

(m
m

3 )

***

 KPCSmad4-KO (n=14)
 KPCSmad4-WT (n=8)

C57BL/6J

D

R

Neutrophils Macrophages
0

100

200

300

400

%
 li

ve
 s

in
gl

et
s

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l

***

**

KPCSmad4-WT (n=15)
KPCSmad4-KO (n=29)

nu/nu

M

Neutrophils Macrophages
0

100

200

300

%
 li

ve
 s

in
gl

et
s

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l ***

***

KPCSmad4-WT (n=15)
KPCSmad4-KO (n=30)

C57BL/6J

L

Q

P Neutrophil clusterO Neutrophil clusterN Neutrophil cluster

Macrophage cluster

0.0

0.2

0.4

0.6

−20
−10

0
10

4000 8000 12000

scRNA-seq - KPCSmad4-WT and KPCSmad4-KO PDAC tumors

Neutrophil cluster - T3 signature scRNA-seq - KPCSmad4-WT and KPCSmad4-KO PDAC tumors

KPCSmad4-KO vs KPCSmad4-WT PDAC

NES 1.57
FDR 0.11

I Cell type cluster abundance (KO vs WT)

J

G
R

-1

KPCSmad4-WT (C57BL/6J)     KPCSmad4-KO (C57BL/6J)

CD11b

Neutrophils Neutrophils

B

0 24 48 72 96 120
0

5

10

15

Time (hours)

Fo
ld

 c
ha

ng
e 

of
 a

re
a 

***
***

PDAC organoid proliferation

K

F4/80

FS
C

-A Macrophages Macrophages

KPCSmad4-WT (C57BL/6J)       KPCSmad4-KO (C57BL/6J)

R
an

ke
d 

Li
st

 M
et

ric

Rank in Ordered Dataset
16000

R
un

ni
ng

 E
nr

ic
hm

en
t S

co
re

 KPCSmad4-WT WT 2 (n=2)
 KPCSmad4-KO KO 3 (n=2)
 KPCSmad4-KO KO 4 (n=2)

H Cell type clusters



C CXCL pathway

C
xcr2

Cxcl1
Cxcl2Cxcl2

Cxcl
2

Cxcl1
Cxcl2
Cxcl5

CXCL pathway

Figure 3

Cxcl1  − Cxcr2
Cxcl2  − Cxcr2
Cxcl5  − Cxcr2

Tnf  − Tnfrsf1a
Tnf  − Tnfrsf1b

min

max

B CellChat analysis - scRNA-seq KPCSmad4-KO compared to KPCSmad4-WT

p−value

p < 0.01

Commun. Prob.Tgfb1 − (Tgfbr1+Tgfbr2)

Fibr
ob

las
t −

> F
ibr

ob
las

t (W
T)

Fibr
ob

las
t −

> F
ibr

ob
las

t (K
O)

Fibr
ob

las
t −

> N
eu

tro
ph

il (
W

T)

Fibr
ob

las
t −

> N
eu

tro
ph

il (
KO)

Fibr
ob

las
t −

> M
ac

rop
ha

ge
 (W

T)

Fibr
ob

las
t −

> M
ac

rop
ha

ge
 (K

O)

Fibr
ob

las
t −

> M
ali

gn
an

t c
ell

 (W
T)

Fibr
ob

las
t −

> M
ali

gn
an

t c
ell

 (K
O)

Neu
tro

ph
il −

> F
ibr

ob
las

t (W
T)

Neu
tro

ph
il −

> F
ibr

ob
las

t (K
O)

Neu
tro

ph
il −

> N
eu

tro
ph

il (
W

T)

Neu
tro

ph
il −

> N
eu

tro
ph

il (
KO)

Neu
tro

ph
il −

> M
ac

rop
ha

ge
 (W

T)

Neu
tro

ph
il −

> M
ac

rop
ha

ge
 (K

O)

Neu
tro

ph
il −

> M
ali

gn
an

t c
ell

 (W
T)

Neu
tro

ph
il −

> M
ali

gn
an

t c
ell

 (K
O)

Mac
rop

ha
ge

 −>
 Fibr

ob
las

t (W
T)

Mac
rop

ha
ge

 −>
 Fibr

ob
las

t (K
O)

Mac
rop

ha
ge

 −>
 N

eu
tro

ph
il (

W
T)

Mac
rop

ha
ge

 −>
 N

eu
tro

ph
il (

KO)

Mac
rop

ha
ge

 −>
 M

ac
rop

ha
ge

 (W
T)

Mac
rop

ha
ge

 −>
 M

ac
rop

ha
ge

 (K
O)

Mac
rop

ha
ge

 −>
 M

ali
gn

an
t c

ell
 (W

T)

Mac
rop

ha
ge

 −>
 M

ali
gn

an
t c

ell
 (K

O)

Mali
gn

an
t c

ell
 −>

 Fibr
ob

las
t (W

T)

Mali
gn

an
t c

ell
 −>

 Fibr
ob

las
t (K

O)

Mali
gn

an
t c

ell
 −>

 N
eu

tro
ph

il (
W

T)

Mali
gn

an
t c

ell
 −>

 N
eu

tro
ph

il (
KO)

Mali
gn

an
t c

ell
 −>

 M
ac

rop
ha

ge
 (W

T)

Mali
gn

an
t c

ell
 −>

 M
ac

rop
ha

ge
 (K

O)

Mali
gn

an
t c

ell
 −>

 M
ali

gn
an

t c
ell

 (W
T)

Mali
gn

an
t c

ell
 −>

 M
ali

gn
an

t c
ell

 (K
O)

D

Fibroblast
Neutrophil
Macrophage
Malignant cell

TNF pathway

Tnfrsf1a

Tnf

Tn
frs

f1aTn
frs

f1
b

Tn
f

Tnfrsf1a

KPCSmad4-KO

KPCSmad4-WT

−126

1

71

−2

−7

−138

−141

−1

−27

−1

6

−116

5

−4
Fibroblast

Neutrophil

Macrophage

Malignant cell

A CellChat analysis - scRNA-seq
KPCSmad4-KO compared to KPCSmad4-WT

Cx
cl2

Cxcl1Cxcl5
Cxcl2
Ppbp

C
xcr2

Cxcr4

Cxcl1

Cxcl12

Cxcl2

Cxcl2

Cxcl5

KPCSmad4-KO

Fibroblast
Neutrophil
Macrophage
Malignant cell

E

F

0.03
0.04

K

Tgfa
Tnfsf9
Bace2
Calca

Cxcl12
S100a8

Fgf1
Cd38

Angptl4
C3

Cd55
S100a4
Rarres1

Cd14
Hp

Jam2
Il1rn
Il1b

Adm
Tnf

Ligand Activity Plot
From Fibroblasts 
to Malignant cells

(KPCSmad4-KO)

P
rio

rit
iz

ed
 L

ig
an

ds AUPR

0.025
0.030
0.035
0.040

J Ligand Activity Plot
From Malignant cells 

to Fibroblasts
(KPCSmad4-KO)

P
rio

rit
iz

ed
 L

ig
an

ds AUPR

Wnt10a
Wnt4

Wnt7b
Bace2

S100a8
Il18

Gas6
Tnfsf9
Wnt7a
Cd38
Sirpa
Il1rn

Muc4
Ptprc

H2-Eb1
Csf3

Il1a
Il1b

Cafs1
Tnf

H Ligand Activity Plot
From Neutrophils 

to Fibroblasts
(KPCSmad4-KO)

AUPR

0.015
0.020
0.025
0.030

Ecm1
Trf

Fam3c
Gpl1

Almp1
Slglecf
Vegfa
Cxcl1
Cd47

Lgals3
Ltf

Apob
Saraf
Gpc3

Calm3
Try4

H2-Q10
Prnp

Cd14
Il1a

P
rio

rit
iz

ed
 L

ig
an

ds

G

0.020
0.025
0.030
0.035
0.040

P
rio

rit
iz

ed
 L

ig
an

ds

Ligand Activity Plot
From Macrophages 

to Fibroblasts
(KPCSmad4-KO)

Cmtm8
Mmp12

Apob
Anxa1
Apoc2

Try4
Apoe

Col6a6
Adm

H2-D1
H2-Q4
H2-Q6
H2-Q7
H2-K1

S100a8
Ccn1

Tnfsf9
H2-Eb1

Il1a
Il6

AUPR

A
tf3

C
xc

r4
E

gr
1

Fo
s

H
2_

d1
H

2_
k1

H
2_

q4
H

2_
q6

H
2_

q7
Il6 Irf

7
M

uc
1

S
er

pi
ne

1
S

gk
1

S
oc

s3
S

ta
t1

Tx
ni

p
A

ck
r3

G
0s

2
Il1

a
M

m
p1

2
S

lc
7a

2
Tn

fa
ip

3
V

ca
m

1
A

dm
A

tp
1b

1
D

us
p1

Fb
xo

32
H

2_
m

3
M

gp
M

t1
P

lp
p3

S
10

0a
8

S
10

0a
9

Tn
frs

f2
3

C
cn

2
N

os
2

S
lc

2a
1

C
d9

B
cl

2a
1b

C
fb

Il7
r

Ifi
h1

S
pp

1

Il1b

Il1a

Tnf

Predicted target genes
P

rio
rit

iz
ed

 li
ga

nd
s

Regulatory potential
0.0 0.1 0.2

From Malignant cells to Macrophages (KPCSmad4-KO)

A
da

m
ts

1
A

dm
A

m
ot

l2
A

tp
1b

1
B

id
B

irc
2

B
tg

2
C

cl
20

C
el

sr
2

C
sf

3
C

xc
l1

0
C

xc
l2

D
us

p1
E

dn
1

E
gr

1
Fb

xo
32

G
0s

2
H

2_
m

3
H

be
gf

Ic
am

1
Il1

b
Il1

rn
Irf

1
Ju

n
Lc

n2
M

af
f

M
gp

N
fk

bi
z

P
la

t
P

la
u

P
la

ur
P

m
ai

p1
P

pp
1r

15
a

P
tg

s1
S

10
0a

8
S

10
0a

9
S

at
1

S
er

pi
nb

2
Ta

p1
Tl

r2
Tn

f
Tn

fa
ip

3
V

ca
m

1
B

2m
C

as
p4

Fo
s

G
bp

2
G

df
15

G
lrx

Ifi
35

Il1
3r

a1
Irf

9
M

uc
1

P
ls

cr
1

P
ls

cr
2

R
ab

27
a

S
oc

s3
S

ta
t1

S
ta

t3
Tn

frs
f2

1
Tx

ni
p

U
be

2l
6

X
bp

1

Il6

Il1a

Predicted target genes

P
rio

rit
iz

ed
 li

ga
nd

s

Regulatory potential
0.00 0.10 0.20

From Macrophages to Malignant cells (KPCSmad4-KO)

I

B
cl

2a
1b

B
id

C
fb

D
dx

58
G

0s
2

G
bp

2
Il1

a
Il1

b
Irf

7
S

er
pi

na
3m

S
er

pi
nb

2
Ta

p1
Tr

af
1

Tx
ni

p
A

bc
a1

C
ar

2
C

d5
9b

C
dk

n1
a

C
ts

k
D

ra
m

1
G

pn
m

b
Lg

m
n

Lx
n

M
gl

l
N

up
r1

V
w

a5
a

A
m

pd
3

C
d8

2
C

sf
3

S
er

pi
nb

8
Tn

f
Tn

frs
f1

1b
A

dm
A

tp
1b

1
H

2_
m

3
Il1

rn
Lc

n2
P

tg
s1

S
10

0a
8

S
10

0a
9

H
m

ox
1

K
lf1

0
M

xd
1

P
ls

cr
1

P
ls

cr
2

S
ta

t1
A

tf3
Fa

s
Ig

fb
p3

Tl
r3

H
p

C
cl

3
O

as
1a

G
m

70
30

Csf3
Csf1
Il1a
Il1b
Tnf

Predicted target genes

P
rio

rit
iz

ed
 li

ga
nd

s

From Malignant cells to Fibroblasts (KPCSmad4-KO)

Regulatory potential
0.0 0.1 0.2

L

B
cl

2a
1b

C
d8

3
C

fb
C

xc
l1

0
C

xc
l2

C
yl

d
D

dx
58

E
dn

1
E

gr
1

Fo
s

G
0s

2
H

be
gf

Ic
am

1
Ifn

gr
1

Il1
5

Il1
a

Il1
b

Irf
1

Irf
7

M
af

f
N

oc
t

Pa
rp

12
P

la
u

P
la

ur
P

m
ai

p1
P

nr
c1

P
pp

1r
15

a
R

el
S

er
pi

nb
2

S
ta

t5
a

Ta
p1

Tn
fa

ip
3

Tx
ni

p
V

ca
m

1
C

d8
2

C
sf

1
Ju

n
N

fk
bi

z
P

tg
er

4
S

at
1

S
m

ox
S

oc
s3

Tn
f

Tr
ib

1
A

tf3
C

cn
2

C
dk

n1
a

M
yc

S
ta

t1
A

dm
C

cn
d2

C
dh

1
D

di
t3

P
la

t
A

nx
a1

E
pa

s1
M

al
l

M
cl

1

Cxcl12

Fgf1

Tnf

Il1b

Predicted target genes

P
rio

rit
iz

ed
 li

ga
nd

s

From Fibroblasts to Malignant cells (KPCSmad4-KO)

Regulatory potential
0.0 0.1 0.2



CA

D

Figure 4

I
J

Ly
6C

MHC-II

iCAFsiCAFs

myCAFs myCAFs apCAFsapCAFs

M

Live cells
DAPI-

Organoids
PDPN-EpCAM+

KPC organoid + PSC
co-cultures 

RNA-seq
(Figure 5)

PSCs
PDPN+EpCAM-

RNA-seq

KPC organoid 
monocultures

Live cells
DAPI-

Organoids
PDPN-EpCAM+

RNA-seq
(Figure 5)

N

iCAF-associated myCAF-associated

-4 -2 0 2 4

myCAF in vitro
Smooth muscle contraction REACTOME

TGF-β-induced myCAF in vitro
LRRC15+ CAFs

NOTCH signaling HALLMARK
ECM organization REACTOME

IL6 JAK STAT3 signaling HALLMARK
Inflammatory response HALLMARK

TNFα signaling via NFκB HALLMARK
Hypoxia HALLMARK

Interleukin 1 signaling REACTOME
iCAF in vitro

NES

RNA-seq
PSCs (+ KPCSmad4-KO) vs PSCs (+ KPCSmad4-WT)

other

H

Fibroblast cluster

Malignant cell cluster Malignant cell cluster

L

scRNA-seq - KPCSmad4-WT (n=3) and KPCSmad4-KO (n=6) PDAC tumors

Fibroblast cluster

Malignant cell cluster

S
m

ad
4

Il1
a

Il1
b

WT
KO

B Fibroblast cluster

K

scRNA-seq - KPCSmad4-KO vs KPCSmad4-WT - fibroblast cluster 

p-STAT3

ACTIN

Smad4
WT

Smad4 KO
kDa
100
75

37
50

PSCs + CM

KO 3 KO 4

STAT3
100
75

KPC CM

O

CD49E+ CD56+ CD105+ CD90+
0

50

100

150

200

250

%
 li

ve
 s

in
gl

et
s

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l *** *** *** ***

CAFs

KPCSmad4-WT (n=14-15)            
KPCSmad4-KO  (n=30)

scRNA-seq - KPCSmad4-WT and KPCSmad4-KO PDAC tumors

F

G

KPCSmad4-WT        KPCSmad4-KO

Fibroblast cluster

KO

Itg
a5

 (C
D

49
E

)
Th

y1
 (C

D
90

)

N
ca

m
1 

(C
D

56
)

E
ng

 (C
D

10
5)

WTiCAF-associated
myCAF-associated
other

-4 -3 -2 -1 0 1 2 3

G2M checkpoint HALLMARK
E2F targets HALLMARK

Cell-ECM interactions REACTOME
PI3K AKT mTOR signaling HALLMARK

IL6 JAK STAT3 signaling HALLMARK
NF-kappa B signaling pathway KEGG

Inflammatory response HALLMARK
Oxydative phosphorylation HALLMARK

iCAF in vivo
Interferon gamma response HALLMARK

Interferon alpha response HALLMARK

NES

E

iCAF apCAF myCAF
0

50
100
150
200
250
300
350
400
800

1200

%
 li

ve
 s

in
gl

et
s

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l

**

*
*

KPCSmad4-WT (n=15)            
KPCSmad4-KO  (n=30)

C57BL/6J

myCAF/iCAF
0

100

200

300

400

%
 li

ve
 s

in
gl

et
s

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l ***

KPCSmad4-WT (n=15)            
KPCSmad4-KO  (n=30)

C57BL/6J

C57BL/6J

P < 0.01

HIF1α

ACTIN

100

37
50

KPCSmad4-WT (C57BL/6J)               KPCSmad4-KO (C57BL/6J)



Figure 5

E

G

-2 -1 0 1 2 3

cocx - Respiratory electron transport REACTOME
cocx - Citric acid cycle TCA cycle REACTOME
cocx - Oxidative phosphorylation HALLMARK

cocx - Complex I biogenesis REACTOME
monocx - TGF-β signaling HALLMARK

cocx - TGF-β signaling HALLMARK
monocx - IL6 JAK STAT3 signaling HALLMARK

monocx - EMT HALLMARK
monocx - Interleukin 1 signaling REACTOME

cocx Regulat cholesterol biosynthesis by SREBP REACTOME
cocx - Interleukin 1 signaling REACTOME
cocx - Inflammatory response HALLMARK

cocx - TNFα signaling via NFκB HALLMARK
cocx - IL6 JAK STAT3 signaling HALLMARK

monocx - G2M checkpoint HALLMARK
monocx - MYC targets V1 HALLMARK

monocx - E2F targets HALLMARK

NES

RNA-seq - KPCSmad4-KO vs KPCSmad4-WT organoids B

Il1a
0

1000

2000

3000

4000

5000

N
or

m
al

iz
ed

 e
xp

ec
te

d 
co

un
ts

KPCSmad4-WT cocx (n=5)
KPCSmad4-KO cocx (n=10)

KPCSmad4-WT monocx (n=4)
KPCSmad4-KO monocx (n=10)

**

**

*
KPC organoids

DC

A

Spp1
Vip

Cntf
Lif

Pcdhga7
Edn1
Pthlh

Mmp13
Il1f6

Tnfsf13
Csf3
Tnc

Tnfsf12
Csf1
Il23a

Tnfsf15
Fn1

Ccl20
Tnf
Il1a

Ligand Activity Plot
From KPCSmad4-KO 

to PSCs
P

rio
rit

iz
ed

 L
ig

an
ds AUPR

0.01
0.02
0.03

Ligand Activity Plot
From PSCs 

to KPCSmad4-KO

P
rio

rit
iz

ed
 L

ig
an

ds

a
Gdnf
Tph1

Ptgs2
Areg
Spp1

Cck
Cxcl5
Csf2
Saa1
Csf3

Ccl20
Il6

Serping1
Il33

Cxcl9
Tnfsf11

Il1a

−0.02
−0.01
0.00
0.01
0.02
0.03

AUPR

Ligand Activity Plot
From PSCs to PSCs

(in cocx with KPCSmad4-KO)

Camp
Areg

Ptgs2
Angpt1

Cxcl5
a

Serping1
Spp1

Cck
Lif

Mmp13
Csf3

Tnfsf11
Csf2
Il33

Cxcl9
Saa1

Il6
Ccl20

Il1a

AUPR

0.01
0.02
0.03

P
rio

rit
iz

ed
 L

ig
an

ds

IL6 JAK STAT3 signaling HALLMARKH KPCSmad4-KOKPCSmad4-WT
I NF-kappa B signaling pathway KEGG

scRNA-seq KPCSmad4-WT and KPCSmad4-KO - malignant cell cluster

KPCSmad4-KOKPCSmad4-WT

KPCSmad4-WT cocx (n=5)
KPCSmad4-KO cocx (n=10)

KPCSmad4-WT monocx (n=4)
KPCSmad4-KO monocx (n=10)

J K

F

ACTIN

p-STAT3
100

75

50
37

100

75
STAT3

WT KO 3 KO 4

KPC organoids
Smad4 KOSmad4 kDa

Tgfb1
0

200

400

600

800

1000

1200

N
or

m
al

iz
ed

 e
xp

ec
te

d 
co

un
ts

**
*

KPC organoids

*

L

M N O

KPCSmad4-WT + MEKi (n=5) KPCSmad4-KO + MEKi (n=8)
KPCSmad4-WT + DMSO (n=5) KPCSmad4-KO + DMSO (n=8)

Il1a Tgfb1 Lif Dusp6
0

5

10

15

20

m
R

N
A 

fo
ld

 c
ha

ng
e

*** ***

**
**

******

***
***

*

**

***
***

*

KPCSmad4-WT cocx (n=5)
KPCSmad4-KO cocx (n=10)

KPCSmad4-WT monocx (n=4)
KPCSmad4-KO monocx (n=10)

Il1a Lif Il1r1 Tgfb1
0

5

10

15

20

25
25

45

65

R
el

at
iv

e 
m

R
N

A 
ex

pr
es

si
on

KPCSmad4-WT (n=8)
KPCSmad4-WT + TGF-β (n=7)
KPCSmad4-WT +  TGF-β + TGFBRi (n=4)

KPC organoids

*** ***

**
*

***

*** **

***
***

*** **
**

***

** **
**

KPCSmad4-KO (n=16)
KPCSmad4-KO + TGF-β (n=14)
KPCSmad4-KO +  TGF-β + TGFBRi (n=8)

-3 -2 -1 0 1 2

EMT HALLMARK
TGF-β signaling HALLMARK

Cell-ECM interactions REACTOME
TGF-β signaling pathway KEGG
ECM organization REACTOME

IL6 JAK STAT3 signaling HALLMARK
NF-κB signaling pathway KEGG

NES

scRNA-seq
KPCSmad4-KO vs KPCSmad4-WT - malignant cell cluster 

kDa

ACTIN

SMAD2

STAT3

p-SMAD2/3

p-STAT3

Smad4 WT Smad4 KO
KPC organoids

anti-IL-1α - + - +
100
75
100
75

50

37

50
ACTIN

SMAD2

STAT3

p-SMAD2/3

p-STAT3

Smad4 WT Smad4 KO
KPC organoids

IL-1α - + - +
100
75
100
75

50

37

50

kDa

p44/42

STAT3

HSP60

p-STAT3

p-p44/42

WT KO

TGF-β - + -

KPC organoids
Smad4

50
50
37

100

75
100
75

50

37

kDa

Smad4
KPC organoids



Figure 6

B C

F
scRNA-seq - KvPCSmad4-WT (n=4) and KvPCSmad4-KO (n=4) PDAC tumors

A KvPCSmad4-WT or KvPCSmad4-KO organoids

nu/nu

Tumor weight
Histology

Flow cytometry
Single-cell RNA-sequencing

PDAC tumor

Orthotopic surgery

E Cell type clusters
Acinar cell
B cell
Ductal cell 1
Ductal cell 2
Endothelial cell
Fibroblast
Neutrophil
Macrophage
Malignant cell
Mast cell
Pericyte

KvPCSmad4-KO (n=26)
KvPCSmad4-WT (n=13)

Cxcl1  − Cxcr2
Cxcl2  − Cxcr2
Cxcl5  − Cxcr2

Tnf  − Tnfrsf1a
Tnf  − Tnfrsf1b

G
Tgfb1 − (Tgfbr1+Tgfbr2)

Il1b  − (Il1r1+Il1rap)
Il1a  − (Il1r1+Il1rap)

Fibr
ob

las
t −

> F
ibr

ob
las

t (W
T)

Fibr
ob

las
t −

> F
ibr

ob
las

t (K
O)

Fibr
ob

las
t −

> N
eu

tro
ph

il (
W

T)

Fibr
ob

las
t −

> N
eu

tro
ph

il (
KO)

Fibr
ob

las
t −

> M
ac

rop
ha

ge
 (W

T)

Fibr
ob

las
t −

> M
ac

rop
ha

ge
 (K

O)

Fibr
ob

las
t −

> M
ali

gn
an

t c
ell

 (W
T)

Fibr
ob

las
t −

> M
ali

gn
an

t c
ell

 (K
O)

Neu
tro

ph
il −

> F
ibr

ob
las

t (W
T)

Neu
tro

ph
il −

> F
ibr

ob
las

t (K
O)

Neu
tro

ph
il −

> N
eu

tro
ph

il (
W

T)

Neu
tro

ph
il −

> N
eu

tro
ph

il (
KO)

Neu
tro

ph
il −

> M
ac

rop
ha

ge
 (W

T)

Neu
tro

ph
il −

> M
ac

rop
ha

ge
 (K

O)

Neu
tro

ph
il −

> M
ali

gn
an

t c
ell

 (W
T)

Neu
tro

ph
il −

> M
ali

gn
an

t c
ell

 (K
O)

Mac
rop

ha
ge

 −>
 Fibr

ob
las

t (W
T)

Mac
rop

ha
ge

 −>
 Fibr

ob
las

t (K
O)

Mac
rop

ha
ge

 −>
 N

eu
tro

ph
il (

W
T)

Mac
rop

ha
ge

 −>
 N

eu
tro

ph
il (

KO)

Mac
rop

ha
ge

 −>
 M

ac
rop

ha
ge

 (W
T)

Mac
rop

ha
ge

 −>
 M

ac
rop

ha
ge

 (K
O)

Mac
rop

ha
ge

 −>
 M

ali
gn

an
t c

ell
 (W

T)

Mac
rop

ha
ge

 −>
 M

ali
gn

an
t c

ell
 (K

O)

Mali
gn

an
t c

ell
 −>

 Fibr
ob

las
t (W

T)

Mali
gn

an
t c

ell
 −>

 Fibr
ob

las
t (K

O)

Mali
gn

an
t c

ell
 −>

 N
eu

tro
ph

il (
W

T)

Mali
gn

an
t c

ell
 −>

 N
eu

tro
ph

il (
KO)

Mali
gn

an
t c

ell
 −>

 M
ac

rop
ha

ge
 (W

T)

Mali
gn

an
t c

ell
 −>

 M
ac

rop
ha

ge
 (K

O)

Mali
gn

an
t c

ell
 −>

 M
ali

gn
an

t c
ell

 (W
T)

Mali
gn

an
t c

ell
 −>

 M
ali

gn
an

t c
ell

 (K
O)

CellChat analysis - scRNA-seq -  KvPCSmad4-KO compared to  KvPCSmad4-WT 

Cell type clusters

Tumor volume
by U/S

D

CXCL pathway

Fibroblast
Neutrophil
Macrophage
Malignant cell

I

C
xcl2

C
xc

l3Cxc
l2

Cxcl1Cxcl5
Cxcl3

Ackr3

C
xcr2

Cxcl12
Cxcl5

Cxcl1

Ackr3
Cxcr4

Cxcl2Cxcl12

Cxcl1

Cxcr4
Ackr3

C
xcl2C

xc
l3Cxc

l2

Cxcl2
Cxcl5
Cxcl3
Ackr3

Cxcl5

C
xcr2

CXCL pathway

Fibroblast
Neutrophil
Macrophage
Malignant cell

 KvPCSmad4-KO

Tnfrsf1a

TnfTnf

Tnfrsf1a

Tnfrsf1b
Tnfrsf1b Tn

frs
f1

a

Tn
frs

f1
b

Tnfrsf1a

TNF pathway

 KvPCSmad4-WT

Tnf

TNF pathway

Tnfrsf1a

Tnfrsf1b

Tn
frs

f1
b

Tnfr
sf1

aTnfrsf1a

H

 KvPCSmad4-WT  KvPCSmad4-KO

Cell type clusters

0

200

400

600

Vo
lu

m
e 

(m
m

3 )

Tumor volume

***

J

B
m

p2
C

cl
12

C
sf

1
Il1

a
Il2

3a
Il6 N

fk
b1

N
fk

bi
a

N
fk

bi
z

P
la

u
S

er
pi

nb
8

S
er

pi
ne

1
Tn

fa
ip

3
Tn

ip
1

Tx
ni

p
Fn

1
Id

2
M

ap
3k

8
M

m
p2

C
d8

3
C

xc
l1

C
xc

l2
Il7

r
R

el
S

qs
tm

1
Tr

af
1

E
pa

s1
C

cl
2

Tl
r7

A
tf3

G
m

11
12

7
N

os
2

M
t1

A
da

m
19

A
im

2
C

cl
4

X
bp

1
Fo

sl
1

C
hs

t1
5

E
no

1b
Il1

r1
Itg

a5
P

m
p2

2
P

pa
rg

Tn
frs

f9
Tg

fb
r2

S
10

0a
9

S
er

pi
nb

1a
Itg

b1
K

itl
N

fa
tc

2
O

lr1
H

2_
q1

0
H

2_
q4

Il15
Csf2
Il1a
Il1b
Tnf

Predicted target genes

From Malignant cells to Macrophages (KvPCSmad4-KO)

P
rio

rit
iz

ed
 li

ga
nd

s

Regulatory potential
0.0 0.1 0.2

K

A
m

pd
3

C
fla

r
C

sf
1

E
ts

1
Ic

am
1

Ly
n

N
fk

b1
N

fk
b2

N
fk

bi
z

P
fk

fb
3

P
tg

s2
S

er
pi

ne
1

Tn
fa

ip
2

Tn
fa

ip
3

B
irc

3
C

xc
l1

R
el

Tr
af

3
M

ap
3k

8
C

d8
0

C
is

h
C

re
m

Fc
gr

2b
G

k
Il1

rn
Irs

2
Itg

a5
O

sm
S

lc
7a

11
Fa

s
S

irt
1

F3 N
os

2
A

rrd
c3

O
lr1

A
lo

x5
La

m
b3

N
r4

a2
S

10
0a

9
Tl

r2
H

2_
q4

N
fk

bi
e

Itg
am

C
d1

4

Csf2

Il1a

Il1b

Tnf

Predicted target genes

From Malignant cells to Neutrophils (KvPCSmad4-KO)

Regulatory potential
0.0 0.1 0.2

P
rio

rit
iz

ed
 li

ga
nd

s

Neutrophils Macrophages
0

100

200

300

%
 li

ve
 s

in
gl

et
s 

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l ***

***

KvPCSmad4-KO (n=28)
KvPCSmad4-WT (n=10)

Acinar cell
B cell
Ductal cell 1
Ductal cell 2
Endothelial cell
Fibroblast
Neutrophil
Macrophage
Malignant cell
Mast cell
Pericyte

min

max
Commun. Prob.

0.01 < p < 0.05

p < 0.01

p−value



A
dm

A
tf3

C
cn

d2
C

dk
n1

a
D

us
p6

E
dn

1
E

gf
r

E
lf3

Fo
s

Fs
t

Il1
b

M
m

p2
M

uc
1

P
m

ai
p1

P
tg

s2
S

gk
1

Tn
f

Tn
frs

f2
2

Tn
frs

f2
3

Tr
ib

1
Tx

ni
p

C
sf

2
C

xc
l2

D
us

p1
D

us
p5

H
be

gf
P

la
t

P
la

ur
P

pp
1r

15
a

S
er

pi
nb

1a
Tn

fa
ip

2
Tn

fa
ip

3
Il1

a
A

nk
rd

1
A

re
g

C
cd

c3
S

ftp
d

A
hr

C
as

p4
C

rp
Il1

3r
a1

P
ls

cr
1

P
ls

cr
2

Tn
frs

f2
1

Fgf7

Il6

Il1a

Predicted target genes

Regulatory potential

0.00 0.10 0.20

Figure 7

A
B

H

-3 -2 -1 0 1 2 3

Oxidative phosphorylation HALLMARK
cCAF3

TGFβ-induced myCAF in vitro
Fatty acid metabolism HALLMARK

NF-kappa B signaling pathway KEGG
Collagen degradation REACTOME

Inflammatory response HALLMARK
JAK-STAT signaling pathway KEGG

ECM organization REACTOME
iCAF in vivo

Interleukin 6 signaling REACTOME
iCAF in vitro

IL6 JAK STAT3 signaling HALLMARK
TNF-α signaling via NFkB HALLMARK

NES

scRNA-seq - KvPCSmad4-KO vs KvPCSmad4-WT - fibroblast cluster

iCAF-associated myCAF-associated other

JI

E
Malignant cell cluster

S
m

ad
4

Il1
a

Il1
b

WT
KO

scRNA-seq - KvPCSmad4-WT (n=4) and KvPCSmad4-KO (n=4) PDAC tumors

Malignant cell cluster Fibroblast clusterF G Fibroblast clusterD

P
rio

rit
iz

ed
 L

ig
an

ds

Pglyrp1
Btc

Ccl20
Ocln

Vsig10
Il1rn
Ccl7
Tgfa

Crb3
Fst

Sftpd
Ptprc
Bgn

Clnd2
Alcam
Gpc6
Gas6

Il1a
Tnf
Il1b

Ligand Activity Plot
From Malignant cells 

to Fibroblasts
(KvPCSmad4-KO)

AUPR

0.0100
0.0125
0.0150
0.0175
0.0200
0.0225

C

A
ck

r3
A

kr
1b

3
A

m
pd

3
G

bp
2

G
fp

t2
Il1

a
Il6 Irf

1
M

af
f

N
fk

b1
N

fk
b2

N
fk

bi
z

N
oc

t
Pa

nx
1

P
la

u
P

tg
s2

P
tx

3
S

oc
s3

Tn
fa

ip
2

Tn
fa

ip
3

Tn
ip

1
Tr

ib
1

V
ca

m
1

B
cl

3
B

irc
3

C
fb

C
xc

l1
C

xc
l2

R
el

S
er

pi
na

3n
A

da
m

ts
1

A
m

ot
l2

Il1
rn

M
t1

P
tg

s1
Tl

r2
A

tf3
M

yc
C

cl
3

M
m

p2
Il1

r1
Tl

r3
C

cd
c2

8a
S

am
d9

l
Zf

p3
6

Tgfa

Pdgfd

Il1a

Tnf

Il1b

Predicted target genes

P
rio

rit
iz

ed
 li

ga
nd

s

From Malignant cells to Fibroblasts (KvPCSmad4-KO)

Regulatory potential

0.0 0.1 0.2

From Fibroblasts to Malignant cells (KvPCSmad4-KO)

scRNA-seq - KvPCSmad4-WT and KvPCSmad4-KO PDAC tumors

NMacrophage clusterM Macrophage cluster

Ly
ve

1
Fo

lr2
Ap

oe
Ac

p5
M

sr
1

M
rc

1
Ar

g1
C

cl
3

C
xc

l1
C

xc
l2

Sp
p1

Ve
gf

a
Vc

an
Th

bs
1

M
ki

67
C

dk
1

C
d2

74
H

2-
Aa

H
2-

Ab
1

Irf
1

St
at

1
N

be
a

In
sy

n2
b

B3
ga

lt2

iCAFs apCAFs myCAFs
0

100

200

300

400

500

%
 li

ve
 s

in
gl

et
s 

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l *

**

***

CD90+ CD49E+ CD56+ CD105+
0

100

200

300

%
 li

ve
 s

in
gl

et
s 

co
m

pa
re

d 
to

 W
T 

co
nt

ro
l

***
***

*** *

KvPCSmad4-KO (n=28)
KvPCSmad4-WT (n=10)

KvPCSmad4-KO (n=28)
KvPCSmad4-WT (n=10)

P
rio

rit
iz

ed
 li

ga
nd

s

−0.5
−0.4
−0.3
−0.2
−0.1
0.0

−20
−10
0
10

5000 10000 15000

Neutrophil cluster - T3 signature

NES -1.29
FDR 0.45

L

Rank in Ordered Dataset

R
an

ke
d 

Li
st

 M
et

ric
R

un
ni

ng
 E

nr
ic

hm
en

t S
co

reK Neutrophil cluster



Figure 8

D

E

O

C

-4 -2 0 2 4

TGF-β-induced myCAF in vitro
LRRC15+ CAFs

myCAF in vitro
KRAS signaling UP HALLMARK

cCAF3
myCAF in vivo

Cholesterol homeostesis HALLMARK
Collagen formation REACTOME
ECM organization REACTOME

ECM-receptor interaction KEGG
E2F targets HALLMARK

iCAF in vivo
TNF signaling pathway KEGG

Oxidative phosphorylation HALLMARK
IL6 JAK STAT3 signaling HALLMARK
Interleuking 6 signaling REACTOME

iCAF in vitro

NES

RNA-seq - PSCs (+ KvPCSmad4-KO) vs PSCs (+ KvPCSmad4-WT) 

-4 -2 0 2 4

monocx - TGF-β signaling HALLMARK
cocx - TGF-β signaling HALLMARK

cocx - TNFα signaling via NFκB HALLMARK
cocx - Hypoxia HALLMARK

monocx - Cholesterol homeostasis HALLMARK
cocx - Cholesterol homeostasis HALLMARK

monocx - E2F targets HALLMARK
cocx - Fatty acid metabolism HALLMARK

cocx - Citric acid cycle TCA cycle REACTOME
cocx - MYC targets V1 HALLMARK

cocx - Respiratory electron transport REACTOME
cocx - Oxidative phosphorylation HALLMARK

cocx - E2F targets HALLMARK

NES

RNA-seq - KvPCSmad4-KO vs KvPCSmad4-WT organoids 

-3 -2 -1 0 1 2

Hedgehog ligand biogenesis REACTOME
Signaling by Hedgehog REACTOME

ECM organization REACTOME
Glycolysis HALLMARK

Mitotic spindle HALLMARK
Cholesterol homeostasis HALLMARK

Signaling by Hippo REACTOME

NES

scRNA-seq
KvPCSmad4-KO vs KvPCSmad4-WT - malignant cell cluster 

A

iCAF-associated myCAF-associated other

KP Smad4 WT PDAC KP Smad4 KO PDAC 

Macrophage

apCAF

iCAF

myCAFNeutrophil
IL-1

JAK/STAT signaling

PDAC progression

IL-1

(a)

F
kDa

100

75
100

75
50

37

p-STAT3

STAT3

ACTIN

JAKi -      +    -     +    -     +
Smad4 WT Smad4 KO

KvPC organoids H

STAT3

JAKi -      +    -     +    -     +
Smad4 WT Smad4 KO

KPC organoids

kDa

p-STAT3

ACTIN

100

75
100

75
50

37

I

B

Live cells
DAPI-

Organoids
PDPN-EpCAM+

KvPC organoid + PSC
co-cultures 

RNA-seq

PSCs
PDPN+EpCAM-

RNA-seq

KvPC organoid 
monocultures

Live cells
DAPI-

Organoids
PDPN-EpCAM+

RNA-seq

KvPCSmad4-WT cocx (n=4)
KvPCSmad4-KO cocx (n=8)

KvPCSmad4-WT monocx (n=4)
KvPCSmad4-KO monocx (n=7)

K

0

5

10

15

20

Tu
m

ou
r g

ro
w

th
 (f

ol
d 

ch
an

ge
)

 KPCSmad4-WT + vehicle (n=8)
KPCSmad4-WT + JAKi (n=11)
KPCSmad4-KO + vehicle (n=14)
KPCSmad4-KO + JAKi (n=14)

***

Tumor growth N

0

5

10

15

20

%
 P

os
iti

ve
 a

re
a 

/ t
ot

al
 a

re
a

CC3 stain quantification

***
**

KPCSmad4-WT + vehicle (n=8)
KPCSmad4-WT + JAKi (n=11)
KPCSmad4-KO + vehicle (n=14)
KPCSmad4-KO + JAKi (n=14)

JAKi sensitivity

KvP Smad4 WT PDAC KvP Smad4 KO PDAC 

Macrophage

apCAF

myCAF
Neutrophil

PDAC progression

(b)

KPCSmad4-WT and KPCSmad4-KO 

organoidsJ

C57BL/6J

Histology
U/S

Flow cytometry

PDAC tumor

Orthotopic surgery

2-week treatment
AZD1480 (JAKi)

U/S

Ve
hi

cl
e

CC3 IHC

JA
K

i

M

Smad4 WT Smad4 KO
KPC PDAC tumors

L

liver
metastases

diaphragm
metastases

lung
metastases

KPCSmad4-WT + vehicle (n=8)
KPCSmad4-WT + JAKi (n=11)
KPCSmad4-KO + vehicle (n=14)
KPCSmad4-KO + JAKi (n=14)

***
***

***
***

***
***

**

0
10
20
30
40
50
60
70
80
90

100

%
 m

ic
e 

w
ith

 m
et

as
ta

se
s

ACTIN

p-STAT3

p-SMAD2/3

Smad4 KO
KO3 KO4

Smad4
WT

KvPC organoids

STAT3

SMAD2

kDa

100
75
100

75

50
75

50

37

3 24 48 72 96 120 144
0

20

40

60

80

100

120

Hours

C
el

l v
ia

bi
lit

y 
co

m
pa

re
d

to
 o

w
n 

D
M

SO
 c

on
tro

l (
%

)

***
***

*** ***

***

KPCSmad4-WT + JAKi (n=3)
KPCSmad4-KO + JAKi (n=3)

KvPCSmad4-WT + JAKi (n=3)
KvPCSmad4-KO + JAKi (n=3)

G

24 48 72 96 120 144
0

20

40

60

80

100

120

Hours

C
el

l v
ia

bi
lit

y 
co

m
pa

re
d

to
 D

M
SO

 c
on

tro
l (

%
)

3