Organoid technology for personalized pancreatic cancer therapy

Pancreatic ductal adenocarcinoma is a highly aggressive tumor type with a 5-year survival rate below 5% [1]. It currently ranks as the third leading cause of cancer-mortality, but is expected to become the second leading cause in the U.S.A. by 2030 [2, 3]. The low and non-durable response rate to current treatments is attributed to the inherent aggressive nature of pancreatic cancer cells, the dense desmoplastic reaction and the molecular heterogeneity observed in this tumor type. In addition, most pancreatic cancers appear to be non-immunogenic, explaining the refractoriness to immunotherapeutic compounds [4, 5]. Phase III drug trials are plagued by high dropout rates and low initial clinical correlations, explaining the limited success rate [6]. If interpatient biological variability, as evident by the characterization of the molecular subtypes [7] and long-term survivors of the disease [8], is to be addressed as well, development of viable models for personalized medicine is next in line. Personalized medicine, or precision medicine, is an approach for improving the outcome of cancer therapies by grouping patients based on their predicted individual responses to therapy. Limitations of classical preclinical models, such as 2D monolayer cell cultures, genetically engineered mouse models, and patient-derived tumor xenografts, include a lack of comparability with normal tissue [9, 10], culturing times exceeding months [11], foreign stromal components [12, 13] and selection of more aggressive phenotypes [14, 15].

Conversely, patient-derived organoids (PDOs) can be generated from extracted patient tumor tissue (primary tumor and metastases) within two to four weeks following surgery or biopsy via endoscopic ultrasound [16, 17] (Fig. 1). PDOs recapitulate histopathologic and genomic profiles of the tissue of origin while maintaining genomic stability throughout passaging [18‐20]. Organoids represent cell cultures kept in an organ-type specific matrix for 3-dimensional growth by avoiding cell attachment to the surrounding plate. 3D cultures derived from established monolayer cell lines are called spheroids, whereas patient-derived organoids are defined as multicellular extracellular matrix (ECM)-dependent units derived directly from patient tissue. The extracted tissue is first minced, then enzymatically digested and placed in a matrix (collagen or Matrigel) together with a culture medium containing growth factors and modulators enriching epithelial cells. The cells form microscopically visible organ-like structures within two days, with passaging and propagation ready once 80% confluency is reached [13, 17]. Organoids have been used to simulate a variety of diseases. Intestinal organoids, for instance, have been utilized for the study of cystic fibrosis, inflammatory bowel disease and Clostridium difficile infection. Cancer has been studied utilizing organoids in the context of neoplastic transformation and therapeutic screening, and pancreatic cancer is at the forefront of organoid technology applications. However, there are major challenges pertaining to the implementation of organoid technology in personalized pancreatic cancer care. These challenges can be divided into different phases of the organoid workflow process, as depicted in Fig. 2. The establishment phase entails extraction of tumor tissue and initial culturing. Here, ensuring a good neoplastic cellularity is key for a successful initial outgrowth, a challenge highlighted by Tiriac and colleagues in a recent multicenter study [17]. While resection has remained the main extraction method, introducing novel biopsy techniques may broaden potential therapeutic applications. The next phase, termed ‘propagation’, describes the conditions in which the neoplastic cell clusters can multiply for further passaging. There are various culture methods currently in use to minimize deviation from the in vivo material. The third and fourth phases encompass different approaches to therapeutic assays of promising novel compounds. Drug screening in the in vitro setting has traditionally consisted of a short phase of cell line screening and the effect measures of half maximal inhibitory concentration (IC₅₀). With the advent of experimental therapeutic organoid assays, more complex yet faster assays are under development. These powerful viability assays, together with drug response models and drug-specific classifiers could provide the basis for rapid therapeutic decision-making.

This review will focus on the applications of PDOs for high-throughput drug screening and personalized therapy in pancreatic cancer, utilizing the previously outlined development steps of establishment, propagation, drug screening and response prediction. The review will conclude with a proposed organoid workflow and a remark on the potential for organoids to be introduced in the pipeline of promising chemotherapeutics.

3D cell culture started with the hanging drop tissue method developed by Ross Harrison in 1906 [21]. In 1977, the study of the extracellular matrix of chondrosarcoma led to the development of an ECM substrate with basal membrane characteristics, today known as Matrigel [22]. This enabled the characterization of ECM proteins and their role in tissue morphogenesis. The parallel progress of stem cell biology led in 2003 to the discovery that progenitor cells generated complex structures resembling the tissue of origin when embedded in Matrigel [23]. The Clevers laboratory pioneered culturing methods for epithelial organoids with cells from a variety of organs, even without stem cell markers, starting in the early 2000s. In 2013, the laboratory grew budding cyst-like structures from pancreatic duct cells using the Wnt-Lgr5-Rspo signaling axis [24]. In 2015 the first patient-derived pancreatic tumor organoid model was published as a result of a collaboration between the Clevers and Tuveson laboratories. This protocol includes Wnt-activating serum-free media over a dome of Matrigel [13].

Other methods for establishing pancreatic cancer organoids have followed. The Muthuswamy lab cultured cells in an overlay over a Matrigel bed without Wnt-signaling medium [25], while the Skala group developed a PDO culture model with visible innate fibroblasts by initially growing them in mice [26]. In 2019, the first comprehensive protocol for establishing pancreatic cancer organoids from resected tissue was published by Baker et al. [16] (Clevers and Tuveson groups). This protocol enables long-term propagation of PDOs and has, in a slightly modified form, been employed for high-throughput screening (HTS) of drugs [20, 27]. A limitation of this model is the absence of cancer-associated fibroblasts (CAFs) after a few days in culture media. Another research group maintained normal pancreatic tissue organoids in collagen matrices and Transwell cell culture plates with an air-liquid interface (ALI) [28]. At this stage, the ALI method is mainly used to produce kidney and intestinal organoids [29], whereas the use of collagen-based hydrogel may provide a solution for automated processes like bioprinting [30].

To construct a robust personalized therapy workflow for use in the clinic, novel biopsy technologies have to be combined with optimized extraction and establishment protocols. This will increase tumor cell viability, minimize normal cell contamination and increase establishment rates. The first patient-derived tumor pancreatic model by the Clevers and Tuveson laboratories showed establishment rates of 75% and 83%, respectively. With the use of fine-needle aspiration, Tuveson and colleagues achieved an 87% success rate, however only 66% of organoids survived the fifth passage [17]. According to the authors, a success rate goal of at least 90% should be set for clinical implementation [17]. While the lack of a tailored medium composition is thought to contribute to the failure of growth in some organoids, the enrichment of distinctive subclones is also a cause for concern. The selection of specific protein-dependent pancreatic cancer subclones without identifiable genetic mutations has been described [27, 31], showing an unmet need to define ideal media compositions for specific tumor types.

Obviously, success rates vary between extraction methods. In addition to novel biopsy techniques, other forms of tumor cell extraction have gained interest in recent years. Circulating tumor cells (CTCs) are primary tumor and/or metastatic cells shed into the blood stream to seed metastatic lesions. Among many differing technical solutions for the separation of seeding-capable tumor cells from peripheral blood, Zhang et al. demonstrated that, with a microfluidic CTC-chip, brain-metastatic breast cancer cells were able to undergo 3D-coculturing with fibroblasts and extracellular matrix components on the same chip and to maintain strong mutational concordance [32].

Useful companion biomarkers, shortened assay times, increased biomimicry, and optimized viability assays and protocols are all important aspects of the pancreatic PDO model in need of further optimization. Still, knowledge of the interplay between the TME, pancreatic cancer subtypes and drug response is limited [52], which means that clinical implementation needs to be validated on a per-drug basis. Each compound will affect and be affected by the organoid model through several distinct and unelucidated mechanisms. However, with the technical solutions presented in this review as a basis, a cautious description of a future optimized clinical workflow for pancreatic cancer organoids could be as follows: (1) A less invasive novel biopsy technique is used to harvest tumor tissue with good cellularity. Transcriptomic characterization of the tumor tissue reveals a subtype requiring a specific documented culture medium composition. Establishment time is less than one week, and no fibroblast overgrowth or wild-type contamination is observed. (2) Cells are dissociated and resuspended in a bioprinter-specific hydrogel formulation and seeded with precision using laser-assisted bioprinting. The seeding is carried out on a microfluidic chip in coculture with fibroblasts and vascular endothelial cells. (3) Optical metabolic imaging is employed to investigate intra-tumoral heterogeneity prior to the introduction of chemotherapeutic compounds to each chip platform. The viability assay time is less than 24 hours. With the help of a classifier tool, only one concentration per chemotherapeutic compound is required. (4) The classifier tool is based on the observed intra-tumoral variability, the viability assay and companion biomarkers specifically developed for pancreatic cancer organoids, creating a robust response prediction. A chemotherapeutic regimen is introduced or adjusted accordingly.

The use of PDOs may alleviate the struggles of drug development by expanding the in vitro testing capabilities. Future promising therapies, like photodynamic therapy and chemotherapeutics combined with CDK 4/6-inhibitors [65, 66] could benefit from organoid models. Studies of clinical correlation are ongoing, and a phase II trial of pancreatic PDOs is currently being designed [67]. Because of the limited therapeutic advances in pancreatic cancer that have so far been made, hope is being placed on translational models such as PDOs to facilitate the transition of research data to clinical practice. Here, we have presented an overview of PDO models, focusing on their application and optimization, with the overarching goal to facilitate personalized pancreatic cancer treatment.