Organoids Derived from Neoadjuvant FOLFIRINOX Patients Recapitulate Therapy Resistance in Pancreatic Ductal Adenocarcinoma

Pancreatic ductal adenocarcinoma (PDAC) is an exceptionally lethal cancer. Most PDAC patients present with metastatic disease at the time of diagnosis, rendering them ineligible for surgical resection. The average PDAC tumor progression time from resectable to advanced stage is about one year (1). Consequently, most PDAC patients succumb to the disease within a year of diagnosis.

Gemcitabine was the standard treatment for pancreatic patients for more than a decade (2, 3). A more recent chemotherapy regimen, FOLFIRINOX (comprising 5-fluorouracil, irinotecan, and oxaliplatin), improved the overall and progression-free survival when compared with gemcitabine—particularly for patients with advanced disease and in the adjuvant setting (4, 5).

FOLFIRINOX treatment often leads to a stabilization of the disease, or partial response occurs in a minority of the PDAC patients, while others do not respond by showing disease progression during treatment (6). However, even among the responders, most patients develop rapid therapy resistance, resulting in disease progression after treatment. In the clinic, tumor progression following initial regression to chemotherapy presumably originates from the failure of complete eradication of all subclones of cancer cells by current chemotherapy regimens. Such incomplete eradication alters the selection pressure acting on subclones that ultimately favor the outgrowth of resistant clones (7, 8).

Underlying mechanisms of chemoresistance include alterations in several cellular processes, such as metabolic pathways, epithelial–mesenchymal transition (EMT), and angiogenesis resulted in a chemorefractory phenotype in PDAC (9–12). Additionally, stromal tumor components such as cancer-associated fibroblasts and tumor-associated macrophages may promote tumor resistance to chemotherapy (13).

Genomic studies support the notion of homogeneity in disease-initiation events identifying common mutations in four genes: KRAS, TP53, CDKN2A, and SMAD4 (14). Despite this uniformity in aggressiveness and initiating genetic events, PDAC patients display extensive heterogeneity in the response to treatment. This highlights the prominent role of phenotypic alterations in the development of chemoresistance in pancreatic cancer cells (15).

Understanding the PDAC tumor evolution will be pivotal in delineating mechanisms of chemoresistance and developing strategies to address them. Currently, most studies focused on drug resistance to gemcitabine (16). However, the introduction of the standard of care involving the FOLFIRINOX regimen warrants research into the therapy resistance of PDAC to this novel chemotherapeutic regimen.

In the last decade, comprehensive 3D models of various tumors have been developed recapitulating the biology of primary tumors (17, 18). These patient-derived organoids demonstrated great potential as a research model to predict therapy response in pancreatic cancer and other gastrointestinal cancers (19, 20). However, until now treatment response of residual tumor cells after neoadjuvant FOLFIRINOX has not been investigated in such a 3D patient-derived organoid model. In this study, we generated organoids from resected tumor material of PDAC patients, with or without neoadjuvant treatment before surgery. We investigated the sensitivity/resistance of organoid lines to the individual drugs of the FOLFIRINOX regimen. We demonstrated the utility of organoids from neoadjuvant treated patients as a preclinical model to discover potential targets of therapy failure in PDAC patients.

Fresh tumor tissues were obtained from PDAC patients operated in Erasmus MC between 2019 and 2020. This study was performed according to the principles of the Declaration of Helsinki, approved by the local medical ethics committee (MEC-2015-085) and all patients provided written informed consent. Patients included in the study completed eight cycles of FOLFIRINOX chemotherapy before resection or were treatment naïve at the time of surgery.

Tumor tissues were collected and minced into 2–3 mm³ fragments in basic medium [AdDMEM/F12 (Gibco), 1X Hepes (Gibco, #15630), 1X glutamax (Gibco, #35050061), 100 μg/mL piromocin (InvivoGen, #ant-pm-1)]. Tissue fragments were digested in basic media supplemented with 1 mg/mL liberase (TH; Roche, #5401135001), 10 μg/mL DNase I (Sigma, #4263), and 10 μmol/L ROCK inhibitor (Tocris, #1254) for 30 minutes in 37°C.

Digested tissues were filtered and washed with phosphate-buffered saline containing 5% serum to deactivate the enzymes. Pancreatic cells were mounted in matrigel (Corning, #356231) and were cultured in serum-free organoid growth media composed of basic medium supplemented with 1 mmol/L N-acetyl L-cysteine (Sigma, #A9165), 1 mmol/L nicotinamide (Sigma, #N0630), 100 ng/mL mouse recombinant NOGGIN (PeproTech, #250-38), 500 ng/mL recombinant RSPONDIN (PeproTech, #120–38), 25% serum-free WNT conditioned media (21), 50 ng/mL mouse recombinant EGF (Gibco, #PMG8041), 100 ng/mL FGF10 (PeproTech, #100-26), 10 nmol/L gastrin I (Sigma, #G9020), 500 nmol/L A83-01 (Tocris, #2939), and 10 μmol/L Y-27632 dihydrochloride (Tocris, #1254).

We used the AllPrep DNA/RNA mini kit (QIAGEN) to extract both DNA and RNA from organoid samples. Organoids were extracted from matrigel domes using Cell Recovery Solution (Corning, #354253) before adding lysis buffer.

Snap-frozen tumor tissues were used for RNA extraction. Tumor tissues were homogenized [Ultra-Turrax T25 homogenizer (Janke & Kunkel)], and lysed simultaneously. RNA was extracted using the AllPrep DNA/RNA mini kit (QIAGEN).

Paraffin-embedded primary tumor tissue blocks were used for DNA extraction. Five 10-μm-thick FFPE tissue sections were deparaffinized and hematoxylin-stained. Tumor areas were selected and microdissected manually into 5% Chelex 100 Resin (Bio-Rad) and cell lysis solution (Promega), using sterile scalpels. DNA was extracted by proteinase K (Roche) digestion by overnight incubation at 56°C.

Extracted DNA was analyzed with next-generation sequencing (NGS) using a panel of 58 cancer-associated genes, including driver genes frequently mutated in PDAC patients (ref. 12; Supplementary Table S1). Library preparation was performed using the Ion AmpliSeq library kit Plus (Thermo Fisher Scientific) with two multiplex PCR reactions (10 ng of DNA per reaction). The Ion AmpliSeq kit protocol was applied to generate amplicon libraries. Libraries were sequenced on the Ion S5XL Semiconductor Sequencer (Thermo Fisher Scientific).

Variants were detected and analyzed by VARIANT CALLER v5.10.0.18 (Thermo Fisher Scientific) and annotated in a local galaxy pipeline (www.galaxyproject.org) using ANNOVAR. Exonic or splice-site variants with at least 5% frequency were selected for analysis. Synonymous point mutations, non-frameshift alterations, and mutations with frequencies greater than 1% in the 1000 Genome database (considered common polymorphisms) were filtered out.

Organoids were cultured in the same culture condition till confluency. Thereafter, organoids were digested with TrypLE (Gibco 12604-013) for 5 to 10 minutes at 37°C, assisted with mechanical digestion into single cells. 1,000 cells were seeded in 5 μL matrigel domes in a clear-bottomed 96-well plate (3707, Corning). When mature organoids were formed, chemotherapeutic drugs were added. Oxaliplatin (ACCORD, infusion; six doses in the range of 1.2 × 10^–6 mol/L to 4.02 × 10^–5 mol/L), 5-FU (ACCORD, infusion; six doses in the range of 3.8 × 10^–6 mol/L to 1.2 × 10^–4 mol/L), SN-38 (TOCRIC #2684; six doses in the range of 5 × 10^–10 mol/L to 1.6 × 10^–8 mol/L), gemcitabine (SANDOZ, infusion; six doses in the range, 7 × 10^–10 mol/L to 2.4 × 10^–8 mol/L) were tested. FOLFIRINOX regimen was given in five different concentrations by combining the three chemotherapeutic compounds ranging from 6 × 10^–7 mol/L to 1 × 10^–5 mol/L for oxaliplatin, 1.9 × 10^–6 mol/L to 3 × 10^–5 mol/L for 5-FU, and 1 × 10^–10 mol/L to 5 × 10^–9 mol/L for SN-38. Because FOLFIRINOX combination is more potent than monotherapies of its individual components, the range of concentrations was centered around IC25 of its components. FOLFIRINOX drug concentrations represent summed concentrations of its individual components. Cell viability was obtained at baseline day 0 before adding chemotherapeutics and 72 hours after incubation with chemotherapy drugs [CellTiter-Glo 3D (#G9681, Promega); SpectraMax I3 plate reader]. Growth rate values (GR) were obtained by calculation of the baseline median luminescence values of day 0, untreated, and drug-treated day 3 wells (22). Drug concentrations were log transformed, and GR value of each data point was plotted to obtain dose–response curves. GR₅₀ values were obtained, and response between two groups was compared using an unpaired t-test with Welch correction (GraphPad Prism 9.0).

For processing organoids, full matrigel domes were fixed in 2% paraformaldehyde and 0.2% glutaraldehyde. The matrigel domes were mounted with low melting agar and kept in 4°C before processing for paraffin sectioning. Matrigel-agar blocks were paraffin-embedded, and 4-μm sections were mounted on Superfrost plus slides (Menzel-Gläser), stained with hematoxylin and eosin (H&E; HE600, Ventana, #06917259001) and evaluated by a clinical pathologist.

IHC staining for keratin-19 (CK19) and transcription factor SOX-9 was performed on 4-μm-thick paraffin sections by automated IHC using the Ventana Benchmark ULTRA (Ventana Medical Systems Inc). Sequential 4-μm-thick (FFPE) sections were stained using the Ultraview universal DAB detection kit (#760-700, Ventana).

Antibodies used included 0.11 mg/mL concentration of CK19 mouse monoclonal antibody (clone: A53-B/A2.26; Cell Marque) and 1/750 concentration of SOX-9 anti-goat polyclonal antibody (R&D Systems).

Poly-A–enriched directional RNA-seq libraries were prepared (Kapa mRNA HyperPrep), and subsequent paired-end read sequencing was performed using the Novaseq 6000 (Illumina; 2 × 150 cycles, v1.0 chemistry). After quality control with FastQC, raw sequencing reads were trimmed used Trimmomatic (v 0.36; ref. 23) and aligned to the human reference genome hg38 with TopHat (v 2.0.9). The remaining reads were counted through HTSeq (v 0.6.0; ref. 24). We performed further downstream analyses using Python (NumPy and Pandas packages), R studio (ggplot, DESeq2), Perseus (v 1.5.5.3; ref. 25), Google Sheets, and GraphPad Prism (v 8.3.0). The RNA-seq data set consisting of normalized transcripts per million reads (TPM) for 60,583 genes was log₂ transformed and grouped into 10 different groups, each containing global transcriptomes of two samples (organoid and their respective primary tumor). We filtered the data set to retain genes with one or more read counts (in all 20 samples), resulting in a final matrix consisting of 17,848 genes. Heatmaps display hierarchically clustered genes after preprocessing the data sets with k-means (Euclidean average distance defining the separation between two clusters). To identify differentially expressed genes (DEG), we separated the organoid and primary tumor data and used a combination of two filtering criteria: the first being significantly responding (DESeq2; P < 0.1), and the second having a minimum TPM raw (non–log-transformed) value of at least 1 in all samples. Pathway analysis was performed using the Reactome and enrichR databases (reactome.org; P < 0.05).

RNA-seq data are available in the GEO repository with the identifier GSE169321.

We have generated a library of 10 organoids from resectable and borderline resectable PDAC patients who received complete eight cycles of neoadjuvant FOLFIRINOX chemotherapy or from patients who were treatment naïve at the time of resection (Fig. 1A). The 10 tumor tissues were collected consecutively in time. Clinical characteristics of patients are shown in Table 1.

All the organoid lines were cultured with organoid growth media containing essential growth factors as described previously in the literature with some optimization (26, 27). To avoid undefined components in the culture medium, we obtained recombinant RSPONDIN and NOGGIN proteins rather than using conditioned media. Besides, by the use of serum-free WNT conditioned media, we reached the serum content of the medium to zero avoiding the introduction of undefined proteins into the culture. Finally, neoadjuvant FOLFIRINOX organoids tolerated withdrawal of FGF10 growth factor that is essential for the growth of treatment-naïve organoids. Neoadjuvant FOLFIRINOX-treated organoid cultures were successfully established growing at least 10 passages with an efficiency of 83%, whereas the efficiency was 62% for treatment-naïve organoids.

Organoid cultures carried known driver mutations of PDACs evaluated by a panel of cancer-associated genes recapitulating primary tumor origin (Fig. 1B; Supplementary Table S1). Our results illustrate that the expected prevalence of driver mutations for PDAC is conserved in our organoid cultures with no significant differences between treatment-naïve and FOLFIRINOX-treated organoid lines (Fig. 1B and C).

Next, we evaluated whether tumor organoids recapitulate histopathologic characteristics of original tumor tissue. Both neoadjuvant FOLFIRINOX-treated and treatment-naïve pancreatic organoids exhibited mixed morphologies consisting of filled and empty lumen shapes with different degrees of filling (Fig. 2A). Although the morphology of the organoids differs between patients, we did not observe any significant morphologic selection associated with FOLFIRINOX treatment (Fig. 2A). A high degree of intratumor heterogeneity was observed in primary tumor lesions (Fig. 2A, bottom), which is conserved in the corresponding organoid cultures (Fig. 2A, top and middle).

Additionally, organoids recapitulate specific features of primary tumor tissue by conserving the expression of epithelial markers (CK19) and expression of pancreatic lineage markers (SOX9). As shown in Fig. 2B, expression of CK19 and SOX9 is highly preserved in both treatment-naïve and FOLFIRINOX-treated samples.

To further investigate whether organoids recapitulate their corresponding primary tumor tissue at the molecular level, we performed mRNA sequencing on 10 tumor organoid cultures (treatment-naïve: O1-O5 and FOLFIRINOX-treated: O6-O10) and 10 corresponding matched tumor tissues (T1-T5 and T6-T10).

Gene-expression profiles were strongly correlated between organoid lines and their corresponding tumor tissues, with a median Pearson correlation coefficient of 0.85 (range, 0.81–0.88; Fig. 3A). Principal component analysis showed dispersed and heterogeneous transcription profiles among organoids with a partial separation between treatment-naïve and FOLFIRINOX-treated organoids (Fig. 3B). Tumor samples were separated from organoids. Genes that are differentially expressed between paired organoids and tumor tissues are thought to be mainly related to stromal components, or immune cells that are presented in primary tumor lesions but absent in organoids. Further, organoids are exposed to artificial culture conditions to promote cell growth and proliferation. To elucidate these differences, differential gene-expression analysis was performed between cancer organoids and paired tumor tissues. About 2,256/17,848 genes were differentially expressed between organoids and tumor tissues. As shown in the heatmap, most DEGs are highly expressed in primary tumor lesions compared with organoids (Fig. 3C). Upregulated genes were enriched for pathways predominantly involved in the tumor microenvironment, immune response, and extracellular matrix (Fig. 3C, top right). Whereas downregulated genes were enriched for pathways related to DNA synthesis, mitosis, and proliferation of cancer cells (Fig. 3C, bottom right). These DEGs represent only a minority (12%) of the global gene-expression profile that is congruent with a high correlation (Fig. 3A) observed between paired tumor and organoid samples. These results indicate that the global gene expression of primary tumor lesions is well conserved in organoid cultures.

To evaluate whether surviving cancer cells after neoadjuvant FOLFIRINOX developed resistance to the treatment, we first tested the drug response of five organoid lines (O1–O5) derived from treatment-naïve patients, and five organoid lines (O6–O10) derived from neoadjuvant FOLFIRINOX-treated patients to the individual components of FOLFIRINOX. We quantified cellular growth inhibition metrics of each condition (GR) 72 hours after drug exposure (Fig. 4 and A–J).

Drug response to oxaliplatin indicates four of five FOLFIRINOX-treated organoid lines were markedly resistant when compared with the treatment-naïve organoids (Fig. 4A). One of five FOLFIRINOX-treated organoids (O9) showed higher sensitivity to oxaliplatin compared with the rest of FOLFIRINOX-treated samples and clustered with treatment-naïve lines. Computerized tomography evaluation of the primary tumor of this patient (T9) indicated that this was the only patient with a partial response (PR) to neoadjuvant FOLFIRINOX (Table 1). The drug-sensitivity results are congruent with the clinical data suggesting the biology of this organoid line is different from the other neoadjuvant patient samples, particularly in the development of chemoresistance to oxaliplatin. We found a significant difference between the GR₅₀ values of FOLFIRINOX-treated versus treatment-naïve organoid lines (difference between means ± SE = 15.7 μmol/L ± 5.6) to oxaliplatin monotherapy (P = 0.03; Fig. 4D).

Drug response to 5-FU varies between different organoid lines (Fig. 4B), with no significant difference between FOLFIRINOX-treated and treatment-naïve groups (difference between means ± SE = 8.2 μmol/L ± 6.2; P = 0.21; Fig. 4E).

Further, we assessed the sensitivity of the organoids to irinotecan therapy (active metabolite SN38; Fig. 4C and F). Treatment-naïve organoids demonstrate heterogeneous responses to SN38, as three patients (O1, O2, and O4) were sensitive to the drug, while two patients (O3 and O5) were less sensitive to SN38. However, all neoadjuvant FOLFIRINOX-treated organoids were clustering together, showing lower sensitivity to SN38. Our data illustrated a significant difference in sensitivity to SN38 between the neoadjuvant and treatment-naïve (difference between means ± SE = 2.1 nmol/L ± 0.8; P = 0.04).

Next, we exposed organoid lines to different concentrations of the FOLFIRINOX combination regimen (obtained by mixing appropriate concentrations of oxalipatin, 5-FU, and SN38) and calculated GR values. All the five FOLFIRINOX-treated organoids displayed a significantly lower sensitivity to the combination treatment when compared with treatment-naïve lines (Fig. 4G and I; difference between means ± SE = 8.4 μmol/L ± 3.00; P = 0.02). Organoid line O3 displayed less sensitivity to FOLFIRINOX, reflecting the clinical resistance of the patient to adjuvant FOLFIRINOX (Table 1).

To test whether the chemoresistance is specific to FOLFIRINOX, we exposed all the organoid lines to gemcitabine as a negative control. As it is demonstrated in Fig. 4H and J, both treatment-naïve and FOLFIRINOX-treated lines respond equally to gemcitabine (P = 0.90).

Next, the influence of organoid passage numbers on drug sensitivity was tested by comparing the GR₅₀ values obtained from passages three and eight for two organoid lines (O1 and O8) exposed to oxaliplatin and SN-38 monotherapies. We did not find any significant difference between the GR₅₀ values obtained from different passages of the same organoids, regardless of the drug used or whether they were treatment-naïve or FOLFIRINOX-treated (Supplementary Fig. S1).

We performed DEG analysis in organoid cultures and corresponding tumor tissues. We observed 304 DEGs between the treatment-naïve and neoadjuvant FOLFIRINOX groups in primary tumor samples, whereas, in organoid lines, we identified 1,388 DEGs between these two classified groups. Among these, 58 DEGs were common between primary tumors and organoids (Fig. 5A). Heatmap visualization displayed the fold-change patterns of these 58 overlapping DEGs in both primary tissues and organoid lines (Fig. 5B). Among these genes, 32 were upregulated and 26 were downregulated. Supplementary Fig. S2 illustrates the expression pattern of all individual samples. These overlapping genes provide a set of candidate genes that are independently supported by primary tissues and corresponding organoid cultures.

Pathway analysis of upregulated genes revealed enrichment for pathways mainly involved in genomic instability (genes involved: TOP2A, EME1, and KIF18A), alternative energy metabolism (genes involved: SLC2A1, PLA2G4D, MBOAT2, and PKM), and disorders of transmembrane transporters (genes involved: NET1 and SLC2A1; Fig. 5C). Conversely, downregulated genes displayed enrichment for pathways mainly engaged in complement system (genes involved: C3and C4BPB), immune system (genes involved: SERPINB1, BLNK, OLFM4, CXCL8, CXCL3, and IL1R2), endoplasmic reticulum stress (genes involved: CXCL8, CXCL3, and CXCL6), and senescence (gene involved: CXCL8; Fig. 5C).

In this study, we have generated a unique library of organoids from surviving cancer cells obtained from resectable or borderline resectable PDAC tumors from patients who received eight cycles of neoadjuvant FOLFIRINOX treatment as a part of the Dutch pancreatic cancer clinical trial. We also generated organoids from treatment-naïve PDAC lesions that have not been under any selection pressures of chemotherapy. Comprehensive characterization of the organoids confirmed that they recapitulate their matched primary tumor tissues genetically, histologically, and transcriptionally. To our knowledge, no study investigated treatment responses of residual tumor cells after completion of neoadjuvant FOLFIRINOX treatment. FOLFIRINOX-treated organoids showed a resistance profile to oxaliplatin and irinotecan (SN38) monotherapy as well as to FOLFIRINOX combination therapy compared with treatment-naïve organoid lines, suggesting a clonal sweep of resistant clones under the selection pressure of these drugs (7). However, we did not observe any particular resistance profile to 5-FU monotherapy. Furthermore, organoids in both cohorts displayed equal sensitivity to gemcitabine, suggesting that the chemoresistance observed toward FOLFIRINOX is specific to this therapy. These data suggest that the chemotherapy regimen for PDAC patients needs to be determined at the individual level, because adjuvant therapy with oxaliplatin and irinotecan may not be beneficial for all patients after neoadjuvant FOLFIRINOX therapy.

The organoid line O9 derived from tumor material of a patient with PR to neoadjuvant FOLFIRINOX was the only exception. This particular organoid line (O9) was sensitive to oxaliplatin in vitro but resistant to irinotecan and FOLFIRINOX combination therapy. The oxaliplatin sensitivity of this organoid line recapitulates the better overall clinical response of patient to FOLFIRINOX. Congruently, Tiriac and colleagues (13) generated treatment-naïve organoids and showed that the patient's clinical response to FOLFIRINOX was strongly correlated to the oxaliplatin monotherapy response in vitro.

The clinical drug response of the three of five treatment-naïve patients who received adjuvant FOLFIRINOX therapy mirrored the in vitro drug-sensitivity response of their corresponding organoid models for FOLFIRINOX regimen, indicating that our organoid model may predict clinical drug sensitivity of the patient. However, this result must be verified in a bigger cohort of patient materials. PDAC patients (belonging to the neoadjuvant FOLFIRINOX cohort) were initially all responsive; therefore, they could complete eight cycles of chemotherapy and underwent tumor resection. However, the corresponding neoadjuvant FOLFIRINOX-treated organoids generated from residual tumor cells that escaped the treatment do not represent the initial clinical response of the patient.

Tumor-extrinsic factors (such as cancer-associated fibroblasts and immune cells) that are absent in the organoid cultures may promote therapy resistance in patients (13). Therefore, the impact of tumor microenvironment in the development of chemoresistance of patients should be investigated further in a more complex model.

We evaluated the mutational pattern of organoids by performing targeted NGS on a panel of cancer-associated genes (Supplementary Table S1), including the five known somatic driver gene mutations for pancreatic patients: KRAS, P53, CDKN2A, SMAD4, and ARID1A (9). The prevalence of driver mutations for PDAC in the primary tumors is conserved in our organoid cultures with no significant differences in the ratio of driver gene mutations in treated versus treatment-naïve organoids. Homogeneity in the mutational landscape of FOLFIRINOX-treated and treatment-naïve organoids suggests that resistant clones might originate from heterogeneity in nongenetic factors, such as different levels of receptor activity, differentiation state, and metabolism (8). Consistently, the diversity in the transcriptome profiles of these organoids and the partial segregation of samples based on the FOLFIRINOX treatment regime—both in primary tumor tissues and organoid samples—suggest gene expression could be related to the heterogeneity in treatment response.

We identified 58 core DEGs (treatment-naïve vs. FOLFIRINOX-treated) that overlapped between primary tumor lesions and their corresponding organoid cultures. Some of the pathways associated with alteration in these candidate genes are involved in crucial chemoresistance mechanisms and tumorigenesis of cancer cells, as reported by several studies. For instance, overexpression of PKM and SLC2A genes in the FOLFIRINOX patient samples is associated with altered metabolism leading to the survival of the cancer cells under stress conditions, which eventually confers drug resistance (28, 29). Furthermore, overexpression of the KIF18A gene is strongly associated with the proliferation and survival of cancer cells (30). Conversely, pathways engaged in cellular senescence (CXCL8) and innate immunity (C3, CXCL8) were downregulated in the FOLFIRINOX-treated group. Targeted therapies inducing senescence in PDAC cells lead to increased vascular remodeling in the tumor microenvironment of PDAC, rendering cells susceptible to gemcitabine-based chemotherapy and immunotherapy (12). Finally, downregulation in innate immunity is considered as a prognostic factor mediating the survival and metastasis of cancer cells (31).

In conclusion, our neoadjuvant FOLFIRINOX organoids provide an attractive model to predict responsiveness to therapeutic regimens in vivo and investigate therapy failure in pancreatic cancer patients. Our findings warrant more in-depth investigations to identify crucial elements of therapy resistance of each patient. These investigations might ultimately lead us to design and optimize a personalized strategy for combined chemotherapy regimens following neoadjuvant FOLFIRINOX therapy.

No disclosures were reported.

E.A. Farshadi: Conceptualization, data curation, software, formal analysis, supervision, validation, investigation, visualization, methodology, writing–original draft, project administration, writing–review and editing; the project, designed and performed all experiments, analyzed data sets, and wrote the manuscript. J. Chang: Software and formal analysis; analyzed RNA-seq data and provided related figures. B. Sampadi: Software, formal analysis, visualization, writing–review and editing; contributed with analyzing the RNA-seq data, providing related figures and provided inputs for writing the manuscript. M. Doukas: Writing–review and editing; contributed with all the pathologic assessments and supported collection of patients materials. F. Van 't Land: Data curation; provided patients' clinical information and followed up patients. F. van der Sijde: Data curation, writing–review and editing; contributed with technical assistance during the study. E.E. Vietsch: Data curation, writing–review and editing; contributed with technical assistance during the study. J. Pothof: Supervised Jiang Chang. B.G. Koerkamp: Resources, writing–review and editing; supported collection of patients materials and reviewed the manuscript. C.H.J. van Eijck: Conceptualization, resources, supervision, funding acquisition, writing–review and editing; contributed with supervision, study implementation, writing and reviewing the manuscript, and collection of patient materials.

We would like to thank the members of Prof. Riccardo Fodde laboratory for technical assistance. This study was funded by Stichting Overleven met Alvleesklierkanker.

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.