Integrated Microfluidic Platform for High‐Throughput Generation of Intestinal Organoids in Hydrogel Droplets
Barbora Lavickova, Hannah Kronabitter, Mar Cervera‐Negueruela, Eylul Ceylan, Laura Benito‐Zarza, Rubén López‐Sandoval, Irineja Cubela, J. Gray Camp, Jose L. Garcia‐Cordero
TL;DR
A new microfluidic platform creates uniform intestinal organoids for scalable and reproducible research in drug screening and disease modeling.
Contribution
An integrated microfluidic system for high-throughput, reproducible generation of intestinal organoids in hydrogel droplets.
Findings
The platform enables uniform encapsulation of intestinal stem cells in hydrogel microparticles.
Healthy and tumor-derived organoids were successfully cultured with high homogeneity and reproducibility.
The system supports efficient transfer of microparticles to aqueous media for downstream applications.
Abstract
Organoid research offers valuable insights into human biology and disease, but reproducibility and scalability remain significant challenges, particularly for epithelial organoids. Here, we present an integrated microfluidic platform that addresses these limitations by enabling high‐throughput generation of uniform hydrogel microparticles embedded with primary‐derived human adult intestinal stem cells. Our platform includes a cell distribution system for homogeneous cell encapsulation and a microfluidic oil‐removal module for efficient particle transfer to aqueous media. We demonstrate the successful culture and differentiation of both healthy‐ and tumor‐derived intestinal organoids within these microparticles, achieving high homogeneity and reproducibility. This integrated microfluidic approach holds promise for scalable and standardized organoid production, with potential applications…
Genes, proteins, chemicals, diseases, species, mutations and cell lines named across the full text — each resolved to its canonical identifier and authoritative record.
Click any figure to enlarge with its caption.
FIGURE 1
FIGURE 2
FIGURE 3
FIGURE 4Peer Reviews
No public reviews on file for this paper yet. If you reviewed it on a platform where reviews are public (OpenReview, ICLR, NeurIPS, ICML), you can paste yours below so the community can read it here.
Videos
No videos yet. Explain this paper in a talk, walkthrough, or lecture? Add one.
Taxonomy
Topics3D Printing in Biomedical Research · Cancer Cells and Metastasis · Innovative Microfluidic and Catalytic Techniques Innovation
Introduction
1
Organoid research stands at the forefront of biomedical innovation, providing unprecedented insights into human development and disease. However, the lack of reproducibility and scalability in organoid production presents a significant obstacle. This challenge is particularly prevalent in epithelial models, such as intestinal organoids, which are typically embedded in surface‐attached extracellular matrices (ECM) hydrogel domes, like Cultrex Basement Membrane Extract (BME) or Matrigel [1, 2, 3]. Organoids embedded in these biological hydrogels are often difficult to manipulate [2, 4], and the 3D structure of the dome makes upstream image analysis of individual organoids challenging [5, 6, 7]. Furthermore, they exhibit heterogeneity in size, shape, and cell composition due to spatiotemporal gradients of growth factors and oxygen [2, 8, 9, 10, 11]. Droplet‐based microfluidics allows for the generation of thousands of small uniform particles, therefore offering a promising solution for high‐throughput generation of organoids with both capsules and hydrogel droplets/microparticles, demonstrating significant potential for standardization of organoids and their use in drug screenings [12, 13].
Capsules are suitable for suspension cultures, including brain [14] and pancreatic islets [15], whereas hydrogel‐based droplets/particles are often used for embedded epithelial organoids. These hydrogel microparticles are frequently formed using high‐stiffness synthetic hydrogels, notably gelatin methacryloyl (GelMA) [16, 17] and alginate [18, 19]. While these materials are user‐friendly, as they can be polymerized within seconds by UV or ion bath and are not temperature sensitive, they often inadequately support organoid growth. Alternatively, basement membrane‐like matrices such as Matrigel [20, 21, 22] are sometimes used. However, these matrices present challenges due to their temperature sensitivity and insufficient droplet stability, which can hinder their long‐term use and applicability in bioreactors. Moreover, many encapsulation methods struggle to ensure uniform cell distribution within droplets and capsules [21], due to cell clumping and aggregation. This can be overcome by a high cell density encapsulation [23, 24], the addition of a magnetic stirrer [25], or a distribution device [26].
Although oil‐free droplet generation devices are gaining popularity [15, 27], they are more complex [23, 28] and rely on specific material compositions [13, 15]. Therefore, oil‐based droplet microfluidics methods, such as flow focusing, remain the most popular and straightforward techniques to produce highly monodisperse droplets across a wide range of volumes [12]. Depending on the type of particle, various oil removal methods [29], including demulsification [30, 31], oil absorption by membranes [32, 33], and filtration [34], are employed. These methods are manual and may result in cell death, particle loss, and oil residues. To address these issues, on‐chip oil removal solutions have been introduced [35]. However, these devices were developed for fast polymerization UV‐based hydrogels [16], high‐viscosity oils, and high‐stiffness particles, rendering them unsuitable for the soft hydrogel materials necessary to support organoid growth.
Overall, despite the existence of various microfluidic strategies for particle generation, they generally fail to provide the stability and homogeneity required for high‐throughput floating bioreactor cultures of epithelial organoids. Specifically, these methods are hindered by the lack of appropriate 3D matrices, inconsistent cell loading, or a persistent reliance on labor‐intensive manual steps. To address these challenges, we introduce a microfluidic platform featuring an integrated cell distribution system and a particle generation device that enables the controlled production of thousands of highly uniform hydrogel microparticles embedded with primary‐derived human adult stem cells, serving as precursors for intestinal organoid growth. We demonstrate that optimizing hydrogel composition is essential for the culture of epithelial organoids, while maintaining the stability for bioreactor cultivation, manipulation, and distribution into well plates. Lastly, our microfluidic oil removal module facilitates the transition of microparticles from oil to aqueous media, minimizing manual handling and particle loss, thereby enhancing the overall efficiency and scalability of organoid production.
Results
2
Generation of Homogeneous Hydrogel Droplets
2.1
We hypothesized that achieving uniform mature organoids requires the initial generation of homogeneous hydrogel droplets with consistent volumes and even distribution of cells within each droplet. To accomplish this, we employed a flow‐focusing microfluidic device (Figure 1a,b; Figure S1a). This device facilitated the rapid production of thousands of highly uniform hydrogel droplets (Figure 1c) within minutes, each with a volume in the nanoliter range. The particle diameters can be controlled by varying the flow rates of the oil and the hydrogel precursor, as well as by modifying the dimensions of the microfluidic channels (data not shown). For our purposes, we chose a droplet diameter of approximately 350 µm, which is optimal for generating intestinal organoids of significant size while retaining ease of manipulation. A challenge associated with water‐in‐oil droplet formation is the introduction of an oil fraction and its subsequent removal following particle polymerization. To mitigate this, we employed PTFE membranes, which exhibit a high affinity for oil, to absorb the oil fraction prior to droplet culture (Figure S1b).
*Generation of organoid droplets: (a) Schematics of the microfluidic chips used for droplet generation. A single cell solution is flushed through a distribution chip to ensure cell homogeneity. Droplets are generated in a flow‐focusing device where oil disrupts the hydrogel flow. This allows for high‐throughput generation of miniaturized hydrogel culture systems, with increased organoid reproducibility. Schematic was partially created with BioRender.com. (b) Brightfield images of cell distribution device pillars of teardrop shape at the inlet and outlet. The cell encapsulation device with the flow‐focused generation of cell‐hydrogel droplets (scale bars 250 µm). (c) Brightfield images of droplets in oil and media, showing the cell density with and without cell distribution device (scale bars 250 µm). Droplets densely packed with cells are indicated by red arrows, nearly empty droplets are indicated with blue arrows. (d) Mean pixel intensities of droplets generated with and without the cell distribution device on day 0. Data were extracted from brightfield images. (e) Quantitative comparison of the coefficient of variation (CV) in cell‐laden droplets on day 0 across experiments. Each dot represents the coefficient of variation in a single experiment (n = 4 biological replicates). P‐value of < 0.05 was expressed as *, p‐value < 0.01 was expressed as ** and p‐value < 0.001 was expressed as ***. The normality of the distribution was verified with a Shapiro‐Wilk test, and a t‐test was done to calculate the p‐values. (f) Size analysis of colon organoids in a bioreactor on day 8 formed with or without distribution device, with coefficient of variation (CV) highlighted above (n organoids ≥ 300). (g) Quantitative fluorescence analysis of cell viability of cells after encapsulation on day 0 in domes or droplets by live/dead staining (n samples = 4). (h) Brightfield images of cell‐hydrogel droplets composed of 100% BME (Matrigel) and 1.5 mg/mL Collagen I ‐ 25% BME (Matrigel) ECM mix (scale bars 250 µm). (i) Quantitative fluorescence analysis of cell viability of intestinal organoids at day 7 of culture in BME (Matrigel) and 1.5 mg/mL Collagen I ‐ 25% BME (Matrigel) ECM mix domes by live/dead staining (n samples = 3). (j) Brightfield images of healthy colon organoids grown on day 6 (scale bars 250 µm) and mIF staining of healthy colon organoids grown in a bioreactor on day 8 (SOX9 turquoise, FABP1 yellow, MUC2 green, and Ki67 red, scale bars 100 µm) comparing the two hydrogels, BME (Matrigel) and 1.5 mg/mL Collagen I ‐ 25% BME (Matrigel) ECM mix. (k) Quantification of Log2 transformed expression levels and heatmap of z‐score normalized expression levels across selected genes for the two ECM conditions. Data were averaged across replicates. Statistical significance between conditions was assessed using the Wilcoxon test, with p‐values adjusted for multiple testing using the Benjamini‐Hochberg method. P‐value of < 0.05 was expressed as *, p‐value < 0.01 was expressed as ** and p‐value < 0.001 was expressed as **.
While microfluidic droplet formation produced highly homogeneous hydrogel droplets, the encapsulated cell density varied significantly despite the use of a single‐cell suspension. Some droplets exhibited dense cellular packing, whereas others remained nearly empty (Figure 1c). High‐speed camera observations revealed pulsatile flow caused by cell clumping and inhomogeneous cell dispersion within the hydrogel precursor. To address this issue, we implemented a microfluidic distribution device featuring a series of pillar arrays with varying diameters and pitches prior to encapsulation. This design, inspired by a previously reported device [26], aimed to break up cell clumps and homogenize cell distribution in the precursor mixture. However, we observed that the circular shape of the pillars led to significant device clogging (data not shown). To mitigate this issue, we optimized the pillar design by adjusting their shape, size, and spacing (Figure S1c). Teardrop‐shaped pillars proved to be the most effective, as they maximized cell yield and minimized clump formation. At the entrance of the distribution device, larger cell clumps were evident, but at the exit, these clumps were significantly reduced (Figure 1b). This optimization resulted in hydrogel droplets with a more uniform cell distribution and no visible high‐density clumps (Figure 1c).
To confirm the improvement in cell distribution, we measured the mean pixel intensity in droplets, which serves as an approximation of cell densities. On day 0, droplets produced from the same batch of cells using the distribution device exhibited reduced dispersion and greater consistency in mean pixel intensity compared to those generated without the device (Figure 1d). This was further supported by a reduced coefficient of variation in the mean intensities of droplets generated with the distribution device across different experiments (n = 4) (Figure 1e). By day 8, organoids in droplets generated with the distribution device exhibited a 10% lower coefficient of variation compared to those generated without it (Figure 1f), confirming increased reproducibility and reduced size variation. Additionally, the high cell viability (95%–100%) immediately after encapsulation, with no significant difference compared to cells directly seeded in dome culture (Figure 1g), indicates that the encapsulation process imposes low stress on the cells.
Basement membrane‐like matrix (BME), such as Matrigel, is the most commonly used matrix for intestinal organoid growth and has been previously used for hydrogel droplet formation [21, 22, 23, 24]. However, its soft properties result in unstable particles (Figure 1h), complicating oil extraction and leading to droplet merging and disintegration over time. To enhance droplet stability while maintaining support for epithelial organoid culture [36], we added type I collagen to the BME. We used live/dead staining of intestinal organoids grown in hydrogel domes on day 7 (Figure 1i; Figure S2a,b) to assess various mixtures. We observed the highest viability (85%), comparable to pure BME, in a mixture containing a final collagen concentration of 1.5 mg/mL and 25% (v/v) BME. Comparable viability was also observed with a 50% (v/v) BME concentration. However, higher collagen concentrations (2.5 mg/mL) resulted in lower cell viability (60%–70%). We therefore selected the formulation containing 25% (v/v) BME and a final collagen concentration of 1.5 mg/mL. This composition maintained high cell viability and also produced stable droplets with higher integrity (Figure 1h), which prevented merging, simplified oil removal, and allowed for easy manipulation and centrifugation. Additionally, the droplets remained stable over time, supporting both static and bioreactor cultures. Organoids grown in this extracellular matrix (ECM) mixture displayed a dense morphology with elongated columnar cells characteristic of differentiated intestinal organoids, with less budding morphology observed compared to those grown in pure BME (Figure 1j). Multiplex immunofluorescence (mIF) staining (Figure 1j) confirmed the presence of diverse cell types, including stem cell and proliferation markers (SOX9^+^, Ki67^+^), enterocytes (FABP1^+^), and goblet cells (MUC2^+^). Bulk RNA sequencing (Figure 1k; Figure S2c) of organoids grown in ECM mix or BME revealed similar expression patterns, with slight downregulation of stemness markers (LGR5, SOX9) and upregulation of FABP2, GUCA2A, indicating a slightly higher level of differentiation in the ECM mix domes.
Organoid Development and Homogeneity
2.2
To evaluate the biocompatibility of the encapsulation process and confirm if the organoids faithfully recapitulated those grown in traditional culture and their parental tissues, we cultivated droplets containing adult intestinal stem cells for 8 days, a common established timeline for intestinal organoids [2, 37]. While encapsulated organoids cultivated in static conditions in a 6‐well plate underwent coalescence over time, forming larger, more heterogeneous agglomerates, organoids cultured in the bioreactor retained their individuality and remained strikingly uniform through day 8 (Figure S3). Therefore, we cultured the droplets in the bioreactor throughout the study to prevent the organoid fusion.
Subsequently, we tracked organoid maturation and cellular organization using brightfield microscopy, Hematoxylin and Eosin (H&E) staining, and mIF (Figure 2a). By day 3, the single cells self‐assembled into organoid aggregates (r = 75 µm) that began to grow and differentiate. By day 6, these organoids grew in size (r = 118 µm) and exhibited a proliferative phenotype marked by SOX9 and Ki67 expression. By day 8, the organoids further increased in size (r = 160 µm) and matured as indicated by the presence of strong FABP1 expression (Figure 2a,b). Similar results were observed for proximal small intestinal organoids from the duodenum (Figure S4a,b). The whole‐mount staining of F‐actin and collagen on day 8 (Figure 2c) demonstrates individual organoid development supported by the hydrogel particles. Furthermore, the size of the organoids could be modulated by varying the numbers of cells per droplet (Figure 2d). Specifically, 5000 cells/µL (approximately 130 cells/droplet) resulted in organoids with a radius of ∼120 µm, while 7500 cells/µL (approximately 200 cells/droplet) produced larger organoids with a radius of ∼230 µm.
*Development of colon organoids in droplets and comparison to traditional dome cultures: (a) Brightfield images (scale bars 250 µm), H&E staining (scale bars 100 µm), and mIF staining (SOX9 turquoise, FABP1 yellow, and Ki67 red, scale bars 100 µm) of colon organoids in hydrogel droplets cultured in a bioreactor on days 3, 6, and 8. (b) Size development analysis of colon organoids over several days, with average size highlighted above (n organoids ≥ 235). (c) Whole‐mount staining of colon organoids in droplets on day 8. (d) Size development analysis of duodenum organoids formed from different amounts of encapsulated cells, with average size highlighted above (n organoids ≥ 58). (e) mIF staining of organoids in hydrogel droplets cultured in a bioreactor on day 8 (SOX9 turquoise, FABP1 yellow, Ki67 red, MUC2 green, CA1 orange, CHGA pink, ECAD purple, and CEACAM5 white, scale bars 100 µm). (f) Relative mRNA expression of different genes of organoids grown in droplets over several days normalized to β‐actin gene expression (2 biological replicates). (g) Quantification of mIF staining of organoids in droplets cultured in a bioreactor on day 6 and day 8 (n organoids ≥ 30). (h) Quantitative comparison of the coefficient of variation in organoid sizes between organoids grown in domes (1:12 dilution) and droplets on day 6. Each dot represents the coefficient of variation in a single experiment (n = 5 biological replicates). P‐value of < 0.05 was expressed as *, p‐value < 0.01 was expressed as ** and p‐value < 0.001 was expressed as ***. The normality of the distribution was verified with a Shapiro‐Wilk test, and a t‐test was done to calculate the p‐values. (i) Brightfield images (scale bars 500 and 250 µm) and H&E staining (scale bars 100 µm) of organoids grown in domes and droplets cultured in a bioreactor on day 8. (j) Quantitative fluorescence analysis of cell viability in intestinal organoids at days 3, 6, and 8 of culture in domes or droplets by live/dead staining (n samples = 4), and representative fluorescent images of organoids grown in domes and droplets cultured in bioreactor on day 8. P‐value of < 0.05 was expressed as *, p‐value < 0.01 was expressed as ** and p‐value < 0.001 was expressed as ***. The normality of the distribution was verified with a Shapiro‐Wilk test, and a t‐test was done to calculate the p‐values. (k) mIF staining of organoids in domes (1:12 dilution) and hydrogel droplets cultured in bioreactor on day 8 (SOX9 turquoise, FABP1 yellow, Ki67 red, and CA1 green, scale bars 100 µm). l–n) Bulk RNA sequencing of organoids grown in domes (1:6 dilution) and droplets (three technical replicates, two biological replicates). (l) Quantification of Log2 transformed expression levels and heatmap of z‐score normalized expression levels across of selected genes. Data were averaged across replicates. Statistical significance between conditions was assessed using the Wilcoxon test, with p‐values adjusted for multiple testing using the Benjamini‐Hochberg method. P‐value of < 0.05 was expressed as *, p‐value < 0.01 was expressed as * and p‐value < 0.001 was expressed as **. (m) Volcano plot of differentially expressed genes between domes (left) and droplets (right) subsets (p < 0.05, log2 Fold Change > 1). (n) Plot showing significantly enriched Gene Ontology biological processes of genes in colon organoids grown in domes (p‐value < 0.05, q‐value < 0.2).
A key benefit and hallmark of the intestinal organoid model is the ability to differentiate into various intestinal epithelial cell types. mIF of day 8 organoids (Figure 2e; Figure S4c) confirmed the presence of diverse cell types, including stem/undifferentiated cells (SOX9^+^), transit‐amplifying cells (Ki67^+^), colonocytes (FABP1^+^, CA1^+^), or enterocytes (FABP1^+^, APOA4^+^), enteroendocrine cells (CHGA^+^), and goblet cells (MUC2^+^). While the presence and number of crypts were not quantified due to the challenging assessment of the organoid shape upon compression within the droplet, we could localize undifferentiated cells (SOX9^+^) patterned toward the edges of the organoid and differentiated cells (FABP1^+^, CA1^+^) in the core (Figure 2e; Figure S4c). We also observed a higher number of goblet cells (MUC2^+^) in colon organoids and a higher prevalence of enteroendocrine cells (CHGA^+^) in small intestine organoids, which is concordant with previous reports [38, 39]. qPCR (Figure 2f) and image analysis of mIF staining (Figure 2g; Figure S4d) of the main canonical markers over time revealed downregulation in the expression of stemness markers (LGR5, SOX9), stable expression of proliferation markers (MKI67/Ki67), and upregulation of differentiation markers for goblet cells (MUC2), enteroendocrine cells (CHGA), and enterocytes/colonocytes (FABP1; FABP1, and CA1) by day 8, which highlight the ability of the organoids to homeostatically grow and differentiate within the droplets.
To further characterize the organoids and benchmark them to the traditional dome culture, we compared cells encapsulated in hydrogel droplets to those cultured in domes (Figure 2h,i). However, due to differences in volume and shape, monitoring and segmenting individual organoids in domes is challenging [5], especially in densely populated cultures. Additionally, organoids can be located in various z‐planes, making segmentation difficult due to image overlapping structures at various heights. This necessitates significant dilution of the cell suspension prior to dome seeding to prevent organoid merging and allow for individual monitoring, which in turn requires large and often costly volumes of hydrogels and media. While we diluted the cell suspension for dome culture 6‐ to 12‐fold, this still often resulted in difficulties in organoid segmentation at later time points (Figure S5). Therefore, in this study, we compared organoid sizes on day 6. Organoid generation in droplets showed high reproducibility across biological replicates (Figure S6a), with an average coefficient of variation (CV) of 20% for droplet‐derived organoids compared to 50% for dome‐grown organoids (Figure 2h). Organoids formed in droplets grew significantly larger (r dome = 46 µm, r droplet = 118 µm) and more uniform (CV dome = 59%, CV droplet = 19%) (Figure 2i; Figure S6b) compared to those in domes.
In intestinal organoid culture, domes create nutrient, growth factors, and oxygen gradients, leading to poorly perfused regions and resulting in variability in organoid size, morphology, and development [2, 8, 9, 10, 11]. Although various dome volumes are utilized in the literature, larger domes are prone to developing steeper gradients. We hypothesized that reducing the dome size would mitigate the formation of spatial gradients, thereby enabling a more accurate comparison with our droplet system. Consequently, dome cultures were restricted to 3.5 µL. This volume, though lower than standard protocols, was established as the minimum size required to achieve dome stability while minimizing these gradients. Nevertheless, we were able to observe differences between both culture approaches in terms of homogeneity, viability, and cell type localization. Although initial cell viability was lower in droplet cultures on day 3, likely due to stress from encapsulation, this pattern reversed by day 8 (Figure 2j; Figure S6c). The increased viability in droplets compared to domes by day 8 is likely due to higher proliferation rates and improved nutrient supply in the bioreactor droplet culture system. Live/dead fluorescence images showed a high number of dead cells at the center and increasing viability toward the edges (Figure 2j), indicating the presence of a necrotic core within the dome. Similarly, mIF stainings (Figure 2k) revealed notable differences in marker distribution between organoids grown in domes and those in droplets. In dome culture, the edges exhibited organoids with high SOX9 and Ki67 expression, indicating stemness and proliferation, whereas the center showed organoids with elevated levels of FABP1 and CA1 expression, suggesting increased differentiation due to limited growth factors. This pattern is well‐documented [2, 8, 11] and reproducible. In contrast, organoids in droplets contained markers for both proliferation and differentiation, resulting in more homogeneous single organoids. While significant differences were observed in individual patterns, mIF analysis (Figure S6d) showed less variability between groups overall. Organoids grown in droplets displayed slightly lower expression of differentiation markers (FABP1, CA1). To further investigate these differences, we performed bulk RNA sequencing (RNA‐seq) of day 8 organoids. Both groups displayed similar profiles of standard colon‐specific marker genes (Figure 2l), with differentiation markers slightly lower, and the proliferation marker (MKI67) upregulated in the droplet‐grown organoids, consistent with mIF and live/dead analysis. Prolonged culture (e.g. day 10) may further increase mature marker levels while maintaining stem/proliferative cells within droplet organoids. Interestingly, differential analysis (Figure 2m) revealed the presence of cytokine and chemokine genes (CXCL1, CXCL8, and CCL20) in dome conditions. Gene Ontology analysis of upregulated genes in domes (Figure 2n) indicated enrichment of genes associated with wound healing, hypoxia response, and extracellular matrix reorganization, aligning with the lack of oxygen and nutrient supply in dome centers. We also observed upregulation of genes associated with hypoxia response in droplets cultured in static conditions (Figure S6e–g), highlighting the importance of enhanced nutrient and gas exchange within the bioreactor.
Altogether, these findings confirm the biocompatibility and the feasibility of 3D cell culture using hydrogel scaffolds. The results underscore the advantages of culturing epithelial organoids individually in small droplets, providing a more controlled microenvironment, higher reproducibility, and increased homogeneity, which would benefit their subsequent study upon treatment. Moreover, integrating this method with a bioreactor setup offers potential for high‐throughput organoid generation, paving the way for scalable and standardized organoid production.
Tumor Organoids Development in Hydrogel Droplets
2.3
Having established droplet bioreactor culture for healthy intestinal organoids, we further applied the technique to patient‐derived tumor organoids (Figure 3a). Mirroring the development of healthy organoids, cell aggregates formed by day 3 (r = 71 µm) and matured into 3D structures by day 6 (r = 120 µm). Prolonged tumoroid cultures did not lead to an increase in the size of the organoids (Figure 3a,b), but resulted in the development of dense aggregates. H&E staining of the tumoroids confirmed their dense structure with a lack of luminal space (Figure 3c). This aligns with the fact that patient‐derived tumoroid cultures are known to adopt donor‐specific morphologies [40]. As expected, tumoroids grew larger in droplets probably due to the larger cell concentration, while tumoroid sizes in domes were significantly smaller (Figure 3c). In contrast to healthy intestinal organoids, which showed clear patterning and differentiation over time, tumoroids did not show a patterned cell distribution (Figure 3d). This is likely due to the lower amount of growth factors in the tumor organoid media, and particularly the lack of Wnt, which has been linked to the observed patterning [11]. To evaluate the potential use of the generated tumoroids as a screening platform, we transferred the fully developed organoid droplets to a 96‐well plate and treated them with SN‐38 [2], chemotherapy drugs Capecitabine, combination chemotherapy drug FOLFOX (5‐Fluorouracil:Leucovorin:Oxaliplatin 25:5:1) [23, 41], and target therapy drug Bortezomib [42] (Figure 3e). The tumoroids displayed heterogeneous sensitivity. Specifically, SN‐38 induced clear dose‐dependent cell death, whereas Bortezomib showed limited efficacy, and no significant cytotoxicity was observed with FOLFOX or Capecitabine. This differential response is likely attributed to donor‐specific biological resistance [21, 43, 44]. As the primary objective of this work is to establish the organoid generation technology rather than to perform deep biological profiling, further characterization of these resistance mechanisms was outside the scope of this study. However, overall, these proof‐of‐concept results demonstrate that the system can be adapted to different organoid cultures and highlight the platform's potential for screening applications.
Growth and morphology of colon tumor organoids: (a) Brightfield images of colon tumor organoids in hydrogel droplets on day 3, 6, 8, and 12 (scale bars 250 µm). (b) Size development analysis of colon tumor organoids over several days, with average size highlighted above (n organoids < 110). (c) Brightfield images (scale bars 250 µm) and H&E staining (scale bars 100 µm) of organoids grown in domes (1:6 dilution) or droplets. (d) mIF staining of organoids in hydrogel droplets and domes (1:6 dilution) on day 12 (ECAD white, Ki67 red, SOX9 turquoise, and CEACAM5 pink, scale bars 100 µm). (e) Drug treatment of bioreactor grown tumor organoids. Dose‐response curves to compounds FOLFOX (5‐Fluorouracil: Leucovorin:Oxaliplatin = 25:5:1, the concentration specified in the graph is based on 5‐Fluorouracil), SN‐38, Bortezomib, and Capecitabine, quantified by fluorescence analysis of cell viability by live/dead staining (n samples = 3). Schematic was created with BioRender.com.
Automated Oil Extraction Device and Organoid Spotting
2.4
The complete integration of droplet generation, polymerization, and oil extraction is essential to establish a truly high‐throughput, automated platform for uniform organoid culture. Despite the automation of microfluidic droplet generation, downstream processing often relies on manual protocols [29]. Current oil removal methods, including hydrophobic membrane absorption [32, 33] and chemical demulsification [30, 31], involve substantial manual intervention, introducing variability and the risk of material loss. To address these challenges and further automate the organoid generation process, we integrated the system with a polymerization chamber and a microfluidic oil‐removal device. We adapted the oil‐media coflow strategy described by Mohamed et al. [16], modifying it to accommodate temperature‐sensitive soft hydrogels and low surface tension oils (Figure 4a; Figure S7a). Compared to the original UV‐cross‐linked and mineral oil‐based systems, our approach required the integration of a long‐term polymerization chamber and specific geometric optimizations (detailed below). Furthermore, surface modification via a siliconizing reagent was essential to establish a highly hydrophobic interface, thereby enabling stable co‐flow with fluorinated oil.
*Oil Removal Device and Organoid Spotting: (a) Schematic overview of oil removal device integration. Schematic was partially created with BioRender.com. Brightfield image of coflow of oil and water fraction in extraction chamber (scale bars 500 µm). (b) Brightfield images of droplets in extraction with or without demulsifier pulsed into the device (scale bars 250 µm). (c) Comparison of hydrogel droplets generated with manual oil removal method or with the oil removal device. Brightfield images of encapsulated cells on day 0 (left). Brightfield images (scale bars 500 and 250 µm) and H&E staining (scale bars 250 µm) of colon organoids in hydrogel droplets cultured in a bioreactor on day 8 (right). (d) Quantitative fluorescence analysis of cell viability of intestinal organoids generated by manual oil removal or with oil removal device at days 0, 2, 5, and 7 of culture by live/dead staining (n organoids =11). P‐value of < 0.05 was expressed as *, p‐value < 0.01 was expressed as ** and p‐value < 0.001 was expressed as **. The normality of the distribution was verified with a Shapiro‐Wilk test and a t‐test was done to calculate the p‐values. (e) Size development analysis of organoids generated with the oil removal device over several days, with average size highlighted above (n organoids ≥ 240). (f) mIF staining of organoids in hydrogel droplets generated with the oil removal device on day 8. (g) Quantification of mIF staining of organoids in droplets with manual oil removal method or with oil removal device (n organoids ≥ 45). (h) Schematic of organoid distribution from bioreactor to a well plate by organoid spotting. (i) Representative brightfield images of organoids and tumoroids development after distribution in a well plate (scale bars 250 µm).
After droplet generation, the aqueous precursor droplets were introduced into the polymerization chamber (Figure S7b), where they were maintained at 37°C for 20 min. This chamber featured two outlets: one for waste disposal and the other connected to the oil removal device that leverages particle buoyancy. During droplet generation, the bottom waste outlet remained open, allowing oil to escape while droplets floated on top of the oil. Following polymerization, the waste outlet was closed, and the chip inlet was opened to direct the flow of polymerized particles into the oil removal device. The oil containing the hydrogel particles was co‐introduced into the oil removal device (Figure S7c) along with oil containing the perfluorooctanol (PFO) demulsifier, delivered using pulse‐width flow modulation. The demulsifier and oil were then mixed within a serpentine channel, where PFO disrupted the surfactants at the oil‐water interfaces of the droplets, destabilizing them. Consequently, when particles entered the extraction oil‐media coflow chamber, they exhibited a higher affinity for the aqueous medium, transitioning from the oil phase to the aqueous solution for collection.
The extraction process relied heavily on the presence of a demulsifier (Figure 4b). In its absence, droplets retained their oil affinity, leading to pulsatile flow or droplet extrusion through the pillars toward the oil outlets. However, in the presence of the demulsifier, droplets transitioned smoothly to the media or buffer for collection (Figure 4b). The device geometry (Figure S7d) was optimized to maximize droplet yield, minimize clogging, and reduce oil carryover. Effective coflow establishment and particle transfer relied on surface modification with chlorinated organopolysiloxane, a 1:6 chamber‐to‐input width ratio, and a low oil‐to‐media flow ratio, as well as a tapered chamber entrance, the latter of which proved essential for minimizing particle attachment to device structures. Overall, the device exhibited minimal oil carryover and reduced droplet loss compared to manual methods. Notably, the on‐chip oil removal resulted in a 2.5‐fold increase in organoid yield on day 8 compared to manual PTFE extraction. Despite being composed of a soft hydrogel, the droplets maintained their integrity post‐oil extraction (Figure 4c). Furthermore, cell viability following the extraction process was similar to manual extraction, exceeding 95% on day 0, suggesting that the device does not induce additional stress on the cells and that the incubation time with PFO has minimal cytotoxic effects on the resulting particles (Figure 4d). As anticipated, the encapsulated cells developed into colon organoids with diameters <200 µm over eight days in culture (Figure 4e), exhibiting morphology similar to those extracted manually or grown in standard dome culture (Figure 4c; Figure S8). Additionally, mIF staining of the droplets revealed characteristic colon organoid markers (Figure 4f,g), consistent with those obtained via manual methods (Figure 2e). These results indicate that the integrated microfluidic approach is readily applicable for generating hydrogel droplets suitable for single organoid growth.
Lastly, to simplify the workflow, we tested if our microfluidic‐based pipeline can work alongside liquid handling tools to transfer suspended organoids to well plates. This facilitates a seamless process from organoid generation to assay, allowing for applications such as drug screening, organoid imaging, and analysis typically performed in well‐plate formats. Thanks to the stability of the hydrogel particles, this transfer can happen either immediately after organoid precursor generation or later during culture. The latter method offers advantages, including improved oxygenation and nutrient supply in the bioreactor, easier media exchange, and the potential for organoid sorting prior to distribution. To assess the feasibility of transferring individual organoids, we used a commercially available benchtop dispenser robot (Dispen3D, Seed Bioscience, Figure 4h) to transfer organoids from the bioreactor to 96 well‐plates. Our results showed that the dispensing process does not compromise organoid viability, as they continue to grow after being transferred to the well plate (Figure 4i). This capability opens up future possibilities for applications such as drug testing, or co‐culture with individual or specific numbers of organoids.
In conclusion, the stability of the ECM mixture and the microparticles allows for a combination of different automation methods and high‐throughput droplet generation. These automation methods minimize manual handling and simplify organoid manipulation, which has potential to further enhance reproducibility of organoids and expand their applications.
Conclusion
3
Here, we introduced an integrated microfluidic process for generating hydrogel droplets that enables 3D culture, growth, and large‐scale production of intestinal organoids and tumoroids. By incorporating a distribution device before the droplet formation unit, we achieved more uniform cell distribution within the droplets. The integration of a hydrogel polymerization chamber and an oil removal device after droplet generation allows for hands‐free hydrogel production, eliminating manual handling and reducing droplet loss. Additionally, we optimized the ECM composition of the droplets to ensure their stability over time while maintaining proper organoid development, allowing for the culture of organoids in bioreactors and facilitating easy handling for downstream applications, such as transferring to well plates.
The established system reduces variability in organoid culture by providing uniform 3D scaffolds in a reproducible manner. Organoids grown in these droplets showed higher individual homogeneity in size and distribution of canonical markers compared to traditional dome cultures, which often exhibit organoids with higher differentiation and cell death in the center, and organoids with higher stemness at the edges of the dome. Furthermore, we observed lower expression of cytokines and genes associated with hypoxia response in droplet cultures, highlighting the importance of enhanced nutrient and gas exchange within the droplets. These findings underscore the advantages of culturing epithelial organoids individually in small droplets, creating a more controlled microenvironment that enhances reproducibility and homogeneity. However, extending this characterization to detailed morphometrics, particularly crypt formation, would be valuable for further refining metrics of organoid homogeneity. Moreover, the potential of this system to support long‐term culture via bioreactor‐enhanced transport represents an exciting avenue for future study. This optimization could facilitate the emergence of rare cell types or lineages requiring prolonged differentiation, such as late colonocytes.
We believe this work provides proof‐of‐concept for an integrated microfluidic process for generating hydrogel droplets for human organoid and tumoroid culture. This system has the potential to be applied to various epithelial cell organoids, as well as co‐culture systems involving stromal cells or immune cells. Our results suggest that our approach opens up potential avenues for scalable, high‐throughput, and reproducible culture of diverse epithelial organoids and tumoroids. Such capabilities are particularly valuable for advancing application potential in drug screening, genetic engineering, and personalized medicine, where consistent and efficient organoid production is essential.
Materials and Methods
4
Material
4.1
All the material is specified in Table S1.
Human Tissue Samples
4.2
The list of donors is specified in Table S1. Human intestinal tissue resections, along with concurrent data collection and experimental procedures, were conducted within the framework of the non‐profit foundation HTCR (Munich, Germany) and University Center for Gastrointestinal and Liver Disease (Clarunis; Basel, Switzerland), including informed patient consent. The framework of the HTCR Foundation has been approved by the ethics commission of the Faculty of Medicine in the Ludwig Maximilian University (no. 025‐12) and the Bavarian State Medical Association (no. 11142). The framework of the University Center for Gastrointestinal and Liver Disease was approved in accordance with the Helsinki Declaration and reviewed and approved by the ethics committee (Ethics Committee of Basel, EKBB, no. 2019‐02118).
Visceral surgery with partial resection of intestine was performed on these oncology patients, from which we used micro‐ and macroscopically tumor‐free tissue as well as tumor material.
Organoid Culture and Tumoroid Culture
4.3
Human intestinal organoid lines were maintained in organoid media as previously described [45] up until passage 20 at 37°C, 5% CO_2_, and 95% humidity incubators, with a few modifications (Table S1). The organoids lines were regularly tested for mycoplasma by MycoAlert PLUS Mycoplasma Detection Kit (Lonza) according to the manufacturer's instructions.
Tumor organoid lines were maintained until passage 26 in IntestiCult Organoid Growth Medium (Human, StemCell Technologies) without supplement and Advanced DMEM:F12, 1x Glutamax in a ratio of 1:1 supplemented with 0.05 mg/mL Primocin (Invivogen).
Healthy organoids and tumor organoids were fragmented by Gentle Cell Dissociation reagent (Stemcell Technologies) weekly at 1:4–1:8 depending on the confluence and growth rate. Washing steps were performed with Advanced DMEM:F12 (Gibco) with 0.01 M Hepes (Gibco), 1x Glutamax (Gibco), and 100 U Penicillin and Streptomycin (Gibco) referred to as DMEM:F12+++. The fragments were seeded on Tissue Culture treated well plates (CytoOne) in 25 µL domes in 100% growth factor‐reduced Matrigel for healthy organoids (Corning) or Cultrex Basement Membrane Extract, PathClear for tumor organoids (R&D Systems), respectively. After 15–30 min polymerization in the incubator, organoid media supplemented with 10 µM Y‐27632 was added. Media exchange was performed every two to three days until use.
Fabrication of Microfluidic Devices
4.4
The molds were generated using standard photolithography, based on the supplier protocol. The silicon wafers were primed in an oxygen plasma machine for 7 min (ATTO RF, Diener electronic GmbH), and SU‐8 photoresist (GM1075, Gersteltec Sàrl) was spin‐coated onto the wafer to yield the desired height. After relaxation and soft bake, the spin‐coating process was repeated for wafers that required heights above 200 µm. The final heights for the different devices were as follows: distribution device (200 µm), droplet device (250 µm) (Figure S1c), and oil removal device (400 µm; Figure S7d). Wafers were exposed with a laser writer (µMLA, Heidelberg Instruments), followed by a post‐exposure bake. The unexposed photoresist was developed using propylene glycol monomethyl ether acetate (Sigma–Aldrich), and the resulting wafers were hard baked. Detailed steps are provided in Table S1.
The microfluidic devices were fabricated with polydimethylsiloxane (PDMS, Sylgard) by soft lithography. Each of the wafers was treated with chlorotrimethylsilane (Sigma–Aldrich) and covered with a PDMS solution of 10:1 elastomer to cross‐linker ratio. The molds were then degassed in a desiccator and baked at 70°C for a minimum of 2 h. The polymerized PDMS chips were removed from the molds, punched to create inlets and outlets, and plasma‐bonded to a PDMS‐coated glass slide.
Organoid Culture in Hydrogel Droplets
4.5
Organoids were washed with DMEM/F12+++, collected, and centrifuged. Then, organoids were resuspended in TrypLE Express Enzyme (1X) (Gibco) with 350 U DNAse (Roche), 20 µM Y‐27632 (STEMCELL Technologies), and 1.5‐2 mM DL‐Dithiothreitol solution (DTT) (Sigma–Aldrich) for 8 min at 37°C and an extra 5 min after harsh pipetting. After pelleting, cells were resuspended in DMEM:F12+++, strained through a 40 µm cell strainer (Pluriselect), centrifuged, and resuspended in organoid media with 20 µM Y‐27632 and counted with the Cell Countess (Invitrogen). Collagen I solution (Biomatrix, 6.2 mg/mL) was neutralized with Neutralization Solution (Biomatrix). Cells resuspended in organoid media were mixed with ECM components to create the final ECM‐cell mixture. This mixture consisted of 1.5 mg/mL neutralized collagen, 25% (v/v) growth factor‐reduced Matrigel (Corning), and a final cell concentration of 5x10^6^ cells/mL (unless indicated otherwise). The mixture was then subjected to the encapsulation and oil removal process.
Organoid media was supplemented with 10 µm Y‐27632 for the first three days. Droplets were cultured using the CERO bioreactor (80 RPM with 2 s rotation period), unless stated otherwise. Static droplets were cultured in ultralow attachment six‐well plates (Corning). Every three days, 70% of organoid media was changed by centrifuging the droplets in media at 190 RCF for 3 min, prior to the media exchange.
Domes were used as controls. Specifically, we diluted the initial cell‐hydrogel suspension 1:6 (8 x 10^5^ cells/mL) and 1:12 (4 x 10^5^ cells/mL) ratios and seeded two 3.5 µL domes in a well of a Tissue culture‐treated 24‐well plate (CytoOne). The domes were polymerized for 20 min at 37°C, and 500 µL of organoid media with 10 µM Y‐27632 for the first three days was added. Every three days 100% of the media was renewed.
Brightfield images were taken with the Leica DMi (Leica) and Nikon Ti2W (Nikon).
The cell distribution was measured from brightfield images of droplets in media on day 0, by mean intensity approximation. The individual droplets were segmented by morphological operations with Canny edge detection and particle analysis (FIJI), and then the mean intensity was measured for each droplet.
Organoid size was measured from brightfield images by segmentation (Figure S5a), either by Halo AI for organoids in droplets (Indica Labs, version 3.6) or in Python for organoids in domes. In brief, when using HALO AI, single organoids were automatically detected using a deep learning algorithm (DenseNet AI V2 Plugin) trained to distinguish organoids from background. When using Python, organoids were detected in brightfield images as follows: images were inverted, the scale was normalised by applying quantiles on the image histogram, and Multiotsu thresholding. Detected regions were filtered according to size and aspect ratio to filter out segmentation artifacts. All steps in Python were performed using algorithm implementations from scikit‐image library. Detection was manually validated, and those objects that did not correspond to organoids were removed.
Further analysis was performed with Python. The variability in organoid size and cell distribution was expressed in terms of the coefficient of variation, defined as the ratio of the standard deviation divided by the mean value.
Microfluidic Encapsulation
4.6
The encapsulation chip was functionalized using 1H, 1H, 2H, 2H‐Perfluoro‐1‐octanol (Merck/Sigma) for 30 min, washed with deionized water, and dried. The distribution device was primed with 1% bovine serum albumin (BSA) (Miltenyi Biotec) in phosphate buffered saline (PBS) (Gibco) for at least 5 min. Microfluidic encapsulation was performed via flow focusing device as indicated in Figure S1a, with detailed experimental setup provided in Table S1. Pressure and flow rate were controlled with Flow EZ pressure controllers and flow unit (Fluigent).
The 3M Novec 7500 oil supplemented with 1.5% w/v dSURF surfactant (Fluigent) was injected into the flow focusing device at 15 µL/min. The hydrogel cell suspension was kept on ice and injected at a pressure of 250 or 300 mbar, through a water‐cooled tube, to the flow focusing device or to the distribution device, respectively. The hydrogel precursor was sheared into monodisperse droplets at the merging points of both fluids in the flow focusing device. The droplets were collected in an Eppendorf tube or a polymerization pool and left at 37°C for 20 min to polymerize.
To remove droplets from oil (Figure S1b), the polymerized droplets were passed twice over a hydrophilic PTFE membrane (Merck/Millipore) in a volume of 50–75 µL to enable oil uptake of the PTFE membrane and then droplets were resuspended in desired media, or passed through oil removal device as described below.
On‐Chip Oil Removal
4.7
Before use, the oil‐removal chip was treated with Sigmacote (Merck/Sigma) for 30 s, dried, and rinsed with deionized water. Upon connection to the fluidic system, a droplet‐free co‐flow was established (details below) to equilibrate the system and coat the extraction chamber with BSA.
After hydrogel droplets were generated, the droplets were collected in an incubation chamber (Figure S7b and Table S1), placed on top of a heating unit (Inheco), covered with a portable microscope incubator (OkoLab stage top incubator), and polymerized for 20 min at 37°C. After polymerization the droplets in oil were introduced into the oil removal device (Figure S7c), at the same time 20% (v/v) perfluorooctanol (PFO)‐oil solution was pulsated into the device, using on/off state of 2‐way microfluidic valve (2‐SWITCH, Fluigent). Both fluids were mixed in the serpentine region of the chip, before entering to the extraction chamber, where a coflow was established between 1% BSA in Hanks' Balanced Salt Solution (HBSS, Gibco) and the droplet solution, allowing the droplets to transfer to the water‐based fraction. After oil extraction, the solution containing droplets was collected, spun, and replaced with culture media.
RNA Isolation, Quantitative PCR, and Bulk RNA Sequencing
4.8
For quantitative PCR (qPCR) and bulk RNA sequencing, 2 mL of organoid droplets suspended in media and 2 domes (1:6 dilution, 3.5 µL sized) were collected in 1% BSA‐coated Eppendorf tubes, centrifuged, and resuspended in a dissociation mixture. The dissociation mixture was composed of 5 mg/mL Collagenase type II (Thermo Fisher Scientific), 1.84 U/mL Dispase (Gibco), 800 U/mL DNase (Merck), 20 µM Y‐27632, 0.1% BSA (Miltenyi Biotec), and HBSS (Gibco). The organoid‐dissociation mixture was incubated for 30 min at 37°C and 500 rpm. Throughout the incubation time, organoids were additionally mechanically dissociated by pipetting every 5 min. Dissociated cells were pelleted and resuspended in the desired lysis buffer and frozen at −20°C or −80°C.
For qPCR experiments, 2 mL of organoid droplets suspended in media were used for RNA collection. All material was collected in Eppendorf tubes containing 350 µL RNA lysis buffers (Zymo Research). RNA was extracted using the Quick‐RNA MiniPrep (Zymo Research) following the manufacturer's instructions. qPCR analysis was performed using biological duplicates and technical triplicates. cDNA synthesis was performed using the Transcriptor First Strand cDNA Synthesis kit (Roche). qPCR was run on the LightCycler 96 (Roche Diagnostics) using LightCycler 480 SYBR Green I Master (Roche). β‐actin was used as a housekeeping gene. Primers are shown in Table S1.
For bulk RNA sequencing, the samples were processed with Organoid DRUG‐seq service (ALITHEA genomics). Reads were mapped to the human GRCh38 genome assembly. Differential gene expression analysis was performed using the DESeq2 package [46].
Organoid fixation and Histogel embedding
4.9
The organoid domes and droplets were gently washed once with 1X PBS and subsequently fixed overnight at 4°C in 4% paraformaldehyde (Sigma). After fixation the material was stored at 4°C, before use. The collected material was embedded into a HistoGel (Epredia) microarray mold, as previously described [47, 48]. The molds were processed on an automated tissue processor HistoCore PEARL (Leica), and embedded in paraffin. The formalin‐fixed paraffin‐embedded (FFPE) blocks were sectioned on a microtome (3 µm) and mounted onto Superfrost Plus Gold (Epredia).
Hematoxylin and Eosin Staining
4.10
H&E staining was done manually using the H&E Staining Kit (Abcam). The sections were deparaffinized and hydrated in distilled water. The sections were incubated in Hematoxylin for 5 min and rinsed two times in water. Then, the sections were incubated in Bluing Reagent for 20 s and rinsed and rinsed two times in water. The slides were dipped in absolute alcohol, and the excess was blotted off. The section incubated in Eosin Y Solution for 2–3 min and rinsed, and dehydrated using absolute alcohol. Finally, the slide was cleared, mounted in synthetic resin, and scanned by a brightfield whole‐slide scanner (Hamamatsu, NanoZoomer S360).
Multiplex Immunostaining
4.11
mIF was automated on Ventana Discovery Ultra automated tissue stainer (Roche Tissue Diagnostics), as described in [47, 48]. Slides were baked first at 60°C for 8 min and subsequently further heated up to 69°C for 8 min for subsequent deparaffinization with Discovery Wash (Ventana). This cycle was repeated twice. Heat‐induced antigen retrieval was performed with Tris‐EDTA buffer pH 7.8 (CC1, Ventana) at 92°C for 40 min. After each blocking step with the Discovery Inhibitor (Ventana) for 16 min, the Discovery Inhibitor was neutralized. Primary antibodies (Table S1) were diluted in 1X Plus Automation Amplification Diluent (Akoya Biosciences). Primaries were detected using according to anti‐species secondary antibodies conjugated to HRP (OmniMap anti‐Rabbit HRP, Ventana; OmniMap anti‐Mouse HRP, Ventana; OmniMap anti‐Rat HRP, Ventana; and OmniMap anti‐Goat HRP, Ventana). Subsequently, the relevant Opal dye (Opal 480, Opal 520, Opal 570, Opal 620, Opal 690, and Opal 780, Akoya Biosciences) was applied. After the first sequence, the steps are repeated in each subsequent sequence: Antibody neutralization and HRP denaturation step was applied to remove residual antibody and HRP, Inhibitor, Antibody, Secondary HRP, Opal dye. In the sixth sequence, instead of opal dye, Opal TSA reagent was applied, and immediately after that, Opal dye 780 (Akoya Biosciences) was applied. Lastly, samples were counterstained with 4',6‐Diamidino‐2‐phenylindole (Ventana). Sequential order of the primary antibodies, as well as corresponding dyes, was determined during establishment runs. Neutralization of HRP and denaturation of the proteins were performed after every primary antibody cycle in order to avoid cross‐bleeding and cross‐reacting antibodies. Slides were mounted manually using ProLong Gold Antifade Mountant (Thermo Fisher Scientific). Slides were dried for at least 2 h prior to imaging.
Image analysis of mIF images was performed with HALO AI (Indica Labs, version 3.6). Briefly, single organoids were automatically detected using a deep learning algorithm (DenseNet AI V2 Plugin) trained to distinguish organoids. Following validation, organoids were annotated as individual regions of interest.
The HighPlex FL v.4.2.14 module was used to perform nuclear segmentation based on DAPI+ cells (assisted by HALO's integrated AI‐default ‘nuclear segmentation type’) and specific marker identification. For quantification, DAPI+ nuclei and markers for each distinct cell type of interest were merged (taking membranous and nuclear signals into account). Positive detection thresholds were established for each of the markers, taking into account their nuclear or cytoplasm nature and integrated into the HighPlex FL analysis module. The HighPlex FL analysis module was deployed on previously generated regions of interest of the organoids using integration of the classifier in the module. Further analysis was performed with Python.
Whole Mount Staining
4.12
Permeabilization and blocking were next performed by incubating organoids in 1x PBS containing 0.5% Triton X‐100 (Sigma) for 2 h at room temperature and blocked with 2% normal donkey serum in PBS overnight at 4°C. All stainings were performed in a blocking buffer (2% normal donkey serum in PBS, 0.05% Triton X‐100). Samples were incubated with conjugated primary antibody (Table S1) for 32–48 h, washed three times for at least 2 h with PBS, followed by incubation in blocking buffer containing DAPI (1:1000, Invitrogen), Hoechst (1:1000, Thermo Fisher Scientific), and Alexa Fluor Plus 647 Phalloidin (1:2000, Thermo Fisher Scientific) for 24 h. The organoids were cleared by D fructose (Sigma) for a few hours or overnight and imaged using an CV8000 confocal microscope (Yokogawa). Image contrast was adjusted for visualization purposes.
Cell Viability
4.13
Cell viability was assessed by LIVE/DEAD Viability/Cytotoxicity Kit (Thermo Fisher Scientific) according to the manufacturer's instructions. Organoid cultures were washed with DMEM:F12+++ and the Calcein AM, ethidium homodimer‐1, and DAPI were added in DMEM:F12+++. The incubation time ranged from 30 to 90 min at 37°C. To stop the staining, a wash with base media was performed. Live/dead staining was visualized on Nikon Ti2 CSU‐W1. The maximum intensity z‐projection fluorescence images were analysed in FIJI. Positive life threshold and positive dead threshold were established, and the number of pixels above the threshold was counted. The viability was expressed as the percentage of live cells, where the number of live positive pixels was divided by the total number of positive pixels.
Drug Testing
4.14
Tumoroids in droplets, grown in a bioreactor for 12 days, were transferred into a 96‐well ultralow attachment U‐bottom plate (Primesurface) by a multichannel pipette, and cultured for additional 24 h in the well plate, before treatment. The tumoroids were treated with 3 individual drugs and one combinatorial drug treatment. The highest drug concentrations were chosen as follows: 1 µM Bortezomib (Tocris), 200 µM SN‐38 (Tocris), 50 µM Capecitabine (Selleckchem), and the combinatorial FOLFOX with 50 µM 5‐Fluorouracil (Tocris), 2 µM Oxaliplatin (Tocris), 10 µM Leucovorin (Selleckchem). We included a no‐treatment control and vehicle control with DMSO. The treatment was performed on technical triplicates and incubated for 48 h. Cell viability assessment, imaging, and analysis were executed as described above.
Organoid Spotting
4.15
Organoids were spotted using the Dispen3D device from SEED Bioscience, according to the standard protocol. Shortly, around 400 organoids were resuspended in 1 mL of methylcellulose‐based medium, loaded in the machine, and dispensed into a 96 ultralow attachment well plate, based on impedance measurements, with the goal to achieve dispersion of 1 organoid per well. A threshold for organoid detection was established manually based on the size of the organoids.
Author Contributions
B.L. and J.L.G.C. conceived the study. The manuscript was written, and the data were analyzed by B.L., H.K., and M.C.N. with the support of R.L.S., J.G.C., and J.L.G.C. B.L. and H.K. designed and performed most of the experiments with the help of M.C.N. and R.L.S. The microfluidic devices were designed, fabricated, and optimized by M.C.N., E.C., and L.B.Z. I.C. performed mIF stainings.
Conflicts of Interest
All authors are employees of F. Hoffmann‐La Roche Ldt. The company provided support in the form of salaries for authors but did not have any additional role in the study design, data collection and analysis, decision to publish or preparation of the manuscript. F. Hoffmann‐La Roche Ldt. has filed a patent application on the technology described herein. B.L., H.K., M.C.N., E.C., L.B.Z., and J.L.G.C. are named as inventors on the patent application.
Declaration of AI‐Assisted Technologies in the Writing Process
AI‐assisted technologies were used to spell‐ and language‐check written text. Following the use, the text was thoroughly reviewed and revised as necessary, assuming full responsibility for the publication's content.
Supporting information
Supporting File 1: advs73411‐sup‐0001‐SuppMat.docx.
Supporting File 2: advs73411‐sup‐0002‐SuppTable1.xlsx.
Supporting File 3: advs73411‐sup‐0003‐SuppMat.docx.
The reference list from the paper itself. Each links out to its DOI / PubMed record.
- 1N. Gjorevski , N. Sachs , A. Manfrin , et al., “Designer Matrices for Intestinal Stem Cell and Organoid Culture,” Nature 539 (2016): 560–564, 10.1038/nature 20168.27851739 · doi ↗ · pubmed ↗
- 2J. Y. Co , J. A. Klein , S. Kang , and K. A. Homan , “Suspended Hydrogel Culture as a Method to Scale up Intestinal Organoids,” Scientific Reports 13 (2023): 10412, 10.1038/s 41598-023-35657-9.37369732 PMC 10300005 · doi ↗ · pubmed ↗
- 3M. Hofer and M. P. Lutolf , “Engineering Organoids,” Nature Reviews Materials 6 (2021): 402–420, 10.1038/s 41578-021-00279-y.33623712 PMC 7893133 · doi ↗ · pubmed ↗
- 4P. Kakni , R. Truckenmüller , P. Habibović , and S. Giselbrecht , “Challenges to, and Prospects for, Reverse Engineering the Gastrointestinal Tract Using Organoids,” Trends in Biotechnology 40 (2022): 932–944, 10.1016/j.tibtech.2022.01.006.35221125 · doi ↗ · pubmed ↗
- 5T. Kassis , V. Hernandez‐Gordillo , R. Langer , and L. G. Griffith , “Orga Quant: Human Intestinal Organoid Localization and Quantification Using Deep Convolutional Neural Networks,” Scientific Reports 9 (2019): 12479, 10.1038/s 41598-019-48874-y.31462669 PMC 6713702 · doi ↗ · pubmed ↗
- 6N. Landon‐Brace , N. T. Li , and A. P. Mc Guigan , “Exploring New Dimensions of Tumor Heterogeneity: The Application of Single Cell Analysis to Organoid‐Based 3D In Vitro Models,” Advanced Healthcare Materials 12 (2023): 2300903, 10.1002/adhm.202300903.37589373 PMC 11468421 · doi ↗ · pubmed ↗
- 7J. M. Matthews , B. Schuster , S. S. Kashaf , et al., “Organo ID: A Versatile Deep Learning Platform for Tracking and Analysis of Single‐organoid Dynamics,” PLOS Computational Biology 18 (2022): 1010584, 10.1371/journal.pcbi.1010584.PMC 964566036350878 · doi ↗ · pubmed ↗
- 8S. E. Park , S. Kang , J. Paek , et al., “Geometric Engineering of Organoid Culture for Enhanced Organogenesis in a Dish,” Nature Methods 19 (2022): 1449–1460, 10.1038/s 41592-022-01643-8.36280722 PMC 10027401 · doi ↗ · pubmed ↗