Skip to main content
NCBI home page
Cell Communication and Signaling : CCS logoLink to Cell Communication and Signaling : CCS
. 2026 Jul 10;24:533. doi: 10.1186/s12964-026-03052-2

S100A4 promotes collective invasion of pancreatic ductal adenocarcinoma and its blockade suppresses metastasis

Dong Woo Son 1, Ye-jin Lim 1, Chae Won Yu 1, Min Jung Cha 1, Eun Ji Lee 1, Hyori Kim 2, Heechan Lee 3,4, Seung Mo Hong 5, Hahnbeom Park 3, Suhwan Chang 1,✉
PMCID: PMC13637410  PMID: 42432705

Abstract

Most solid tumors invade peritumoral tissues as multicellular units while maintaining cell-cell interactions and epithelial characteristics, a process known as collective invasion. Our previous proteomic analysis revealed that S100-family proteins were upregulated in the PDAC secretome. Therefore, we aimed to elucidate the functional roles of these proteins in collective invasion and PDAC metastasis. To this end, we established a 3D collective invasion model using co-culture spheroids composed of PDAC cells and pancreatic stellate cells (PSCs). Using this model, we found that S100A4 contributes to collective invasion by mediating PDAC cell-PSC interactions, with S100A4 knockdown inhibiting collective invasion and S100A4 overexpression promoting it. Additionally, depletion of S100A in PDAC cells reduced trans-endothelial migration and cancer cell adhesion to endothelial cells. Mechanistically, PDAC cell-derived extracellular S100A4 promotes LAMB3 expression in PSCs. S100A4 depletion significantly reduced histological tumor aggressiveness in a pancreatic tumor orthotopic model and inhibited metastasis. To explore the therapeutic potential of these findings, we developed de novo S100A4-targeting inhibitory peptides using AI-guided structure modeling and confirmed their inhibitory effects on collective invasion. Collectively, our findings identify S100A4 as a critical driver of collective invasion and PDAC metastasis and support extracellular S100A4 as a potential target for AI-guided therapeutic intervention.

Supplementary Information

The online version contains supplementary material available at https://doi.org/10.1186/s12964-026-03052-2.

Introduction

Pancreatic cancer is one of the most lethal types of cancer, with the lowest five-year survival rates across all cancer types. Pancreatic ductal adenocarcinoma (PDAC) is the most common subtype, constituting approximately 90–95% of all pancreatic cancers, and is associated with a 5-year overall survival rate of only 11% [1, 2]. PDAC is characterized by its highly aggressive nature with local invasion and distant metastasis [3]. In a previous study, we performed proteomic analysis on tumor interstitial fluids from patient-derived xenograft models of PDAC and breast cancer [4]. The tumor secretome, which consists of proteins released by cancer cells and their surrounding microenvironment, is a valuable source of potential therapeutic or diagnostic targets [5]. We identified 87 proteins highly expressed in PDAC, among which S100-family proteins were prominently represented. S100-family proteins are a group of calcium-binding proteins expressed exclusively in vertebrates and comprise 21 members with high structural similarity [6]. Dysregulated expression of S100-family proteins is commonly observed in many human cancers [7]. Previous studies have also demonstrated that secreted S100 family proteins are markedly enriched in PDAC [8]. S100P has been reported to be highly expressed in PDAC, and extracellular S100P activates pro-invasive signaling and enhances epithelial-mesenchymal transition (EMT) through RAGE-mediated mechanisms [9]. S100A4 expression is significantly upregulated in PDAC and is associated with EMT progression and poor prognosis [10]. S100A10 is significantly overexpressed in PDAC tissues and correlates with aggressive tumor phenotype and poor prognosis [11]. Therefore, we hypothesized that enriched S100P, S100A4, and S100A10 promote tumor aggressiveness and metastatic potential in PDAC.

Invasion is a fundamental step in the early stages of metastasis. Cancer cells can invade peritumoral tissue through various modes, including amoeboid, mesenchymal, and collective invasion [12]. Several lines of evidence suggest that many solid tumors, including PDAC, invade cohesively as multicellular units within stromal tissues and blood vessels while maintaining cell-cell contact and epithelial-like characteristics. This phenomenon is known as collective invasion [13–16]. A recent study suggested that the collective invasion phenotype is observed more frequently than single-cell invasion in patient samples across multiple types of tumors [17]. A recent study reported that S100P promotes collective invasion of PDAC cells. They conducted a chemorepellent-induced defect (CCID) assay to evaluate collective invasion [18]. Thus, we investigated whether S100-family proteins regulate the collective invasion of PDAC cells using a spheroid-based 3D collective invasion model. Furthermore, after tumor invasion, cancer cells undergo intravasation and extravasation during metastasis [19]. We assessed the functions of S100-family proteins in these processes using a trans-endothelial migration assay, which recapitulates the process by which cancer cells cross the vascular endothelium [20, 21].

In this study, we sought to investigate the functional roles of S100P and S100A4 to metastatic dissemination of PDAC. Using PDAC-pancreatic stellate cell (PSC) co-culture spheroids and endothelial interaction model, we examined whether S100P and S100A4 regulate key steps of metastasis, including migration, collective invasion, trans-endothelial migration, and in vivo metastasis. Finally, this study provides insight into how S100A4 may promote PDAC aggressiveness within the tumor microenvironment and suggests the therapeutic potential of blocking extracellular S100A4 through de novo-designed inhibitory peptides.

Materials and methods

Cell culture

The human pancreatic cancer cell lines BxPC-3, AsPC-1, and MIA PaCa-2 were maintained in RPMI-1640 (SH30027.01, Cytiva, Marlborough, MA, USA) supplemented with 10% fetal bovine serum (FBS) (SH30919.03, Cytiva) and 1% penicillin-streptomycin (Cytiva). Human embryonic kidney 293Tx cells and human pancreatic cancer cell line Panc-1 were cultured in Dulbecco’s modified Eagle medium (DMEM) (SH30243.01, Cytiva) with 10% FBS and 1% penicillin-streptomycin. The human patient-derived pancreatic cancer cells 115,026, 110,621, and 19,224 were cultured in RPMI-1640 supplemented with 5% FBS, 1% penicillin-streptomycin, 20 ng/mL EGF, 4 µg/mL hydrocortisone, and 4 µg/mL transferrin. Human pancreatic stellate cells (hPSCs) were incubated in stellate cell medium (#5301, ScienCell, Carlsbad, CA, USA). Human umbilical vein endothelial cells (HUVECs) were purchased from Lonza (C2519A) and maintained in Endothelial Cell Basal Medium-2 (CC-3156, Lonza, Basel, Switzerland) with EGM-2 SingleQuot (CC-4176, Lonza). All cells were maintained at 37 °C in a 5% CO2 humidified cell incubator. The patient-derived PDAC cells were obtained with permission from the institutional review board of the Asan Medical Center (S2013-0744-0009).

Plasmids and molecular cloning

The pLKO.1-TRC cloning vector (plasmid #10878, Addgene, Watertown, MA, USA) was used for shRNA cloning. The shRNA fragment was cloned into the backbone plasmid using the restriction enzymes EcoRI and AgeI through standard cloning techniques. Target shRNA sequences were designed using publicly available bioinformatic tools (https://portals.broadinstitute.org/gpp/public/seq/search). Overexpression vectors were generated using the gateway cloning system. Gateway BP Clonase Ⅱ enzyme mix (11789-020, Invitrogen, Carlsbad, CA, USA) and gateway LR Clonase Ⅱ enzyme mix (11791-020, Invitrogen) were used for the gateway cloning according to the manufacturer’s instructions. pDONR207 (12536017, Invitrogen) was used for the entry clone with the BP reaction and pLenti6.3/To/V5-DEST (A11144, Invitrogen) was used for the destination vector with the LR reaction. The primers used for shRNA cloning and the Gateway cloning system are listed in Supplementary Table S1.

Lentiviral transduction

To generate recombinant lentivirus, human embryonic kidney 293Tx cells were co-transfected with lentiviral target vectors and lentiviral packaging vectors (psPAX2, pMD2.G) with Lipofectamine 3000 reagent (L3000075, Invitrogen) according to the manufacturer’s instructions. The lentiviral supernatant was harvested twice at 12-h intervals and stored at 4 °C. Then, target cells were incubated with virus supernatant and 5 µg/mL hexadimethrine bromide (H9268, Thermo Fisher Scientific, Waltham, MA, USA) for 2 days and selected using 0.5–1 µg/mL puromycin or 5–8 µg/mL blasticidin, according to the selection marker of the recombinant vectors.

Co-culture spheroid model

Two methods were used to generate co-culture spheroids: the ultra-low attachment (ULA) plate method and hanging-drop (HD) method. For the ULA method, 1 × 103 PDAC cells and 3 × 103 PSCs were incubated in a round-type 96-well ULA plate (34896, SPL Life Science, Gyeonggi-do, Korea) with 100 µL of complete medium for 24 h at 37 °C in a 5% CO2 incubator. For the HD method, the cells were prepared in complete media (5 × 102 PDAC cell and 1.5 × 103 PSCs in each 20 µL of a drop). Subsequently, 20 µL of cell suspension was deposited onto the lid of a petri dish using a multi-channel pipette. To prevent evaporation, phosphate-buffered saline (PBS) was placed at the bottom of the dish. The lid was then placed onto the PBS-filled petri dish and incubated for 24 h at 37 °C in a 5% CO2 incubator.

Spheroid-based cellular viability assay

The Promega® CellTiter-Glo 3D cell viability assay kit (G9683, Promega, Madison, WI, USA) was used to assess the cellular viability of tumor spheroids according to the manufacturer’s instructions. Briefly, spheroids were generated in a 96-well ULA plate and then transferred individually to each well of a 96-well cell culture plate with 100 µL of culture media at each time point, such as 24 h, 48 h, and 72 h after spheroid formation. Then, 100 µL of CellTiter-Glo 3D reagent was added to each well and incubated for 30 min at room temperature (RT) with gentle shaking for the first 5 min. Subsequently, the luminescence signal was measured using a microplate reader with 0.5-s integration time (PowerWave HT 340, BioTek, USA).

3D collective invasion model

To perform the 3D collective invasion model, the matrix was prepared prior to use, and all steps were carried out on ice. The matrix consisted of a collagen solution (#04902, Stemcell Technologies, Vancouver, Canada) at a concentration of 0.75 mg/mL and Matrigel (#354234, Corning, NY, USA) at a concentration of 4 mg/mL. Spheroid embedding was carried out in two steps. First, 95 µL of the matrix solution was spread at the bottom of a 48-well cell culture plate and incubated for 30 min at 37 °C until polymerization. Next, spheroids were placed on the solidified matrix, and 75 µL of matrix solution was added to cover them entirely. The matrix was then incubated for 40 min at 37 °C in a 5% CO2 incubator. Subsequently, 500 µL of complete medium containing 20 mM HEPES was added. For the analysis of collective invasion, microscopy images were captured and analyzed using ImageJ. Collective invasion was quantified based on the relative invasion area, which was calculated as follows: Relative invasion area = (total invasion area − spheroid central area) / spheroid central area.

Spheroid-based migration assay

The 48-well cell culture plate was coated with 95 µL of matrix, composed of the same components as in the collective invasion model, and incubated for 30 min at 37 °C. Subsequently, fluorescence-labeled co-culture spheroids were placed on the matrix, and 300 µL of complete medium was added. After incubation for 48 to 96 h, fluorescence images were captured using a fluorescence microscope and analyzed with ImageJ. Quantification was performed by measuring the relative migration area based on fluorescence signals and normalizing it to the central area of the spheroids.

Trans-endothelial migration assay

HUVECs were seeded at 1 × 105 cells in the upper chamber of a transwell chamber (8 μm pore size, #3422, Corning) that was pre-coated with 50 µL of 0.2% gelatin for 15 min at RT. The chamber was incubated for 24 h at 37 °C in a 5% CO2 incubator. Afterwards, the chamber was washed with PBS twice to remove non-attached HUVECs. Subsequently, co-culture spheroids were placed in the upper chamber with 200 µL of RPMI media containing 0.5% FBS, and 800 µL of complete medium was added to the lower chamber. After 48 h of incubation, non-migrated cells in the upper chamber were removed using a cotton swab, washed twice with PBS, and then fluorescence images were captured using a fluorescence microscope and analyzed with ImageJ. Quantification was performed by counting the transmigrated cells.

HUVEC adhesion assay

In total, 1 × 105 HUVECs were seeded into each well of a 48-well cell culture plate pre-coated with 0.2% gelatin and incubated at 37 °C in a 5% CO2 incubator for 24 h. Once the HUVEC monolayer reached 100% confluence, as confirmed by microscopy, each well was washed twice with PBS. Subsequently, GFP-labeled PDAC cells were seeded with 500 µL of serum-free medium. After incubation for 2 h at 37 °C, non-adherent cells were washed out twice with PBS. Fluorescence images were then captured, and quantification was performed by counting the GFP-expressing cells.

RNA extraction and quantitative real-time polymerase chain reaction (PCR) analysis

Total RNA was extracted using Tri-RNA reagent (FATRR 001, Favorgen Biotech Corp., Ping Tung, Taiwan). cDNA was synthesized with 1 µg of RNA using PrimerScript 1st strand kit (6110, Takara, Kusatsu, Japan), according to the manufacturer’s instructions. Real-time PCR was performed using CFX Connect Real-time PCR Detection System (Bio-Rad) and EzAmp™ qPCR 2X Master Mix (EBT-1801, ELPIS-BIOTECH, Daejeon, Korea). GAPDH was used as an internal control. Thermal cycling conditions comprised 1 step at 95 °C for 3 min followed by 40 cycles at 95 °C for 10 s and at 60 °C for 30 s. Each measurement was conducted in triplicate. The PCR product quality was confirmed by melting-curve analysis. Relative gene expression was calculated by the 2−(∆∆Ct) method based on cycle threshold (Ct). The primers used for qPCR analysis are listed in Supplementary Table S1.

Protein extraction and western blot analysis

Whole cell lysates were extracted using radioimmunoprecipitation assay buffer (RC2002-050-00, Biosesang, Gyeonggi-do, Korea) supplemented with a protease inhibitor cocktail (05892791001, Roche) for 15 min on ice, followed by centrifugation at 15,000 rpm for 15 min at 4 °C. The protein extracts were quantified using a BCA assay kit (23225, Thermo Fisher Scientific). The samples were separated on 15% SDS-PAGE gels and transferred onto 0.45 μm polyvinylidene difluoride membranes (10600023, GE Healthcare, Chicago, IL, USA). The blots were blocked with 5% skim milk (232100, BD Biosciences, Franklin Lakes, NJ, USA) and incubated with primary antibodies overnight at 4 °C on a rotator. Subsequently, the blots were washed three times with 0.1% TBST buffer, incubated with secondary antibodies for 1 h at room temperature on a rotator, and then washed five times with 0.1% TBST. Protein detection was performed using enhanced chemiluminescence (ECL) reagents, specifically the WestGlow™ FEMTO ECL chemiluminescent substrate (BWF0100, Biomax, Gyenggi-do, Korea) and WestGlow™ PICO PLUS ECL chemiluminescent substrate (BWP200, Biomax). The antibodies used in the western blot analysis are described in Supplementary Table S2.

Indirect co-culture using a transwell system

Indirect co-culture of PDAC cells and PSCs was performed using a transwell chamber system with a 0.4 μm pore size. In total, 2 × 105 115,026 cells were seeded in the lower chamber, while 2 × 105 PSCs were seeded in the upper chamber. RPMI medium containing 5% FBS was used for incubation. After 48 h of incubation, cells were harvested using Tri-RNA reagent for RNA extraction and real-time PCR analysis.

Pancreatic orthotopic model using PDAC co-culture spheroids

All mice were maintained in a pathogen-free facility of the Laboratory of Animal Research at Asan Institute for Life Science and animal studies were conducted at the Asan Institute for Life Science and approved by the Institutional Animal Care and Use Committee of Asan Institute for Life Science (2024-20-224).

Ten-week-old NOD-scid gamma (NSG) mice were obtained from Jung-A Bio Laboratories. For the pancreatic tumor orthotopic model using co-culture spheroids, 25 co-culture spheroids were resuspended in 100 µL of Matrigel (354234, Corning) and injected into the pancreas using a Hamilton syringe (Hamilton, Reno, NV, USA). Tumor growth was monitored using the IVIS Spectrum (PerkinElmer, Shelton, CT, USA).

In vivo metastasis analysis of pancreatic cancer cells with S100A4 knockdown

All animal studies were conducted at the Asan Institute for Life Sciences in accordance with protocols approved by the Institutional Animal Care and Use Committee (IACUC) of the Asan Institute for Life Science (2023-40-264). Female NSG mice were randomly assigned to three experimental groups (n = 6 per group) and intravenously injected via the tail vein with shCON, shS100A4, or shS100P AsPC-1 cells. Briefly, 1 × 106 cells suspended in 100 µL sterile phosphate-buffered saline (PBS) were administered to each mouse on Day 0. Cell suspensions were freshly prepared immediately before injection. Prior to the main study, a pilot experiment was performed to evaluate the ability of intravenously injected AsPC-1 cells to establish liver and lung metastases and to determine the optimal experimental endpoint. NSG mice were injected with 1 × 106 AsPC-1 cells via the tail vein and sacrificed at either 4 weeks (n = 2) or 5 weeks (n = 2) after injection. Based on the pilot study findings and previous reports, a 4-week endpoint was selected for subsequent experiments. Mice were monitored throughout the study and euthanized 28 days after tumor cell inoculation. Livers and lungs were harvested, fixed in 10% neutral-buffered formalin, paraffin-embedded, and subjected to hematoxylin and eosin (H&E) staining. Metastatic burden was evaluated by quantifying the metastatic area in H&E-stained tissue sections.

Immunohistochemistry

Specimens were fixed in 10% formalin for 48 h and embedded in paraffin. Paraffin Sect.  (4 μm thickness) were deparaffinized with xylene and rehydrated through a diluted ethanol (100%, 90%, 80%, and 70%) for 10 min each. The sections were immersed in 3% hydrogen peroxide containing 10% methanol to quench endogenous peroxidase activity. Antigens were retrieved using a citrate buffer (pH 6.0) containing 0.05% Tween 20 for 20 min. To reduce nonspecific staining, the sections were treated with blocking buffer consisting of 1.5% normal horse serum in PBS for 30 min. Subsequently, the sections were incubated with primary antibody diluted in 100 µL of blocking buffer at 4 ℃ overnight. The sections were then washed twice with 0.1% PBST and twice with PBS. For detection, the VECTASTAIN ABC-HRP kit (PK-6200, Newark, CA, USA) was used. The sections were incubated with a biotinylated secondary antibody (1:100) for 1 h, washed twice with 0.1% PBST and twice with PBS, and then incubated with prepared ABC reagent (a mixture of 10 µL of reagent A and 10 µL of reagent B in 1 mL of PBS) for 30 min. After two rinses with PBS, the sections were developed using the Dako Liquid DAB + Substrate Chromogen System (K3468, Dako) and counterstained with hematoxylin. The slides were mounted with a coverslip using Permount (SP15-500, Fisher Scientific), and the sections were observed under a light microscope.

Sirius red staining

Specimens were fixed in 10% formalin for 48 h and embedded in paraffin. Paraffin Sect.  (4 μm thickness) were deparaffinized with xylene and rehydrated using graded ethanol solutions (100%, 90%, 80%, and 70%) for 2 min each. The sections were incubated with Picro-Sirius Red solution (ab246832, Abcam) for 60 min at RT. Subsequently, the sections were rinsed quickly in two changes of 0.5% acetic acid solution and then washed with absolute ethanol twice. The slides were mounted with a coverslip using Permount (SP15-500, Fisher Scientific), and the sections were observed under a light microscope.

De novo design of S100A4-targeting peptides

Using the S100A4 crystal structure (PDB: 2Q91 [22]) as input, we generated 1,000 57-amino-acid binder backbone scaffolds using RFdiffusion [23]. ProteinMPNN [24] then designed corresponding sequences for these scaffolds, resulting in 2,000 candidate binders. These candidates were filtered using IEFFEUM [25] to remove those predicted to be unstable (ΔG < 0 kcal/mol) and with AlphaFold2 [26] to select those with high confidence metrics (pLDDT ≥ 80 and iPAE ≤ 10), yielding 44 candidates. Selectivity against homologous proteins was evaluated by filtering against S100A1, S100A6, and S100P using AlphaFold2 (iPAE > 10), leaving 23 candidates that are highly specific to S100A4. Finally, four candidates were selected for experimental validation based on their predicted Rosetta binding energy [27, 28] (ddG) to S100A4 (Supplementary Table S3). The selectivity of the four peptide candidates over S100 family is calculated by AlphaFold multimer, confirming its relative affinity to S100A4 target (Supplementary Table S4).

Pull-down assay

To confirm the secretion of S100P and S100A4 by the PDAC cells, we performed a pull-down assay. First, PDAC cells were transiently transfected with vectors encoding fusion proteins, which are S100 proteins tagged with the Fc region of immunoglobulin (Ig)G. The medium was replaced using fresh culture media, and then cell supernatants were harvested after 24 h. Subsequently, 500 µL of cell supernatant was incubated with 20 µL of protein G agarose beads (#17061801, Cytiva) for 3 h at 4 ℃ on a rotator. The beads were then washed three times with PBS containing 0.5% NP-40 and collected by centrifugation at 2,000 rpm for 5 min at 4 °C. Subsequently, the beads were boiled in 1× sample loading buffer for 10 min, and the supernatant was collected for western blot analysis.

To assess the binding affinity of the designed peptides, we conducted a pull-down assay using biotinylated peptides on the cell supernatants from 293T cells transiently transfected with vectors to ectopically express S100P and S100A4. For the pull-down assay, 500 µL of cell supernatants were incubated with 1 µg of biotinylated peptides for 3 h at 4 °C on a rotator, followed by incubation with 10 µL of streptavidin magnetic beads (88816, Thermo Fisher Scientific) for 1 h 30 min at 4 °C on a rotator. The beads were then washed 5 times with PBS containing 0.5% NP-40 and collected using a magnetic separation rack. Subsequently, the beads were boiled in 1× sample loading buffer for 10 min, and the supernatant was collected for western blot analysis.

To obtain biotinylated peptides, EZ-Link NHS-PEG4-Biotin (21330, Thermo Fisher Scientific) was used. 1 µg of each peptide (#56, #58, #59, and #60) was initially reacted with a 20-fold molar excess of biotin in a total volume of 100 µL of reaction buffer (PBS). The reaction was performed according to the manufacturer’s instructions. Briefly, 1 µg of peptides was dissolved in the reaction buffer, and the appropriate volume of 20 mM NHS-PEG4-Biotin solution was added. The mixture was incubated on ice for 2 h. Subsequently, unreacted NHS-PEG4-Biotin was removed using an Amicon® Ultra centrifugal filter (0.5 mL, 3 K MWCO) (UFC500324, Millipore, Burlington, MA, USA). The biotinylated peptides were recovered at a final concentration of 1 mg/mL.

Statistical analysis

Statistical analyses were performed using GraphPad Prism Software Version 5 (GraphPad Software Inc., San Diego, CA, USA). All data are presented as the mean ± SEM. The comparison between two groups was performed using Student’s t-test. *P < 0.05, **P < 0.01, and ***P < 0.001.

Results

Establishment of a 3D collective invasion model using tumor spheroids

We established a co-culture spheroid model composed of PDAC cells and pancreatic stellate cells (PSCs) to perform functional assays in a three-dimensional context and recapitulate interactions between these two cell types. We initially tested various mixed ratios of PDAC cells and PSCs. The experimental groups included PDAC-only spheroids without PSCs and PDAC: PSC ratios of 1:1, 1:2, 1:3, and 1:4. Cells in the PDAC-only control group did not aggregate, whereas incomplete spherical structures were observed in the 1:1 and 1:2 ratio groups. In contrast, compact spherical structures were observed in the 1:3 and 1:4 ratio groups (Supplementary Fig. S1A). Additionally, we conducted morphological analysis using three parameters: circularity, roundness, and aspect ratio. The 1:3 ratio spheroids exhibited the most spherical structures compared with the other groups (Supplementary Fig. S1B). Based on these results, we used co-culture spheroids with a 1:3 ratio of PDAC cells and PSCs in this study. Next, we assessed spheroid growth as an indicator of cellular viability by measuring spheroid volume using the following equation: Volume = 0.5 × major axis × minor axis2, as described in a previous report [29]. The volume of co-culture spheroids gradually increased over time (Supplementary Fig. S1C). Using fluorescence-labeled cells, we investigated the spatial distribution of PDAC cells and PSCs within the co-culture spheroids and found that both cell types were evenly distributed (Supplementary Fig. S1D). Finally, we confirmed the co-culture spheroid models could be generated using various PDAC cells, including patient-derived PDAC cells (115026, 110621, and 19224) and PDAC cell lines (Panc-1, BxPC-3, MIA PaCa-2, and AsPC-1) (Supplementary Fig. S1E).

Previous studies have established 3D invasion models using ECM gels, such as Matrigel, collagen, or a mixture of both [30–33]. Moreover, several studies have used 3D invasion models based on co-culture spheroids composed of PDAC cells and PSCs to investigate the interactions between PDAC cells and stromal cells [34, 35]. Given the clinical significance of collective invasion as a mechanism of tumor invasion, we established a 3D collective invasion model using co-culture spheroids, as shown schematically in Fig. 1A. To assess the collective invasion phenotype, we evaluated the expanding invasion streams composed of invading cells (Fig. 1E). Initially, we tested three different matrix compositions for embedding co-culture spheroids: collagen-only, Matrigel-only, and a mixture of collagen and Matrigel. These compositions have been commonly used in 3D culture and spheroid-based invasion assays in previous reports. In the collagen-only matrix (final concentration: 1.5 mg/mL), an invasive phenotype was observed; however, the invading cells predominantly exhibited a single-cell invasion phenotype (Fig. 1B, left). In the Matrigel-only matrix (final concentration: 8 mg/mL), invasion streams were not observed (Fig. 1B, middle). Interestingly, the collective invasion phenotype was observed in the matrix composed of a collagen and Matrigel mixture at final concentrations of 1.5 mg/mL and 4 mg/mL, respectively (Fig. 1B, right). A recent study reported that invasion was reduced in highly concentrated collagen matrices [36]. We tested two groups of low-concentration collagen matrices (0.75 mg/mL and 1 mg/mL) while maintaining the same concentration of Matrigel. Although there was no significant difference in the collective invasion phenotype between the two groups, a single-cell invasion phenotype was observed at the leading edge of the invasion stream in the 1 mg/mL collagen group (Fig. 1C). Based on these data, we performed 3D collective invasion model using a matrix composed of 0.75 mg/mL collagen and 4 mg/mL Matrigel throughout this study. Subsequently, we quantified collective invasion by calculating the relative invasion area using the following equation: Relative invasion area = (total invasion area − central area) / central area (Supplementary Fig. S2A). We then asked whether PSCs enhance the collective invasion of PDAC cells. To assess this, we performed a collective invasion assay using two types of spheroids: PDAC-only spheroids and co-culture spheroids. The co-culture spheroids showed a significantly higher relative invasion area than PDAC-only spheroids (Fig. 1D). Finally, we confirmed that the extent of collective invasion phenotype increased over time (Fig. 1E). Additionally, we extended the incubation period until day 17 to evaluate long-term collective invasion (Supplementary Fig. S3).

Fig. 1.

Fig. 1

Open in a new tab

Establishment of a 3D collective invasion model using co-culture spheroids. A Schematic illustrating the 3D collective invasion model. B Matrix composition test: 115,026 co-culture spheroids were embedded in collagen-only, Matrigel-only, or a collagen-Matrigel mixture matrices. C Collective invasion model using different collagen concentrations. The mixture of 0.75 mg/mL collagen and 4 mg/mL Matrigel (left panel) or 1 mg/mL collagen and 4 mg/mL Matrigel (right panel) was used. D Cell type-specific collective invasion model: PDAC-only spheroids vs. co-culture spheroids. Representative images were captured, and relative invasion area was quantified on days 5, 7, 9, and 11. E Representative images of the collective invasion model using 115,026 co-culture spheroids on days 3, 5, 7, and 9. *P < 0.05, **P < 0.01, and ***P < 0.001

Previous studies have used spheroids generated by the hanging-drop (HD) method for 3D invasion assays [36, 37]. Therefore, we performed collective invasion assays using spheroids generated by the HD method, hereafter referred to as HD spheroids. First, we compared the morphology of spheroids generated by the ultra-low attachment plate (ULA) method and the HD method. Although the initial cell numbers were the same, the ULA method formed uniform spheroids, whereas the HD method generated spheroids with variable shapes and sizes (Supplementary Fig. S4A). The volume of HD spheroids gradually increased over time (Supplementary Fig. S4B). We encountered several issues when using HD spheroids in the 3D collective invasion model. First, HD spheroids exhibited highly heterogeneous collective invasion patterns, raising concerns about selection bias during data analysis (Supplementary Fig. S4C, left). Second, the proximity of the spheroids led to aggregation and bridging between spheroids (Supplementary Fig. S4C, middle). Third, the spheroids were not aligned along a single z-axis, which complicated the quantification of collective invasion (Supplementary Fig. S4C, right).

PDAC-derived S100P and S100A4 regulate collective invasion of co-culture spheroids

We examined the gene expression levels of S100P, S100A4, and S100A10 in patient-derived PDAC cells (115026 and 110621) and PDAC cell lines (BxPC-3, MIA PaCa-2, Panc-1, and AsPC-1). Both patient-derived PDAC cell lines, 115,026 and 110,621, exhibited higher expression of S100P and S100A4 than human pancreatic ductal epithelial (HPDE) cells, and S100A10 was also elevated compared with negative control cells (Supplementary Fig. S5A). We identified six clones of 115,026 cells that differentially expressed S100P and S100A4 (Supplementary Fig. S5B). Based on these data, we selected clones with high and low expression levels of S100P and S100A4: clones A and F were designated as 115,026high, and clone B was designated as 115,026low. We then performed knockdown experiments for S100P, S100A4, and S100A10 in clones A and F, and overexpression experiments for S100P and S100A4 in clone B.

We generated stable PDAC cell lines with knockdown of S100P, S100A4, or S100A10, and overexpression of S100P or S100A4. Knockdown and overexpression were validated at the mRNA (Fig. 2A and F) and protein levels (Fig. 2B and G). Cell viability was not altered in PDAC cell line with knockdown or overexpression of S100P or S100A4 (Supplementary Fig. S6A). We generated co-culture spheroids by mixing knockdown or overexpression PDAC cells with wild-type PSCs (Fig. 2D and H). No obvious morphological changes were observed in knockdown co-culture spheroids (Fig. 2C). Next, we assessed spheroid volume growth and evaluated cellular viability using the CellTiter-Glo 3D reagent (Promega). Knockdown of S100P or S100A4 reduced relative spheroid volume growth between day 0 and day 7. In addition, cellular viability was reduced in the S100A4-knockdown co-culture spheroids (Fig. 2E). In contrast, overexpression of S100P enhanced relative spheroid volume growth between day 0 and day 7 (Fig. 2I).

Fig. 2.

Fig. 2

Open in a new tab

Generation of stable PDAC cells with S100-family protein knockdown or overexpression. A and B mRNA and protein levels of PDAC cells with knockdown of S100P, S100A4, and S100A10. C H&E staining of 115,026 co-culture spheroids with knockdown of S100P, S100A4, and S100A10. D Representative images of 115,026 co-culture spheroids with knockdown of S100P, S100A4, and S100A10 at days 0 and 7. E Quantification of relative spheroid volume growth between day 0 and day 7. Relative cell viability at 24, 48, and 72 h after spheroid formation and normalized to the 24 h time point for each spheroid group. F and G mRNA and protein levels of PDAC cells with overexpression of S100P and S100A4. H Representative images of 115,026 co-culture spheroids with overexpression of S100P and S100A4 at days 0 and 7. I Quantification of relative spheroid volume growth between day 0 and day 7. *P < 0.05, **P < 0.01, and ***P < 0.001

We next asked whether PDAC cell-derived S100P and S100A4 regulate cell motility under co-culture condition with PSCs. We conducted spheroid-based functional assays to evaluate collective invasion and migration. In the 3D collective invasion model, knockdown of S100P or S100A4 reduced the relative invasion area (Fig. 3A and Supplementary Fig. S6B). Conversely, overexpression of S100P or S100A4 enhanced the relative invasion area (Fig. 3B). Furthermore, knockdown of S100A10 reduced the relative invasion area (Supplementary Fig. S6C, S7A, and S7B). Next, we established a spheroid-based migration assay using co-culture spheroids to assess cellular migration in a 3D context (Fig. 3C). The quantification method is shown in Supplementary Fig. S2B. Migration of PDAC cells from co-culture spheroids was reduced by knockdown of S100P or S100A4 (Fig. 3D and E). Under co-culture conditions with patient-derived 115,026 PDAC cells, knockdown of S100P and S100A4 in PDAC cells reduced PSC migration from the co-culture spheroids (Fig. 3D).

Fig. 3.

Fig. 3

Open in a new tab

Spheroid-based migration and collective invasion assays using co-culture spheroids with knockdown or overexpression. A Representative images and quantification of collective invasion using 115,026 and 110,621 co-culture spheroids with knockdown of S100P or S100A4. B Representative images and quantification of collective invasion of 115,026 co-culture spheroids with overexpression of S100P and S100A4. C Schematic illustrating the spheroid-based migration assay. D and E Representative images and quantification of the spheroid-based migration assay using 115,026 (D) and AsPC-1 (E) co-culture spheroids. *P < 0.05, **P < 0.01, and ***P < 0.001

Depletion of S100P or S100A4 in PDAC cells inhibits trans-endothelial migration and adhesion to HUVECs

During metastasis, disseminating cancer cells contact endothelial cells and disrupt endothelial junctions to pass through the vascular endothelium, a process known as a trans-endothelial migration [20, 21]. We further asked whether PDAC-derived S100P and S100A4 regulate transmigration across the endothelial cell layer. We conducted a trans-endothelial migration assay using fluorescence-labeled co-culture spheroids to evaluate the behavior of each cell type (Fig. 4A). We found that the knockdown of S100P or S100A4 reduced the trans-endothelial migration of PDAC cells. However, this suppressive effect was not observed in PSCs within the co-culture spheroids (Fig. 4B and C). During the metastatic process, cancer cells undergo rolling and adhesion along the vascular endothelium before transmigration [38, 39]. We assessed the adhesion capacity of knockdown and overexpression PDAC cells to HUVECs (Fig. 4D). Knockdown of S100P or S100A4 inhibited PDAC cell adhesion to HUVECs (Fig. 4E). Interestingly, overexpression of S100A4 enhanced HUVEC adhesion (Fig. 4F).

Fig. 4.

Fig. 4

Open in a new tab

Trans-endothelial migration assay and HUVEC adhesion assay using PDAC cells and co-culture spheroids with knockdown or overexpression. A Schematic illustrating the spheroid-based trans-endothelial migration assay. B and C Representative images and quantification of the trans-endothelial migration assay using fluorescence-labeled 115,026 (B) and AsPC-1 (C) co-culture spheroids with knockdown of S100P or S100A4. (PDAC cells, GFP; PSCs, RFP). Fluorescence images were captured at 48 h (B) and 96 h (C). D Schematic illustrating the HUVEC adhesion assay. E and F Representative images and quantification of HUVEC adhesion assay using GFP-labeled 115,026 cells with knockdown or overexpression. Fluorescence images were captured at 2 h. *P < 0.05, **P < 0.01, and ***P < 0.001

PDAC-derived extracellular S100A4 promotes LAMB3 expression in PSCs

To understand the mechanism underlying S100A4-driven collective invasion, we first conducted bioinformatic analyses using publicly available data. Gene expression data from PDAC tissues were analyzed to identify candidate genes correlated with S100A4 expression. We identified six genes (KRT19, CAPG, AHNAK2, ITGB4, LAMB3, and LAMA3) that were consistently correlated with S100A4 across the GEO database (GSE15471, GSE16515, GSE183795, GSE19281, GSE28735, GSE32688, GSE62165, and GSE62452) and cBioPortal cohorts (QCMG, Nature 2016; TCGA, PanCancer Atlas; CPTAC, Cell 2021). Finally, we selected three candidate genes (ITGB4, LAMB3, and LAMA3) that are associated with extracellular matrix (ECM) organization and cell-ECM interactions. Additionally, we included three candidate genes (MMP9, MMP14, and TGFB1), which have been reported to have pro-metastatic roles in PDAC [40, 41]. Considering that PDAC is enriched with cancer-associated fibroblasts and activated stellate cells, we assessed the gene expression of downstream effector genes under indirect co-culture conditions using PDAC cells and PSCs (Fig. 5A). The results showed that six candidate genes (ITGB4, LAMB3, LAMA3, TGFB1, MMP9, and MMP14) were downregulated in PSCs indirectly co-cultured with S100A4-knockdown PDAC cells (Fig. 5B and Supplementary Fig. S8A). In contrast to PSCs, S100A4-knockdown PDAC cells indirectly co-cultured with PSCs showed no changes in ITGB4 and LAMB3 expression, whereas LAMA3 was upregulated (Fig. 5C). To confirm the effect of extracellular S100A4 on PSCs, we assessed gene expression in PSCs treated with human recombinant S100A4 protein (rhS100A4), inhibitory peptides (#58 and #59) as S100A4 antagonists (see Fig. 7), and RAGE antagonist peptide (RAP) and FPS-ZM1 as inhibitors of RAGE, a major receptor for extracellular S100A4. In PSCs, LAMB3 expression was significantly upregulated following rhS100A4 treatment, whereas this upregulation was suppressed by co-treatment with rhS100A4 and S100A4 antagonists (#58 and #59) or RAGE inhibitors (RAP and FPS-ZM1) (Fig. 5D). Consistently, extracellular S100A4-mediated LAMB3 regulation in PSCs was also observed at the protein level (Fig. 5E). In contrast, MMP9 expression was downregulated following rhS100A4 treatment, and co-treatment with S100A4 antagonists reversed this effect. The expression of MMP14 and TGFB1 was not altered by rhS100A4 treatment (Supplementary Fig. S8B). These data led us to further investigate the role of LAMB3 in PSCs, and we performed a transwell invasion assay and a collective invasion assay following LAMB3 depletion in PSCs. Knockdown of LAMB3 in PSCs reduced PSC invasion (Fig. 5F), and collective invasion was reduced in the co-culture spheroids composed of PDAC cells and LAMB3-knockdown PSCs (Fig. 5G). These results indicate that PDAC cell-derived extracellular S100A4 promotes LAMB3 expression in PSCs and regulates PSC invasion and collective invasion of co-culture spheroids. Collectively, these findings suggest that LAMB3 from PSCs plays a critical role in the collective invasion of PDAC.

Fig. 5.

Fig. 5

Open in a new tab

PDAC cell-derived extracellular S100A4 regulates LAMB3 expression in PSCs. A Schematic illustrating the indirect co-culture assay using 115,026 cells and PSCs using a transwell chamber with a 0.4 μm pore size. B and C Candidate gene expression in PSCs and 115,026 cells under the indirect co-culture condition. D Candidate gene expression in PSCs following treatment with 1 µg/mL rhS100A4 alone or co-treatment with 1 µg/mL rhS100A4 and 2 µM S100A4 antagonists, including S100A4-targeting peptides #58, #59, RAP, and FPS-ZM1. E Western blot analysis of PSCs with treatment of rhS100A4 or co-treatment of rhS100A4 and S100A4-targeting peptide #58. F Representative images and quantification of the transwell invasion assay using PSCs with LAMB3 knockdown. G Representative images and quantification of collective invasion assay using co-culture spheroids composed of wild-type 115,026 cells and LAMB3-knockdown PSCs. *P < 0.05, **P < 0.01, and ***P < 0.001

Fig. 7.

Fig. 7

Open in a new tab

AI-guided design of S100A4-targeting inhibitory peptides and their inhibitory effects on collective invasion and cancer cell adhesion to HUVECs. A Binding poses of four S100A4-targeting peptides predicted by AlphaFold (white: S100A4 homodimer, colored: designed binder). B Schematic illustrating the pull-down assay using biotinylated peptides and streptavidin beads. (a) Incubation of cell supernatants with biotinylated peptides. (b) Pull-down of S100A4-biotinylated peptide complex using streptavidin beads. C and D Pull-down assay and western blot analysis for S100A4 (C) and S100P (D). Cell supernatants were collected from 293Tx cells expressing S100A4 (C) and S100P (D). No peptides, RAGE antagonist peptide (RAP), and non-biotinylated peptides were used as negative controls. E Representative images and quantification of collective invasion assay using115026 co-culture spheroids treated with 10 µM peptides (#56, #58, #59, and #60). Microscopic images were captured and quantified on day 13. F Representative images and quantification of HUVEC adhesion assay using 115,026 cells treated with 2 µM RAP and peptides (#56, #58, #59, and #60). *P < 0.05, **P < 0.01, and ***P < 0.001

Knockdown of S100P and S100A4 inhibits local invasion and metastasis

Next, we asked whether the inhibition of S100P and S100A4 reduces the local invasion and metastasis of PDAC in vivo. We established a pancreatic tumor orthotopic model using co-culture spheroids containing S100P- or S100A4-deficient PDAC cells and PSCs. We injected 25 co-culture spheroids into the pancreas of each NSG mouse (Supplementary Fig. S9). During the experiment, body weight gradually increased across all three groups of mice. At the endpoint, tumor weight was reduced in the S100A4-knockdown group (Fig. 6A). Histological analysis of primary tumors was conducted using H&E staining for shCON (n = 5), shS100P (n = 6), and shS100A4 (n = 3) groups. Notably, in the S100A4-knockdown group, 2 out of 3 mice showed PDAC with a predominant large duct pattern (PLDP), in which large cancer glands occupied more than half of the tumor volume (Fig. 6B). Collectively, our findings suggested that S100A4-deficient PDAC cells generated primary tumors with less aggressive morphology.

Fig. 6.

Fig. 6

Open in a new tab

Pancreatic orthotopic and metastasis model using S100-family knockdown PDAC cells. A Images and tumor weights of primary tumors collected from orthotopic mice at week 6. Body weight curves of mice bearing orthotopic tumors. B Representative images of H&E staining and IHC staining for PanCK at the tumor-normal tissue interface in primary tumors. Images were captured at 4× and 20× magnifications. Classification of primary tumor by histological morphology as either conventional PDAC or PDAC with PLDP. Classification of tumor-normal interfaces as pushing-type or infiltrative-type invasion. Statistical analysis was performed using the chi-squared test. C Representative images and quantification of Sirius Red staining. D Representative images and quantification for metastatic lesions of liver and lung, from the metastasis model generated by AsPC-1 orthotopic injection (n = 6 for each group). *P < 0.05, **P < 0.01, and ***P < 0.001

Next, we analyzed whether knockdown of S100P or S100A4 affects the local invasion of PDAC cells. Initially, we assessed the interface between cancerous tissue and normal pancreatic tissue using H&E staining (Fig. 6B, upper and middle). We classified tumor–normal tissue interface regions as infiltrative-type and pushing-type invasion and quantified them by counting the total number of each type in each mouse group [42]. The numbers of infiltrative-type invasion were significantly reduced in the S100P- and S100A4-knockdown groups (Fig. 6B). Furthermore, we validated this finding by immunohistochemistry (IHC) for PanCK, a pancreatic cancer cell marker (Fig. 6B, lower). The IHC results were consistent with the H&E staining findings. Next, to determine the distribution of collagen in the primary tumor mass, we performed Sirius red staining. Knockdown of S100P or S100A4 reduced collagen-positive areas within the tumor mass. In other words, desmoplastic regions were reduced in S100P- and S100A4-knockdown tumor masses (Fig. 6C). Together, these findings suggest that S100P and S100A4 in PDAC cells contribute to the local invasion of PDAC cells into normal pancreatic tissue and desmoplastic remodeling within primary tumors.

To further evaluate the metastatic capacity of S100P- and S100A4-deficient PDAC cells, we established a metastasis model in AsPC1 cells with S100P or S100A4 knockdown. The number of metastatic nodules in the lungs was significantly reduced in both knockdown groups (Fig. 6D and Supplementary Fig. S10) In the liver, S100A4 knockdown significantly reduce the number of metastatic nodules and S100P depleted cells showed similar pattern (Fig. 6D). These findings demonstrate that S100P and S100A4 contribute to PDAC metastasis.

S100A4-targeting inhibitory peptides suppress collective invasion and cancer cell adhesion to HUVECs

Our findings show that PDAC cell-derived extracellular S100A4 regulates LAMB3 expression in PSCs and contributes to collective invasion in vitro and aggressive histological features in vivo. Next, we aimed to block extracellular S100A4 by designing de novo S100A4-targeting inhibitory peptides. We predicted four de novo peptide sequences with high binding affinity for S100A4 (Fig. 7A and Supplementary Table S3). A detailed description of the computational design workflow is provided in the Methods section.

We generated stable PDAC cell lines expressing S100P or S100A4 fusion proteins tagged with the human IgG Fc region and performed a pull-down assay to confirm the secretion of these proteins from PDAC cells. Extracellular secreted S100P and S100A4 from PDAC cells were successfully detected in pull-down assay (Supplementary Fig. S5C). To evaluate the binding between the de novo S100A4-targeting inhibitory peptides and S100A4 proteins, the peptides were biotinylated and pulled down using streptavidin beads (Fig. 7B). We conducted a pull-down assay using cell supernatant from 293T cells expressing S100P or S100A4 fusion proteins. All inhibitory peptides (#56, #58, #59, and #60) showed higher binding affinity for S100A4 than RAP (Fig. 7C). However, these peptides did not bind to S100P (Fig. 7D). Subsequently, we performed a collective invasion assay with 10 µM inhibitory peptides. All peptides significantly reduced the relative invasion area (Fig. 7E). In another PDAC cell line, only peptide #59 exhibited inhibitory activity against collective invasion (Supplementary Fig. S11). Furthermore, we found that peptides #58 and #59 reduced PDAC cell adhesion to HUVECs (Fig. 7F). Collectively, these findings suggest that de novo S100A4-targeting inhibitory peptides block extracellular S100A4 activity and suppress collective invasion and PDAC cell adhesion to HUVECs.

Discussion

PDAC is a highly malignant tumor with a poor prognosis and is characterized by aggressive local invasion and metastatic dissemination to distant sites [43, 44]. Our previous tumor secretome analysis revealed that S100-family proteins were enriched in the PDAC tissue. In this study, we investigated whether S100-family proteins regulate PDAC invasion and metastasis.

Cancer cells invade peritumoral tissue through various mechanisms, including amoeboid, mesenchymal, or collective invasion [12]. Collective invasion has been reported as a crucial mechanism of cancer invasion in breast cancer [45, 46] and pancreatic cancer [18, 47, 48]. To evaluate collective invasion, we established a 3D collective invasion model using co-culture spheroids composed of PDAC cells and PSCs (Fig. 1A). Previous reports have used collagen, Matrigel, or a mixture of both for 3D invasion models [30–33]. Our results showed that a collagen–Matrigel mixed gel provided optimal conditions for PDAC collective invasion (Fig. 1B). Additionally, previous studies have established 3D invasion model using co-culture spheroids composed of PDAC cells and CAFs or PSCs [34, 35]. Consistent with these studies, we generated a PDAC-PSC co-culture spheroid model and found that PDAC-PSC interactions enhanced collective invasion (Fig. 1D). Using this in vitro model, our findings suggest that PDAC cell-derived extracellular S100A4 mediates crosstalk between PDAC cells and PSCs, thereby promoting collective invasion (Figs. 3 and 5). However, further studies incorporating stromal immune components, as well as phenotypic characterization of invading cells, will help further validate and extend these findings. Moreover, given that S100P and S100A4 are involved in tumor invasion, we evaluated the association between endogenous S100A4 and invasive capacity in PDAC cell lines. S100A4 expression was detected in MIA PaCa-2 cells, but not in BxPC-3 cells (Supplementary Fig. S12A). Consistently, invasive activity was observed in MIA PaCa-2 cells, whereas BxPC-3 cells showed no detectable invasive phenotype (Supplementary Fig. S12B).

Our findings suggest that PDAC cell-derived S100P, S100A4, and S100A10 play a key regulatory role in collective invasion. However, analysis of public datasets showed that S100A10 is highly expressed not only PDAC cells but also normal pancreatic cells, including normal ductal cells, acinar cells, and alpha cells (Supplementary Fig. S12C and S12D). These data suggest that S100A10 is not an ideal therapeutic target owing to potential off-target effects. Nevertheless, S100A10 could become a viable therapeutic target in PDAC if tumor-specific S100A10 protein modifications, such as glycosylation, are identified.

Previous studies identified collective invasion as a crucial mechanism in the vascular invasion of PDAC [47, 48]. Cancer cell adhesion to endothelial cells and trans-endothelial migration are critical steps in intravasation and extravasation during metastasis. To investigate this, we performed a trans-endothelial migration assay using co-culture spheroids. Considering that collectively invading cancer cells may proceed to vascular invasion upon encountering the blood vessels, we designed this assay using PDAC-PSC co-culture spheroids to simulate the interactions among PDAC cells, PSCs, and endothelial cells (Fig. 4A). Finally, we found that PDAC cell-derived S100A4 regulates PDAC cell adhesion to HUVECs (Fig. 4E and F). Collectively, these data suggest that targeting S100A4 has significant potential for inhibiting key steps of metastasis. Further studies are needed to validate these findings and explore the underlying molecular mechanisms.

In 115,026 co-culture spheroids, S100A4 knockdown in PDAC cells reduced PSC migration in the absence of endothelial cells, whereas this effect was not observed when endothelial cells were present. However, this effect was not shown in AsPC-1 co-culture spheroids (Figs. 3D and E and 4B, and C). S100A4 expression was significantly higher in AsPC-1 cells than 115,026 cells (Supplementary Fig. S5A). Based on our data, one possible explanation is that endothelial cells constitutively express RAGE and respond to extracellular S100A4 through RAGE-dependent signaling. Previous studies have reported that extracellular S100A4 promotes endothelial cell migration and angiogenesis by upregulating VEGFR2, MMP-9, and MMP-13 expression. In addition, extracellular S100A4 has been shown to disrupt endothelial monolayer integrity, thereby promoting melanoma cell transmigration [49–51]. Thus, endothelial cells may compete with PSCs for S100A4-mediated signaling, attenuating its effects on PSC migration. Moreover, endothelial cells can release soluble RAGE (sRAGE), which functions as a decoy receptor for extracellular S100A4 [52]. This may further weaken S100A4-mediated regulation of PSC mobility. Therefore, endothelial cells may reshape the crosstalk between PDAC cells and PSCs by modulating extracellular S100A4 signaling. Further studies are needed to clarify how endothelial cells interfere with S100A4-mediated PSC migration in the PDAC microenvironment.

We focused on the PDAC cell-derived S100A4 and its role in mediating interactions between PDAC cells and PSCs. LAMB3 expression in PSCs was downregulated under the indirect co-culture with S100A4-deficient PDAC cells (Fig. 5B) and upregulated following treatment with rhS100A4 (Fig. 5D). These findings suggest a possible mechanism by which extracellular S100A4 promotes stromal remodeling and collective invasion in PDAC. We also looked at cBioportal to check correlation between S100s and LAMB3 and the results are shown in Supplementary Fig, S13. We found strong correlation between S100P, S100A4 or other S100 family (A2, A6, A10 and A11) and LAMB3. Therefore, it is possible that other members of S100 family also involved in the regulation of LAMB3 in PSC. Previous studies have shown that LAM332(LAMB3) is more highly expressed and produced by CAFs than by PDAC cells or normal fibroblasts. The multiplex immunofluorescence (IF) analysis revealed that LAM332 is enriched in CAFs within PDAC patient tissues. Thus, CAFs may represent a major source of extracellular stromal LAM332 [53].

To evaluate the role of S100A4 in tumor aggressiveness and local invasion, we established pancreatic tumor orthotopic model. Our data showed that primary tumors in the S100A4-knockdown group predominantly exhibited PDAC with a PLDP (Fig. 6B). This is classified as a histological subtype of PDAC and exhibits relatively less aggressive clinicopathologic features with improved overall and recurrence-free survival compared to conventional PDAC [54]. Additionally, PDAC cell infiltration into normal pancreatic tissue was significantly reduced in the S100A4-knockdown group (Fig. 6B). These findings suggest that S100A4-knockdown in PDAC leads to the formation of less aggressive histological features and inhibits local invasion of PDAC cells. Furthermore, our histological analysis showed that PDAC cells predominantly invaded with a collective invasion phenotype, whereas single-cell invasion phenotypes were rarely observed (Fig. 6B). This finding is consistent with those of previous research [13]. In addition, our metastasis model demonstrated that S100P and S100A4 regulate distant metastasis (Fig. 6D). Collectively, in vivo experiments showed that S100P and S100A4 play a critical role in PDAC from local invasion to metastasis.

Our in vitro and in vivo results support the potential of de novo S100A4-targeting inhibitory peptides as a strategy to block extracellular S100A4-mediated tumor progression. Therefore, we designed and confirmed the inhibitory effects of de novo S100A4-targeting inhibitory peptides (Fig. 7). Additionally, we confirmed that the positive designs were predicted to have more favorable binding energetics than negative design background. The four positive designs (#56, #58, #59, and #60) showed lower normalized Rosetta ΔG values and were positioned closer to the known Myosin-2 A peptide than to most negative designs (Supplementary Fig. S14A).

In conclusion, our findings highlight the roles of PDAC cell-derived extracellular S100A4 in promoting PDAC aggressiveness and metastasis. We further suggest that a therapeutic blockade of extracellular S100A4 may represent a potential strategy for inhibiting PDAC metastasis.

Supplementary Information

Supplementary Material 1. (2.4MB, pdf)
Supplementary Material 2. (663.2KB, pdf)

Acknowledgements

This research was supported by Korea Health Industry Development Institute (KHDI), funded by the Ministry of Health & Welfare, Republic of Korea (HR21C0198); the University of Ulsan grant funded by the Korean Government (MOE, 2024 − 0912); National Research Foundation of Korea funded by the ministry of Education (RS-2024-00343979, 2021R1A6A1A03040260); and by the Asan Institute for Life Sciences (AILS, Grant 2023IP0116-2, 2025IP0037-1).

Authors’ contributions

Dong-woo Son performed the main experiments, primary data analyses, and drafted the manuscript. Ye-jin Lim and Eun-ji Lee provided technical support in the animal study. Min Jung Cha and Chae Won Yu supported resources. Han Beom Park and Hee Chan Lee performed AI-guided structure modeling and peptide prediction. Seung Mo Hong provided interpretation of the mouse histology/pathology data. Suhwan Chang designed and supervised this study, provided funding, and wrote the manuscript. The authors have read and approved the final manuscript.

Funding

Suhwan Chang reports financial support was provided by University of Ulsan College of Medicine. Suhwan Chang reports a relationship with University of Ulsan College of Medicine that includes: If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Data availability

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Declarations

Competing interests

The authors declare no competing interests.

Footnotes

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

  • 1.Siegel RL, et al. Cancer statistics, 2022. CA Cancer J Clin. 2022;72(1):7–33. [DOI] [PubMed] [Google Scholar]
  • 2.Rawla P, Sunkara T, Gaduputi V. Epidemiology of Pancreatic Cancer: Global Trends, Etiology and Risk Factors. World J Oncol. 2019;10(1):10–27. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Yachida S, et al. Distant metastasis occurs late during the genetic evolution of pancreatic cancer. Nature. 2010;467(7319):1114–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Choi YS, et al. Antibody-mediated blockade for galectin-3 binding protein in tumor secretome abrogates PDAC metastasis. Proc Natl Acad Sci U S A. 2022;119(30):e2119048119. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.da Cunha BR, et al. Cellular Interactions in the Tumor Microenvironment: The Role of Secretome. J Cancer. 2019;10(19):4574–87. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Zimmer DB, et al. Evolution of the S100 family of calcium sensor proteins. Cell Calcium. 2013;53(3):170–9. [DOI] [PubMed] [Google Scholar]
  • 7.Bresnick AR, Weber DJ, Zimmer DB. S100 proteins in cancer. Nat Rev Cancer. 2015;15(2):96–109. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Tian C, et al. Proteomic analyses of ECM during pancreatic ductal adenocarcinoma progression reveal different contributions by tumor and stromal cells. Proc Natl Acad Sci U S A. 2019;116(39):19609–18. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Meng Y et al. S100 Calcium-Binding Protein P and Cathepsin E as Key Mediators in Pancreatic Cancer Tumorigenesis. Biomedicines, 2025. 13(11). [DOI] [PMC free article] [PubMed]
  • 10.Low RRJ, et al. S100 family proteins are linked to organoid morphology and EMT in pancreatic cancer. Cell Death Differ. 2023;30(5):1155–65. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 11.Lin H et al. S100A10 Promotes Pancreatic Ductal Adenocarcinoma Cells Proliferation, Migration and Adhesion through JNK/LAMB3-LAMC2 Axis. Cancers (Basel), 2022. 15(1). [DOI] [PMC free article] [PubMed]
  • 12.Chambers AF, Groom AC, MacDonald IC. Dissemination and growth of cancer cells in metastatic sites. Nat Rev Cancer. 2002;2(8):563–72. [DOI] [PubMed] [Google Scholar]
  • 13.Friedl P, et al. Classifying collective cancer cell invasion. Nat Cell Biol. 2012;14(8):777–83. [DOI] [PubMed] [Google Scholar]
  • 14.Cheung KJ, Ewald AJ. A collective route to metastasis: Seeding by tumor cell clusters. Science. 2016;352(6282):167–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Konen J, et al. Image-guided genomics of phenotypically heterogeneous populations reveals vascular signalling during symbiotic collective cancer invasion. Nat Commun. 2017;8:15078. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Cheung KJ, et al. Polyclonal breast cancer metastases arise from collective dissemination of keratin 14-expressing tumor cell clusters. Proc Natl Acad Sci U S A. 2016;113(7):E854–63. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Bronsert P, et al. Cancer cell invasion and EMT marker expression: a three-dimensional study of the human cancer-host interface. J Pathol. 2014;234(3):410–22. [DOI] [PubMed] [Google Scholar]
  • 18.Nakayama H, et al. S100P regulates the collective invasion of pancreatic cancer cells into the lymphatic endothelial monolayer. Int J Oncol. 2019;55(1):211–22. [DOI] [PubMed] [Google Scholar]
  • 19.Sznurkowska MK, Aceto N. The gate to metastasis: key players in cancer cell intravasation. Febs j. 2022;289(15):4336–54. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Reymond N, d’Água BB, Ridley AJ. Crossing the endothelial barrier during metastasis. Nat Rev Cancer. 2013;13(12):858–70. [DOI] [PubMed] [Google Scholar]
  • 21.Fares J, et al. Molecular principles of metastasis: a hallmark of cancer revisited. Signal Transduct Target Therapy. 2020;5(1):28. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Malashkevich VN, et al. Structure of Ca2+-bound S100A4 and its interaction with peptides derived from nonmuscle myosin-IIA. Biochemistry. 2008;47(18):5111–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Watson JL, et al. De novo design of protein structure and function with RFdiffusion. Nature. 2023;620(7976):1089–100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Dauparas J, et al. Robust deep learning-based protein sequence design using ProteinMPNN. Science. 2022;378(6615):49–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 25.Lee H, Park H. Protein folding stability estimation with an explicit consideration of unfolded states. bioRxiv, 2025: p. 2025.02.10.637420. [DOI] [PMC free article] [PubMed]
  • 26.Jumper J, et al. Highly accurate protein structure prediction with AlphaFold. Nature. 2021;596(7873):583–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 27.Alford RF, et al. The Rosetta All-Atom Energy Function for Macromolecular Modeling and Design. J Chem Theory Comput. 2017;13(6):3031–48. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Frenz B, et al. Prediction of Protein Mutational Free Energy: Benchmark and Sampling Improvements Increase Classification Accuracy. Front Bioeng Biotechnol. 2020;8:558247. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Chen W et al. High-throughput image analysis of tumor spheroids: a user-friendly software application to measure the size of spheroids automatically and accurately. J Vis Exp, 2014(89). [DOI] [PMC free article] [PubMed]
  • 30.Grasset EM, et al. Matrix Stiffening and EGFR Cooperate to Promote the Collective Invasion of Cancer Cells. Cancer Res. 2018;78(18):5229–42. [DOI] [PubMed] [Google Scholar]
  • 31.Gaggioli C, et al. Fibroblast-led collective invasion of carcinoma cells with differing roles for RhoGTPases in leading and following cells. Nat Cell Biol. 2007;9(12):1392–400. [DOI] [PubMed] [Google Scholar]
  • 32.Miyazaki K, et al. Collective cancer cell invasion in contact with fibroblasts through integrin-α5β1/fibronectin interaction in collagen matrix. Cancer Sci. 2020;111(12):4381–92. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 33.Zhang J, et al. Energetic regulation of coordinated leader-follower dynamics during collective invasion of breast cancer cells. Proc Natl Acad Sci U S A. 2019;116(16):7867–72. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 34.Coetzee AS, et al. Nuclear FGFR1 promotes pancreatic stellate cell-driven invasion through up-regulation of Neuregulin 1. Oncogene. 2023;42(7):491–500. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 35.Lee J-H, et al. Microfluidic co-culture of pancreatic tumor spheroids with stellate cells as a novel 3D model for investigation of stroma-mediated cell motility and drug resistance. J Experimental Clin Cancer Res. 2018;37(1):4. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 36.Kuo CT, et al. Three-dimensional spheroid culture targeting versatile tissue bioassays using a PDMS-based hanging drop array. Sci Rep. 2017;7(1):4363. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 37.Moissoglu K, Lockett SJ, Mili S. Visualizing and Quantifying mRNA Localization at the Invasive Front of 3D Cancer Spheroids. Methods Mol Biol. 2023;2608:263–80. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 38.Taichman DB, et al. Tumor cell surface alpha 4 beta 1 integrin mediates adhesion to vascular endothelium: demonstration of an interaction with the N-terminal domains of INCAM-110/VCAM-1. Cell Regul. 1991;2(5):347–55. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Strell C, Entschladen F. Extravasation of leukocytes in comparison to tumor cells. Cell Commun Signal. 2008;6:10. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 40.Knapinska AM, Estrada CA, Fields GB. The Roles of Matrix Metalloproteinases in Pancreatic Cancer. Prog Mol Biol Transl Sci. 2017;148:339–54. [DOI] [PubMed] [Google Scholar]
  • 41.Slapak EJ et al. Matrix Metalloproteases in Pancreatic Ductal Adenocarcinoma: Key Drivers of Disease Progression? Biology (Basel), 2020. 9(4). [DOI] [PMC free article] [PubMed]
  • 42.Ahn B, Kim JY, Hong SM. Combined Infiltrative Macroscopic Growth Pattern and Infiltrative Microscopic Tumor Border Status Is a Novel Surrogate Marker of Poor Prognosis in Patients With Pancreatic Neuroendocrine Tumor. Arch Pathol Lab Med. 2023;147(1):100–16. [DOI] [PubMed] [Google Scholar]
  • 43.Ayres Pereira M, Chio IIC. Metastasis in Pancreatic Ductal Adenocarcinoma: Current Standing and Methodologies. Genes (Basel), 2019. 11(1). [DOI] [PMC free article] [PubMed]
  • 44.Park W, Chawla A, O’Reilly EM. Pancreatic Cancer: Rev Jama. 2021;326(9):851–62. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 45.Cheung KJ, et al. Collective invasion in breast cancer requires a conserved basal epithelial program. Cell. 2013;155(7):1639–51. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Fayad R, et al. EFA6B regulates a stop signal for collective invasion in breast cancer. Nat Commun. 2021;12(1):2198. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 47.Nguyen DT, et al. A biomimetic pancreatic cancer on-chip reveals endothelial ablation via ALK7 signaling. Sci Adv. 2019;5(8):eaav6789. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Hong SM, et al. Three-dimensional visualization of cleared human pancreas cancer reveals that sustained epithelial-to-mesenchymal transition is not required for venous invasion. Mod Pathol. 2020;33(4):639–47. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Herwig N, Belter B, Pietzsch J. Extracellular S100A4 affects endothelial cell integrity and stimulates transmigration of A375 melanoma cells. Biochem Biophys Res Commun. 2016;477(4):963–9. [DOI] [PubMed] [Google Scholar]
  • 50.Hernández JL, et al. Therapeutic targeting of tumor growth and angiogenesis with a novel anti-S100A4 monoclonal antibody. PLoS ONE. 2013;8(9):e72480. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Ochiya T, Takenaga K, Endo H. Silencing of S100A4, a metastasis-associated protein, in endothelial cells inhibits tumor angiogenesis and growth. Angiogenesis. 2014;17(1):17–26. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 52.Dahlmann M, et al. RAGE mediates S100A4-induced cell motility via MAPK/ERK and hypoxia signaling and is a prognostic biomarker for human colorectal cancer metastasis. Oncotarget. 2014;5(10):3220–33. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Tang H, et al. CAFs-derived LAM332 promotes CTCs formation and survival via ITGA3 and contributes to the metastasis of pancreatic ductal adenocarcinoma. Cell Death Dis. 2026;17(1):369. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Kim SJ, et al. Pancreatic ductal adenocarcinoma with a predominant large duct pattern has better recurrence-free survival than conventional pancreatic ductal adenocarcinoma: a comprehensive histopathological, immunohistochemical, and mutational study. Hum Pathol. 2022;127:39–49. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Material 1. (2.4MB, pdf)
Supplementary Material 2. (663.2KB, pdf)

Data Availability Statement

All data supporting the findings of this study are available within the paper and its Supplementary Information.

Articles from Cell Communication and Signaling : CCS are provided here courtesy of BMC

RESOURCES