Colorectal Adenocarcinoma Organoids Enable Morphologic-Based Tumor Evaluation

Introduction: Organoids are stem cell-originated self-organized three-dimensional (3D) cell clusters. Typical two-dimensional cancer cell line models for cancer therapy do not reflect organ structures and tumor heterogeneity. The use of organoids could represent the growth environment of cells in vivo, especially the tumor microenvironment (TME), which could accurately maintain the molecular, cellular, and functional characteristics of original organs or tissues. Moreover, they are expected to reflect their parental tumors' genetic and phenotypic diversity and offer the capability to mimic patient reactions to medical treatments, thereby advancing cancer disease models and drug screening. To produce the tumor organoid model, the droplet-based organoid generation system offers the benefits of high throughput with uniform size control. Additionally, cooperating droplets with hydrogels can produce high-fidelity and reproducible tumor organoids, which involve desired tumor-tumor and tumor-ECM interactions with sufficient physiological relevance in a controllable manner.
Herein, we developed a microfluidic-engineered Colorectal cancer (CRC) organoid droplet setup to encapsulate organoids in microbeads, recapitulate parental tumor phenotypes, and follow with downstream therapeutic screening. Colorectal cancer (CRC) organoids typically involve specific epithelium features, which drive different morphologies such as solid and lumen. We found that the drug response presented different differences between CRC organoids of different morphologies. To identify the morphological difference between the organoids, the progression in computer vision holds the benefit of label-free detection, and it provides a minor invasion that alters the intrinsic organoid dynamic changes. Thus, with an advanced approach, the organoids are further classified and separated based on morphology, assisted by machine learning.

Materials and

Methods: The human colorectal adenocarcinoma organoid line (PDM-359, CCO) was used to demonstrate organoid-laden microbead generation. The organoids were embedded in Matrigel dome and supplemented with Antibiotic & Antimycotic, Noggin, N-acetyl cysteine, Nicotinamide, EGF, Gastrin, HEPES, GlutaMAX, B27, A83-01, SB202190 and Y27632 in DMEM/F12 ++ medium.
FluoroSurfactant was added to the HFE7500 engineered fluid as the continuous phase for microfluidic droplet generation. The harvested cells were resuspended at 15x106 cells/mL in 5 mg/mL of Collagen type I solution as the dispersed phase. Two syringe pumps loaded the continuous and dispersed phases into the microfluidic droplet generation device. The generated cell-laden microbeads were transferred from the microfluidic droplet device and collected in a 24-well plate containing the culture medium.
Before the drug treatment, the CCO microbeads were grown for one week. Subsequently, the microbeads were evenly distributed into a 96-well plate for drug testing. The anticancer drugs were introduced into the organoid culture medium at varying final concentrations, from 100 nM to 5 μM. Calcein AM and propidium iodide (PI) distinguished live and dead cells.
To categorize the CCO microbeads based on their morphologies, they were harvested on day eight and resuspended in 5% GelMA, and LAP served as the photoinitiator. The suspension of microbeads was then transferred into a 6-well plate. A detection model, trained using Yolov5, identified the non-target microbeads and facilitated their photocuring through negative sorting. The non-cured target microbeads were flushed from the wells and gathered for further analysis.

Results, Conclusions, and Discussions: The cell-laden collagen microbeads were prepared with a uniform size of 350 μm. Unlike the collagen microbeads gel dome, the gel dome does not confine space, allowing organoids to grow without spatial limitations. As a result, the size of the organoids within the gel dome can vary from 50 to 300 μm after five days of cultivation. After this period of in vitro culture, the viability of the colorectal cancer organoids was assessed; both the collagen microbeads and the gel dome demonstrated over 98% cell viability.
In our organoid microbeads, we could observe solid and luminal organoids. Thus, with the features of size-uniform, morphological-uniform, cell-matrix uniform, and high reproducibility, the microbeads could be considered a well-controlled tumor organoid model for accurate drug screening. We have chosen six drugs as candidates: Doxorubicin (DOX), Paclitaxel, Fluorouracil (5FU), Brigatinib (Brig), Izorlisib (CH5), and Pictilisib.
To streamline the categorization of organoids, we focused on two morphological types: solid organoids and luminal organoids. In the solid CCO group, DOX and 5-FU exhibited high cytotoxicity at 1 μM concentrations, with the CCOs demonstrating greater sensitivity to 5-FU and Paclitaxel at lower doses. Pictilisib notably induced a significant mortality rate in CCOs, with 500 nM concentrations reducing viability to below 50%. Conversely, in the luminal CCO group, Paclitaxel was less toxic, preserving the morphology of the organoids even at higher concentrations. Additionally, luminal organoids displayed considerable resistance to 5-FU. Brigatinib, however, showed a marked increase in cytotoxicity once the drug concentration reached 1 μM.
Due to varying drug responses, we categorized the CCOs based on their morphology to facilitate downstream analysis and future research. This process typically necessitates a purified sample set and a larger population. To accomplish this, we employed transfer learning to develop a machine-learning detection model using Yolov5, which identifies the organoid microbeads based on their phenotypes/morphologies. This model achieved an accuracy of 94.5%. Moving forward, we plan to utilize our UV projector-based cell capture platform to perform negative sorting of the organoid microbeads across their classified groups.

Acknowledgements (Optional):