Bidirectional crosstalk between cancer cells and cancer-associated fibroblasts in mixed organoid system elicits transcriptomic characteristics of pancreatic cancer with potential therapeutic vulnerabilities

Dear Editor

Recent advancements in organoid technology enable tailored medicine for clinically devastating pancreatic cancer.^1–3 The correlation between the responsiveness of patient-derived pancreatic cancer organoids (PCOs) to chemotherapy suggests their potential as predictive instruments for therapeutic responses.^4,5 However, a PCO, solely composed of pancreatic cancer cells and lacking the interplay within the tumour microenvironment (TME), comes with inherent limitations: first, the understanding of non-genomic characteristics acquired through cellular interplay remains unclear, and secondly, it may exhibit abundant organoid-specific genes, resulting in a skewed transcriptomic and phenotypic disposition disparate from the patient profile.⁶ Cancer-associated fibroblasts (CAFs) play a pivotal role in pancreatic cancer. A composite organoid model encompassing both cancer cells and CAFs has unveiled a propensity for chemoresistance to conventional treatments,^7,8 which is absent within cancer-only organoids.⁹ However, the intricate interplay between cancer cells and CAFs within this mixed PCO-CAFs remains to be explored.

This study introduces a comprehensive transcriptomic analysis of the mixed PCO-CAFs, comparing them with both PCOs and patient tumour tissues. We also identify potential therapeutic targets through this analysis, to provide new avenues for therapeutic intervention.

We generated five corresponding sets of patient tumor tissue, patient-derived PCO, and mixed PCO-CAF as previously outlined (Figure 1A, Table S1).¹⁰ The mixed PCO-CAF consisted of pancreatic cancer cells and CAFs (Figure S1A), closely resembling histologic features observed in the corresponding patient tissue specimens (Figure 1B). Through bulk RNA sequencing analysis, we substantiated that the mixed PCO-CAF exhibits a greater resemblance to the transcriptomic characteristics inherent to those patient tissues compared to PCOs when analyzed with the top 500 variable genes identified in each sample (Figures 1C, S2 and Table S2).

[IMAGE OMITTED. SEE PDF]

Among the four clustered gene sets based on their expression levels (Figure 1D), Geneset 1 (G1) comprised 187 genes that displayed up-regulated immune response-related pathway gene ontology exclusively within the patient tissue (Figure S3A). Conversely, Geneset 3 (G3), which comprised 260 genes, exhibited comparable up-regulation both in the patient tissue and the mixed PCO-CAF compared to the PCO. Within G3, the elevated expression of diverse growth factors and cytokines encompassed within the Receptor-ligand activity and the extracellular matrix (ECM) structure gene ontology, highlighting their integral involvement with cellular communication and matrix-related activities, were consistently discerned within both the patient tissue and the mixed PCO-CAFs (Figure 1E and S3B). These findings suggest that the mixed PCO-CAF not only facilitates the emulation of a stroma-rich TME through ECM production but also orchestrates intercellular signalling communication by virtue of the expression of ligands and corresponding receptors. Furthermore, analysis of RNA sequencing data sourced from The Cancer Genome Atlas (TCGA) dataset for pancreatic adenocarcinoma corroborates the significant enrichment of G3 (Figure 1F).

A single cell-level transcriptome analysis was conducted on the PCO, mixed PCO-CAF, and the corresponding patient tissue. The cellular composition within the mixed PCO-CAF encompassed cancer cells and CAFs of distinct subtypes, comprising myofibroblast-like CAFs (myCAFs; CAF-1 cluster) and inflammatory CAFs (iCAFs; CAF-2 and CAF-3 clusters) (Figure 2A,B). To validate the similarity between cancer cells from tissue and those from PCO-CAF, we integrated cancer cell subsets from the matched patient sample. The Uniform Manifold Approximation and Projection and trajectory analysis unveiled discernible clustering of cancer cells within the mixed PCO-CAF, distinguishing them from those within the PCO, and that the cancer cells from the mixed PCO-CAF exhibited a transition towards states closely resembling those from the patient's tissue (Figure 2C,D). While the cancer cells present in both the mixed PCO-CAF and the patient tissue displayed an enrichment of genes associated with growth factor activity and ECM interactions, the cancer cells within the PCO displayed heightened expression of genes linked to cell division mechanisms (Figure 2E,F and Figure S4A–D). Significant enrichment of gene signatures such as TNFα signalling via NF-κB and epithelial-mesenchymal transition (EMT) was discerned within the cancer cells of the mixed PCO-CAF (Figure 2G,H). Especially, cancer cells in the mixed PCO-CAF showed increased expression of EMT-related genes, such as MMP1 and vimentin, while the expression of epithelial markers was decreased (Figure 2I). Further comparison between the cancer cells in the mixed PCO-CAF and the PCO mono-culture revealed that these alterations are partly derived from the secreted factors in the culture medium (Figure 2J,K and Figure S4E). These findings collectively propose a dynamic reshaping of the transcriptomic landscape within cancer cells in the mixed PCO-CAF system, attributable to the bidirectional communication established between cancer cells and CAFs.

[IMAGE OMITTED. SEE PDF]

We next focused on identifying cellular interactions within the mixed PCO-CAF. Notably, the pancreatic cancer cells not only displayed cancer cell-to-cancer cell interactions but also a heightened up-regulation of interaction weights, primarily evident in their interactions with iCAFs in comparison to myCAFs. These interactions were mediated through a multitude of ligand-receptor pathways, including MIF-CD74, GRN-SORT1, GDF15-TGFR2, HGF-MET, INHBA-ACVR, TNC-SDC and BDNF-NTRK2, by enhanced expression of ligands within CAFs and the corresponding receptors within the cancer cells (Figure 3A). Several of these ligand-receptor interactions were identified within the five patient samples and mixed PCO-CAFs (Figure 3B). When evaluating the therapeutic potential of targeting the aforementioned up-regulated pathways, cancer cells within the mixed PCO-CAF exhibited an elevated vulnerability to targeted therapeutics, surpassing the susceptibility observed in the PCO (Figure 3C). The mixed PCO-CAFs from patient #52 revealed an enhanced response to Trametinib (NRG inhibitor) and Sapitinib (ERBB3 inhibitor) compared to the PCO mono-culture (Figure 3D,E), as shown by the highlighted NRG-ERBB3 interactions in the sample (Figure 3A,B). Additionally, the mixed PCO-CAFs from patients #79 and #70 showed increased susceptibility to Trametinib and/or ANA-12 (NT inhibitor) (Figure 3F,G and Figure S5), which could be inferred from the sequencing data (Figure 3B). It is noteworthy that the cancer cells within the mixed PCO-CAF also exhibited heightened resistance to the gemcitabine and paclitaxel, unlike those within the PCO. This resistance could be attributed to the intricate interplay between cancer cells and CAFs, as previously underscored.^7,8

[IMAGE OMITTED. SEE PDF]

In conclusion, our study demonstrated that the mixed PCO-CAF proficiently recapitulates patient tissue transcriptomic and receptor-ligand engagements when contrasted with the PCO, offering insights into the intricate intercellular communication between cancer cells and CAFs and personalized therapeutic application (Figure 3I).

AUTHOR CONTRIBUTIONS

Conceptualization: J-IC, MJY, DL and SK. Methodology: J-IC, MJY, K-JW, SHPand SK. Investigation: J-IC, KJW, SHP and SK. Formal analysis: J-IC, MJY, DL and SK. Funding acquisition: MJY, DL and SK. Writing—original draft: J-IC, DL and SK. Writing—review and editing: J-IC, MJY, DL and SK. Approval of the final manuscript: all authors.

ACKNOWLEDGEMENTS

We thank Dr Hyuncheol Jung (the University of California San Francisco), Dr Sungsoo Kim (Columbia University Medical Center), Dr Heeyoung Lee (Korea Advanced Institute of Science and Technology) and Dr Soo Ick Cho (Lunit Inc.) for helpful discussions, and Jaebin Yi for language editing. This work was supported by the National Research Foundation of Korea (NRF) grant funded by the Ministry of Science and ICT (MSIT) (2021R1C1C1008619, 2022R1C1C1007289 and 2021R1A2C2005853), Republic of Korea.

CONFLICT OF INTEREST STATEMENT

The authors declare that no conflict of interest.

DATA AVAILABILITY STATEMENT

RNA-sequencing data files are deposited in the GEO database (GSE253561). All other data generated in this study are available from the corresponding author upon reasonable request.

ETHICS STATEMENT

This study was approved by the institutional review board of Ajou University Hospital (AJIRB-BMR-20-222). Written informed consent was obtained from all patients. All procedures were in accordance with the Declaration of Helsinki.

References

Boj SF, Hwang CI, Baker LA, et al. Organoid models of human and mouse ductal pancreatic cancer. Cell. 2015; 160 (1-2): 324-338. doi: [DOI: https://dx.doi.org/10.1016/j.cell.2014.12.021]