PDAC organoids were acquired from a tumor resection performed with curative intent at the University and Hospital Trust of Verona. Written informed consent from the donors for research use of the tissue was obtained prior to acquisition of the specimens. Tissues for the generation of models were collected under protocol number 55859, approved by the local Ethics Committee (Comitato Etico Azienda Ospedaliera Universitaria Integrata) to V.C. (Prog. 3456CESC, 27/09/21). The organoids were cultured in a 6-well plate in 10 μl Matrigel (Corning^®, 356231, 8.3-10.5 mg/ml) droplets. After seeding, the plate was placed in the incubator with bottom-side up for 15 min. For culturing, hCPLT medium was used, which consists of Advanced Dulbecco’s Modified Eagles Medium with Nutrient Mixture F-12 Hams – 500ml (Gibco, #12634-028), GlutaMax – 200mM (Gibco, # 35050-061), Hepes – 1M (Gibco, # 15630-080), Primocin – 50mg/ml (Invivogen # ant-pm-2), B-27 supplement (Gibco, # 17504-044), N-Acetylcysteine (NAC) (Sigma-Aldrich, # A9165-5G), Nicotinamide (Sigma-Aldrich, # N0636), hEGF (Gibco, # PMG8043), TGFβ Receptor inhibitor A83-01 (Tocris, # 2939), FGF10 (Peprotech, # 100-26), mNoggin (Peprotech, # 250-38), R-spondin-1 conditioned medium, Wnt3a conditioned medium, Gastrin (Tocris, # 3006) and Y-27632 Dihydrochloride (Sigma, # Y0503). The cells were harvested using Cell Recovery Solution (Corning^®, 354253) and placed on ice for 30 min. The organoids were spun down at 300 g for 5 min, media was removed and incubated with TrypLE Express for 3 min (Gibco, # 12605-028) in the waterbath for enzymatic disruption. The organoids solution was spun down, resuspended in ice-cold media, counted and used.

PSCs (Klon 2.2) were obtained from Marburg University (15) and cultured in DMEM medium (Gibco, 10829-018) with 1% Penicillin/Streptomycin and 10% Fetal Bovine Serum (FBS). Cells were passaged at 80% confluency using Trypsin 2.5% (ThermoFisher, 15090046).

Human Umbilical Vein Endothelial Cells (HUVEC) (Lonza, C2519AS) were cultured in Endothelial Growth Medium (EGM-2) (Lonza, CC-3156) with 1% Penicillin/Streptomycin and 2% FBS (Gibco, 16140-071). The cells were seeded immediately after thawing.

PBMCs were derived from whole blood-derived buffy coats of healthy donors provided by Sanquin. Ficoll-Paque PLUS (15 ml) was added to a 50 ml Leucosep tube and centrifuged at room temperature (RT) for 30 min at 1000 x g. The blood-derived buffy coat was diluted 1:2 with sterile PBS and 25 ml of the diluted buffy coat was added to the Leucosep tube. The tube was centrifuged at RT for 30 min at 800 x g and plasma was carefully removed by aspiration. The PBMC layer was transferred to a 50 ml conical tube and cells were subsequently washed twice with PBS and centrifuged at RT for 10 min at 300 x g. Cell density was assessed with a cell counter.

After isolation, PBMCs were frozen and thawed upon use. For isolation of immune cell types, PBMCs were thawed and different immune cells isolated with their respective kits according to manufacturer’s protocol: EasySep Human B Cell Isolation Kit (StemCell Technologies, #17954), EasySep Direct Human T Cell Isolation Kit (StemCell Technologies, #19661), EasySep Human CD4+ T Cell Isolation Kit (StemCell Technologies, #17952), EasySep Human CD8+ T Cell Isolation Kit (StemCell Technologies, #17953), EasySep Direct Human Monocyte Isolation Kit (StemCell Technologies, #19669).

All cells were cultured in a humidified incubator at 5% CO₂ and 37°C, and regularly tested for mycoplasma.

PDAC models were established in the OrganoPlate 3-lane (Mimetas, the Netherlands), a microfluidic platform, based on a 384-well plate format with 40 microfluidic chips. The OrganoPlate 3- lane consists of two perfusion lanes, and a middle lane used for ECM gel filling. Culture medium is added to the chips through the perfusion inlets and outlets. Cultures are monitored through the observation window.

In the OrganoPlate 3-lane, both Matrigel and Collagen type I (5 mg/ml, AMSbio, #3447-020-01) were used as ECMs. Collagen type I was neutralized in an 8:1:1 ratio with 1 M Hepes buffer (Gibco, #15630-056) and 37 g/L NaHCO3 (sigma, S5761) reaching a final concentration of 4mg/ml. The collagen mix was kept on ice and seeded within 10 min. PSCs were counted and the appropriate number of cells was resuspended in Collagen type I. ECM (2 μl) was loaded into the gel channel and incubated for 15 min at 37°C and 5% CO2 to allow for polymerization of the gel. After incubation, 50 μl of DMEM medium was added to the gel-inlet to prevent the gel from drying out. Next, HUVECs were seeded (10 000 cells per chip) in 2 μl of EGM-2 medium in the top perfusion channel, 50 μl of the same medium was added to the top medium inlet. The plate was incubated vertically in an angle of 75°, with the gel channels facing downwards to allow the cells to attach to the ECM gel for 2 hours. After 2 hours, 50 μl of medium was added to the outlets of the gel and top perfusion channel. Subsequently, PDAC organoids were resuspended in 2 μl Matrigel and seeded in the bottom perfusion channel. The plate was incubated for 15 min until hCPLT medium was added to the bottom lane onto the PDAC organoids compartment. After that, plates were placed in the incubator on a rocking platform (OrganoFlow, Mimetas, the Netherlands) at an inclination of 7° and an interval of 8 min.

All media was removed from the plate and 80 μl/chip dye was prepared (FITC Dextran 150 kDa (Sigma, Cat#: 46946)). 20 μl of medium was pipetted into the gel and medium in- and outlets and 40 μl of the dye solution was added to the lane containing the endothelial tube in the inlet and 30 μl to the outlet. Leakage of the fluorescent dye from the lumen of the endothelial vessel into the rest of the chip was imaged using the ImageXpress XLS Micro (Molecular Devices). Images were taken for 14 minutes with a 2-minute interval. The images were analysed by extracting the average fluorescence values of the top perfusion channel divided by the average fluorescence value of the bottom perfusion channel for each chip and timepoint, determined in Fiji. The apparent permeability (P_app) value (cm/s) was determined:

ΔC_receiver is the difference between the fluorescence intensity measured in the bottom perfusion channel between t=0 and t=14 min, V_receiver is the volume of the measured region in the ECM channel, Δt is the time between start of the assay and endpoint (14 min), A_barrier is the surface of the ECM barrier with the upper perfusion channel (0.0057 cm²), and C_donor describes the fluorescence intensity measured in the top perfusion channel.

The electrode board was prepared and the OrganoPlate was equilibrated at RT for 30min. The electrode board of the TEER device (Mimetas, the Netherlands) was placed on top of the plate and TEER measurement was performed.

5 μM MitoTracker Deep Red was added to the PBMCs in AIM-V medium. The cells were incubated in the dark at 37°C for 30 min. After the incubation, PBMCS were washed with medium, resuspended in the appropriate volume of medium and seeded in the top lane of the OrganoPlate upon exchanging EGM-2 medium of HUVECs with a 50:50 mixture of AIM-V and EGM-2 containing the PBMCs.

Imaging of the migration was done with the ImageXpress® Micro XLS confocal microscope (Molecular Devices). Montages were created using Fiji. A migration quantification tool made in Fiji, specifically developed for the OrganoPlate 3-lane was used.

The content of the OrganoPlates was fixed with 3.7% Formaldehyde (Sigma, # 252549-1L) in HBSS (+Ca/Mg) (Sigma, # 55037C-1000ML) for 15 min. The plates were washed twice with PBS (Gibco, t# 70013065) for 5 min each. PBS was added to all chips, the plate was sealed and stored until used for immunostaining. For staining, the plates were kept on a rocking device during all incubation steps. The cells were first washed for 5 min with washing solution containing 4% FCS (Gibco/ATCC, cat# A13450) in PBS, permeabilized for 10 min with permeabilization buffer containing 0.3% Triton X-100 (Sigma, #T8787) in PBS and washed again for 5 min. The cells were then blocked with 2% FBS, 2% BSA (Sigma, # A2153) and 0.1% Tween20 (Sigma, # P9616) in PBS for 45 min. The primary antibody was prepared in blocking solution and added to the plate for 24h at RT. CD31 (Dako, #M0823) and ICAM-1 (Biotechne, #BBA3) were used in a 1:100 dilution. Alexa Fluor Plus 488 (Invitrogen, #A32723) was used as a secondary antibody in a 1:250 dilution. The secondary antibody was prepared in blocking solution and added to the plate after washing the plate twice for 3 min and incubated for 24 h. The plate was washed again twice for 3 min after incubation. The cells were washed with PBS once for 1 minute, incubated with NucBlue™ Live ReadyProbes™ Reagent (Hoechst 33342) (Invitrogen, R37605) and ActinGreen™ 488 ReadyProbes™ Reagent (Invitrogen, R37110) and the plate was filled with PBS, sealed and kept in the fridge until imaging.

Calcein-AM (Lifetechnologies, #C3099), NucBlue Live ReadyProbes Reagent (Life technologies, #R37610) and DraQ7 (BioStatus, #DR71000) were used for staining live and dead cells and the nucleus. The reagents were added to the medium and distributed to the perfusion inlets and outlets. The mix was incubated for 45 min on the rocker and fluorescent imaging was performed.

Medium was sampled from the OrganoPlate and added to the pre-ordered plate containing the analyte-specific capture antibodies, which bind to the analytes of interest according to the kit protocol. The Human Magnetic Luminex^® Assay was used (bio-techne, #LXSAHM-11). Samples were analyzed on the MAGPIX xPONENT^® software. The samples were normalized and compared to standards.

For targeting the stroma, several compounds were added as 2 µM solutions to the OrganoPlate after medium was removed on day 4. The following compounds were used: Halofuginone (MedChem Express, #HY-N1584) as a PSC/CAF and SMAD 2/3 inhibitor; Galunisertib (Selleck Chemicals, #S2230) as TGF-β receptor inhibitor; Vismodegib (MedChem Express, #HY-10440) as a SHH pathway inhibitor, Pirfenidone (MedChem Express, #HY-B0673) as a cell cycle inhibitor of CAFs; and Losartan (MedChem Express, #HY-17512) as a TGF-β ligand inhibitor (16).

All statistical analyses were conducted in GraphPad Prism version 9 (GraphPad Software, San Diego, CA, USA) and data was presented as mean ± standard deviation (SD). Differences in survival were assessed using one-way or two-way ANOVA in combination with respective Tukey’s multiple comparison test or Sidak`s multiple comparison. Luminex data was analysed using Kruskal-Wallis test, correction for multiple comparisons by Dunn’s test. A statistical significance of p ≤ 0.05 was maintained. The significances are shown as asterisks in the figures (* = p< 0,05; ** = p< 0,01; *** = p< 0,001; **** = p< 0,0001). Independent experiments are denoted by N, while replicates per experiment are denoted by n. Sample size was chosen based on the variation and standard deviation between samples to ensure significance of the data. F-tests, descriptive statistics and row analysis were performed to ensure similar variance between the groups.

The stroma is considered the major player in shaping the PDAC immune microenvironment. In order, to recapitulate cellular interactions observed in PDAC tumors, we developed a microfluidic-based PDAC model and subsequently applied it in immune migration studies. PDAC organoids and PSCs were characterized and model setups were established in the OrganoPlate 3-lane ( Figures S1 , 1A , B ). These consist of PDAC organoids, PSCs, endothelial and immune cells. First, the middle lane was seeded with Collagen I type ECM containing PSCs. Subsequently, the top lane was seeded with endothelial HUVEC cells, which self-assembled into a tubule under flow conditions. PDAC organoids were cultured in Matrigel in the bottom lane ( Figure 1C ). All cell types were seeded on the same day. After cell seeding, cultures were then placed onto an OrganoFlow for perfusion, which enabled nutrient distribution and waste removal ( Figures 1D, E ). Cultures were allowed to develop for 4 days. For migration studies, the HUVEC tubule was loaded with CMRA labeled PBMCs and the migration towards the PDAC organoids was followed for 72h, cultures were imaged every 24h. Migration of PBMCs was analyzed using an in-house cell counting tool developed in FIJI.

This model setup and variations of it were further used to investigate the role of the tumor microenvironment, particularly the endothelium and PSCs, on immune cell recruitment in PDAC.

Inflammation occurs in response to tissue damage and cancer, which usually results in vascular activation and increased recruitment of immune cells towards the site of inflammation (17). Vascular responses such as changes in barrier function were studied in inflamed HUVEC (exposed to 2.25 ng/ml TNFα) as well as in presence of a PDAC tumor compartment.

HUVEC control tubules showed expected morphology as shown by CD31 immunostaining (18). In Figure 2 , TNFα treated HUVEC tubules show clear increased ICAM expression ( Figure 2A ) and permeability ( Figures 2B–D ), consequently, leading to a better attachment of PBMCs and subsequent migration ( Figures 2E, F ).

However, HUVEC tubules presented a very poor morphology and a very high leakiness in response to TNFα, for this reason migration experiments were further conducted in absence of TNFα. Although, untreated HUVEC tubules in co-culture with PDAC organoids are more organized, these also present a poor barrier function in presence of PDAC organoids, therefore, recapitulating the leaky blood vessels observed in PDAC tumors ( Figures 2D , 3B ).

After confirming that the endothelium did not retain PBMCs or formed a good barrier in presence of PDAC organoids, the role of the PSCs was further evaluated. To bring more complexity into the model the HUVEC-PDAC model was expanded to include PSCs in the middle lane and create a HUVEC-PSC-PDAC model.

PSCs grew along the HUVEC tubule, building an additional barrier ( Figures 3A–C ). Cultures were characterized by actin and nuclei staining and respective 3D reconstruction images ( Figure 3A ) showed a stromal compartment formed alongside a fully developed HUVEC tubule. TEER and BI measurements in different model set ups indicated, that mainly PSCs were responsible for barrier formation ( Figures 3B, C ). Next, PBMCs were allowed to migrate in different culture setups. In presence of both PSCs and PDAC, 20-30% of PBMCs were retained in the stromal compartment ( Figures 3D, E ).

PBMC migration was not observed in HUVEC only controls ( Figure 3F ). PBMCs migrated within 24-72h in the different model setups tested. The highest numbers of PBMCs to reach the PDAC organoids compartment was observed in the model setup containing PDAC and PSCs in the bottom lane ( Figure 3G ).

To further explore the relevance of the stroma in attracting and influencing the distribution of immune cells in our model, chemokine, cytokine and FGF2 levels were determined in culture supernatants, collected from specific chip compartments, using a Luminex panel containing CCL2, CXCL1, CCL4, CXCL10, CXCL13, IL-6, Il-8, IL-10, TNFα, FGF2 and IFN-y ( Figure 4 ).

Data are shown as fold change of PDAC organoids compartment in absence of PSCs. Almost all secreted factors were present in the supernatant of the different culture setups, except for INFy and IL10. These cytokines were present in very low concentrations and no significant differences were observed (data not shown). CCL2 and CXCL13 ( Figures 4A , E ) were increased in the PSC compartment in presence of PDAC organoids (adjacent channel) compared to PDAC organoids growing in absence of PSCs. TNFα was only increased in the PDAC organoids compartment of the co-culture ( Figure 4H ). CXCL1, CCL4, IL8 and IL6 were significantly increased in the PSC and PDAC compartments of the co-culture ( Figures 4B–D, G ). CXCL10 and FGF2 data showed a trend towards increased levels in the co-cultures ( Figures 4F, I ). FGF2 level was significantly increased in the PSC compartment of the co-culture in comparison to PSCs growing in absence of PDAC organoids ( Figure 4I ). Heatmap in Figure 4J summarizes the Luminex data and indicate that compartments of the co-culture setup show an increase in the secretion of most soluble factors measured. These results indicate that the biochemical microenvironment in our model is complex, and that the PDAC organoids-PSCs crosstalk lead to the increase of immune mediators and being those changes sometimes compartment dependent. In addition, these support the notion that PSCs in addition to a physical barrier, seem to shape a biochemical immune microenvironment as well.

PSCs seem to function as a physical barrier as well as contribute with secreted factors that influence PBMCs migration and distribution in vitro preventing them in part from reaching the tumor cells. We next characterized the migratory behavior of isolated PBMC subtypes. T-cells were least likely to migrate towards PDAC organoids, whereas monocytes and B-cells were most likely to migrate to this compartment ( Figure 5A ). Total cell numbers showed, that around 40% of the total B-cell and monocyte population seeded, migrated towards to the organoids, whereas this percentage was much lower for T-cells ( Figure 5B ).

To increase the amount of migrating PBMCs and consequent immune infiltrate into the tumor compartment, several stromal targeting compounds were tested. Halofuginone (PSC/CAF and SMAD2/3 inhibitor), Galunisertib (TGF-β receptor inhibitor), Pirfenidon (cell cycle inhibitor of CAFs), Losartan (TGF-β ligand inhibitor) and Vismodegib (SHH pathway inhibitor) were selected to target the PSCs compartment ( Figure 6 ).

Exposure of the stromal compartment to these compounds decreased TEER, confirming that all compounds influenced PSCs barrier. Halofuginone created the biggest drop in TEER ( Figure 6A ). This effect was accompanied by a significant increase in immune cell infiltration in response to Halofuginone, Losartan and Pirfenidone (24-72h). Migration of PBMCs was slightly increased to the PDAC organoids in the bottom lane and the effect reduced over time ( Figure 6B ). Halofuginone produced the most consistent increase in immune cell infiltration. This seemed to be associated to the induction of pancreatic stellate cell death as determined with a live and dead assay ( Figures 6C, D ).