Organoids are three-dimensional (3D) structures derived from stem cells that emulate the architecture and function of the organ of origin [1]. These structures are a natural progression from the human-relevant model provided by two-dimensional (2D) cell cultures, and are expected to have a lasting impact on biomedical research. The ability to grow miniature human organs in vitro has dramatic implications for the study of human development as well as disease modelling, drug development and regenerative medicine.
Organoids are normally derived from pluripotent stem cells (PSCs) that can be further subdivided into embryonic stem cells (ESCs) and induced PSCs (iPSCs) derived from adult cells, or adult stem cells (ASCs) [2,3]. With the right environmental cues and growth factors in the right combinations, these cells can differentiate into different cell types and self-organize into structures resembling miniature organs. Organoids have been derived from the brain [3,4], liver [5,6], intestine [7,8], kidney [9,10], ear [11,12], and many other organs.
The decades-long journey toward growing organ-like structures from stem cells started with key advances in stem cell biology in the 1980s and 1990s, which provided a foundation for organoids. Among these advances were the discovery in 2006 of iPSCs by Takahashi and Yamanaka [13] in Japan, which let scientists reprogram adult cells into a pluripotent, embryonic-like state, and the first identification of transcription factors regulating organ signaling pathways.
The promise of organoid technology to transform biomedical research was apparent in these early successes, with more physiologically relevant models to study human biology and disease. The translational potential of organoids is broad and diverse. Some of the most important fields of biomedical research already characterized by important findings achieved with organoids are disease modelling, drug screening and toxicology, personalized and regenerative medicine.
Photonics is the science and technology involving the generation, manipulation and detection of photons, elementary particles of light [14]. Photonics spans a broad range of applications, from telecommunications and information processing to medical imaging and diagnostics. In this context, photonics enables the development of sophisticated imaging and manipulation strategies that allow researchers to see and interact with biological systems at unprecedented spatial and temporal resolutions. Photonics relies on tools such as lasers, optical fibers, photodetectors and imaging systems [15]. These technologies exploit the unique optical properties of light—namely, its ability to travel through different media, interact with matter and carry information—to probe and manipulate the states and dynamics of biological structures and processes.
The timeline of the development of photonics in biomedical research can be divided into several milestones. The invention of the laser in the 1960s spawned a new era in optics, providing a coherent and highly confined light source that could be used in a variety of applications [16]. In the following decades, laser technology and optical instrumentation saw continual improvement, leading to the development of more advanced imaging methods such as confocal microscopy [17], two-photon microscopy [18] and optical coherence tomography (OCT) [19]. In the 1990s and 2000s, super-resolution microscopy techniques—such as stimulated emission depletion (STED) microscopy [20] and photoactivated localization microscopy (PALM) [21]—started to emerge, permitting researchers to break the diffraction barrier of classical optical microscopy, and ultimately visualize cellular structures at nanometer-scale resolution.
Photonics is now a key tool of biomedical research, enabling improvements in our ability to understand biological processes and to develop and apply new diagnostic and therapeutic approaches. Important examples include the use of photonics for: high-resolution imaging to study cellular interactions and dynamics in living tissues; functional imaging to study molecular interactions, dynamics of proteins and their functions, or cellular signaling pathways in real time; optical manipulation to provide tools to alter the behavior of biological systems and to trap and manipulate microscopic particles; diagnostic to image tissues non-invasively; and in therapeutics to harness the ability of light interacting with photosensitizing agents to selectively destroy cancer cells and other diseased tissues.
The confluence of organoid technology with photonics-based imaging and manipulation approaches can result in a game-changing paradigm shift in biomedical research by opening windows into complex biological processes, modelling disease and developing new therapeutic approaches. In this section, we focus on the applications of photonics in organoid research, and exemplify how they can be used to investigate human development, disease mechanisms, drug screening, personalized medicine and regenerative therapies.
Organoids are among the most powerful biomedical research tools, and they have become the basis for 3D in vitro models of human organs that recapitulate the architecture and functions of these tissues for studying developmental processes, disease mechanisms and therapeutic interventions. Obtaining real-time insights into the dynamic transformations that lead to organoid formation and maturation is a significant goal of organoid research because these events are difficult to capture and understand when they take place at the single-cell level. Real-time tracking is also important for gauging how an organoid respond to an extrinsic cue (such as activation of a protein or injection of growth factors), as well as its sensitivity to therapeutic agents. Some of the most informative insights into the cellular and molecular dynamics in organoid systems have recently come from using advanced imaging approaches.
One of the most useful imaging methods for organoid studies is light sheet fluorescence microscopy (LSFM), which uses a laser beam to generate a thin sheet of light that can be swept across the specimen [22]. This has revolutionized organoid imaging by providing high-resolution, 3D images of live tissues with minimal photo damage. Unlike confocal imaging, in LSFM, most cells do not have to be imaged at once (Table 1, Fig. 1), which reduces the need for a very high-powered laser. This in turn minimizes exposure of cells to high levels of light and allows for imaging sessions of several hours to days. Consequently, LSFM is well-suited to monitoring organoid development over extended periods. Using LSFM, Held and colleagues [23] were able to characterize the development of kidney organoids over several days, by following proliferating cells, differentiating cell types and tissue morphogenesis. Their study revealed the power of LSFM at capturing images of cell clusters and tissue structures over time. LSFM allowed for a much clearer understanding of long-term developmental processes and showed how kidney organoids form complex tissue structures, similar to those in vivo.
Table 1 . Comprehensive comparison of photonics imaging technologies for organoid analyses
| Features | Conventional confocal microscopy | Light-sheet fluorescence microscopy | Fourier light-field imaging | STED | PALM | STORM |
|---|---|---|---|---|---|---|
| Mechanism | Laser scans sample, collecting fluorescence from a single focal plane | Illuminates thin slices of the sample, collecting fluorescence perpendicular to illumination plane | Captures 3D light-field data using Fourier optics for wide-field 3D imaging | Depletion of fluorescence with a donut-shaped laser | Stochastic activation of photoactivatable proteins | Stochastic activation of standard organic dyes |
| Resolution | ~200-300 nm (limited by diffraction) | ~400-600 nm (depending on optics) | ~500 nm in depth, ~1 µm laterally | 20-30 nm | 10-50 nm | 20-30 nm |
| Fluorophores | Standard fluorophores | Standard fluorophores | Standard fluorophores | Standard fluorophores | Photoactivatable fluorescent proteins | Standard organic dyes |
| Imaging speed | Moderate to fast | Fast | Fast | Faster for continuous imaging | Slower (due to repeated activation cycles) | Slower (similar to PALM) |
| Laser intensity | Moderate to high | Low to moderate | Low to moderate | High (can lead to photobleaching) | Low to moderate | Low to moderate |
| Best for | 2D or 3D imaging of fixed samples, live-cell imaging | Fast 3D imaging, large volumes | 3D imaging of large volumes, high-speed | Fixed and live-cell imaging | Live-cell imaging, single-molecule tracking | High-resolution imaging, multi-color |
| 3D imaging capability | Limited 3D with optical sectioning | Excellent for 3D volumes | Excellent for 3D volumetric data | Limited | Yes | Yes |
| Disadvantages | Limited by diffraction Photobleaching in thick samples Limited depth penetration in live tissue | Relatively low resolution compared to other techniques Limited to optically transparent samples Requires specialized optics | Limited resolution for subcellular structures Requires complex optical setup Higher computation demands | High laser power can cause photobleaching and photodamage Expensive, complex setup Requires precise alignment of lasers | Long acquisition time due to repeated activation cycles Requires special photoactivatable proteins May not be suitable for thick samples | Long acquisition time with repeated dye activation Careful dye selection is required Photobleaching of dyes can occur over time |
STED, stimulated emission depletion; PALM, photoactivated localization microscopy; STORM, stochastic optical reconstruction microscopy; 3D, three-dimensional; 2D, two-dimensional.
Figure 1. This graphic demonstrates the differences between confocal microscopy, light sheet microscopy and Fourier light-field imaging. (A) Confocal microscopy: a laser moves focus point by point, generating a three-dimensional (3D) image through serial point sampling. Each point is rendered in green effect. (B) Light sheet microscopy. Diagram showing how a very thin sheet of light illuminates a thin slice of the sample as it moves through the sample plane. This light sheet (green) will allow for acquisition of high-resolution slices that can be pulled together for reconstruction of a 3D image. (C) Fourier light-field imaging shows that light from the entire 3D volume is captured at once from multiple angles (arrows) and reconstructed into a 3D image that encodes both intensity and angular information.
As single molecules are aggregated into larger entities, they acquire new properties that wouldn’t exist in their single-molecule form. This concept can be complemented by single-molecule, live-cell imaging, which lets us examine how individual molecules move, interact and organize within live cells. With this technique, Liu et al. [24] analyzed the dynamics of live cells within organoids at the single-molecule level. Using highly sensitive fluorescence microscopy, they were able to track the movement, interacting dynamics and organization of individual molecules in real time. From these observations, the researchers gained new insights into the molecular processes underpinning organoid differentiation—how organoids form complex tissue structures—and into how combinations of molecular interactions drive these changes. Understanding these molecular-level mechanisms, which are often obscured by the complexity of tissues, is vital to gaining insight into the precise mechanics of organoid biology. This could, in turn, facilitate identifying specific regulatory molecules that could be targeted for therapeutic development in humans.
Another advanced imaging technique that has been used to great effect in organoid research is Fourier light-field imaging, which can be used for volumetric imaging of organoids [25]. This powerful technique comes close to capturing the same visual details of organoids that are possible in whole organs, allowing for the analysis of cellular morphology and dynamic processes in 3D. Fourier light-field imaging utilizes a hybrid point-spread function for imaging that improves the depth-of-focus and lateral resolution of the image (Table 1, Fig. 1), as well as the ability to capture information from deep within organoids. Using this technique, Liu et al. [26] imaged human organoids and captured images of their complex 3D structures and dynamic processes. Liu and colleagues [26] highlighted that, when it comes to understanding cellular behavior, researchers may be missing important insights if they are not using volumetric imaging. Cells in an organoid organize into different tissue layers and replicate the overall architecture of the organ, which can only be fully understood when imaged in 3D. Spatial and temporal changes to this 3D structure, including both outward cell movement and inward vacuolization to form lumen-like systems, are necessary for organoid development and functioning. Without imaging these dynamic changes, researchers may struggle to fully appreciate what’s occurring within the tissue.
Fluorescent proteins have also played a prominent role in organoid tracking in real time. Thanks to their ability to be genetically encoded, most notably from the green fluorescent protein family, organoids can be programmed with high precision to visualize specific cellular structures or processes in real time. In his review, Nienhaus and Nienhaus [27] describes the use of photoactivatable fluorescent proteins that can be switched from a reversibly non-fluorescent ground state to a reversibly fluorescent state by an external trigger. The light can be applied with extraordinary temporal and spatial precision, which is especially important for super-resolution imaging, when it’s critical to control what regions activate fluorescence at what time. The superior control of light has enabled researchers to study fast motion or transient events within organoids at resolutions that surpass the diffraction limit of light (theoretically, light can’t be focused down more than about 200 nanometers, or less than half the width of a human hair). Organoid super-resolutions microscopy has started to reveal details of dynamic cellular events in organoids, such as cell divisions, cell migratory behaviors and their differentiation, with unprecedented detail. Among other things, such studies have revealed the temporal regulation of organoid development, and the complex interplay of cellular behaviors that typically construct or remodel functional tissue.
However, alongside these labelled imaging methods, there has been an uptake in the use of label-free imaging methods for organoid research. Label-free imaging methods make use of the natural optical properties of endogenous cellular components, such as their absorption, scattering or autofluorescence properties, to generate images. In many cases, this means that cells can be imaged without the need to use labels or dyes. This approach can be particularly advantageous for long-term studies since there are concerns that the introduction of exogenous labels into systems can lead to artefacts caused by the presence of these labels. For example, Pettinato and colleagues [28] used label-free, spectroscopic microscopy to track changes in chromatin structure and biochemical composition in live organoids and real-time track differentiation processes taking place within organoids as the cells alter their shape. The authors also demonstrated that this approach can be useful to monitor changes in chromatin architecture and cellular biochemistry that occur when cells differentiate along distinct lineages. In cases where labelled imaging approaches can lead to artefacts or perturbations, label-free imaging offers a noninvasive way to reveal how organoids develop and differentiate over extended periods, which could make it ideal for studies on the dynamic processes occurring in organoids.
The development of new imaging technologies is likely to result in further advancements in organoid research. Thus, the use of more powerful super-resolution techniques might allow researchers to take images of organoids at even higher resolution, potentially meaning that new details of subcellular structures, for example, or molecular complexes could be seen. In addition, new techniques that integrate real-time imaging with other analytical techniques such as transcriptomics and proteinomics will allow researchers to look at organoid development in the context of molecular and cellular networks. These techniques will further enhance our understanding of mechanisms of organoid biology while also serving as a useful resource for regenerative medicine, disease modelling and the development of more personalized medicines. By allowing us to study organoid formation and differentiation with unprecedented precision, this will, in turn, serve to provide more accurate models of human organs, facilitating better treatment regimens for a range of diseases.
Consequently, in the future, the ability to track organoid development and differentiation in real time will certainly benefit from improved imaging technologies. Many of these methods, including LSFM, single-molecule imaging, Fourier light-field imaging and super-resolution fluorescence microscopy, can be employed for organoid imaging, providing novel insights into the molecular and cellular information driving organoid morphogenesis. These techniques are not only allowing us to better understand organoid biology but are also facilitating studies on developmental processes and disease mechanisms, as well as potentially identifying new therapeutic targets and/or intervention strategies. It is likely that these imaging technologies will continue to evolve and improve, advancing the field of organoid research even more and leading us towards new and more effective treatments for disease.
Within the intricate 3D context of organoids, super-resolution microscopy techniques such as STED, PALM, and stochastic optical reconstruction microscopy (STORM) enable the observation of cellular interactions and their signaling pathways in detail [20,21,29]. These techniques allow exploration of subcellular compartments at the nanoscale resolution far surpassing the diffraction barrier of standard light microscopy.
STED microscopy represents one of the most crucial imaging technologies used, as it exploits a non-classical refinement of fluorescence microscopy, which can provide resolutions as fine as 20-30 nanometers. STED uses a combination of two lasers. The first laser excites the fluorescent molecules, while the second (depletion) laser, shaped like a donut with an empty center, depletes the fluorescence in the surrounding areas (Fig. 2). Only a small region in the center of the donut shape remains fluorescent, effectively shrinking the point of illumination. This allows for the imaging of details beyond the diffraction limit. This high resolution is essential for studying the spatial organization of molecules involved in signaling pathways and allows the researcher to explore the specific spatial positioning of proteins involved in the same cellular process. For example, due to the possibility of quenching the fluorescence signal selectively around the localization of a depletion laser, STED microscopy can be used to image specifically one of the components within the organoid, greatly facilitating the interpretation of images. An example is given in a study by Fang et al. [30], in which STED microscopy was used to monitor how lysosomes communicate with mitochondria in live cells and organoids. Lysosomes are involved in the management of waste on a cellular level and are pivotal in a range of signaling pathways, including those affiliated with apoptosis and metabolism. In this study, the signal resulting from an organic near-infrared small molecule fluorescence probe (that can be used for super-resolution imaging) dynamically followed the lysosome-mitochondria interaction, providing researchers with an approach to follow how these two organelles co-ordinate in time, thereby providing insight into how cellular homeostasis is maintained or how cells respond to stress.
Figure 2. Super-resolution techniques. The principles of how various super-resolution microscopy techniques improve the spatial resolution. (A) In conventional fluorescence microscopy, all fluorophores are excited at once, thus the diffraction-limited areas from closely positioned fluorophores overlap. (B) STED is a patterned illumination approach and based on the reduction of the effective excitation region. The sample is scanned by a tightly focused circular excitation beam. A second, overlapping doughnut-shaped laser beam de-excites all fluorophores except for those in the center. (C) PALM/STORM is a localization-based approach which uses photo-controllable fluorophores to observe single fluorophores at any point in time and localized with high precision and then fit by a Gaussian to obtain a super-resolved image. STED, stimulated emission depletion; PALM, photoactivated localization microscopy; STORM, stochastic optical reconstruction microscopy.
In addition to STED, other localization-based super-resolution techniques such as PALM and STORM. Uses fluorescent proteins that can be photo-activated and switched on and off. A small subset of molecules is activated, and their positions are recorded with precision. After these molecules photobleach, another subset is activated, and the process repeats. By reversibly switching off and on fluorescent molecules attached to specific proteins in a signal, PALM and STORM allow scientists to map the 3D localization of thousands of individual molecules, thus revealing the distribution and the dynamics of many different proteins simultaneously—and at a much higher resolution than previously possible (Fig. 2). Using PALM, it is now possible to map signaling protein organizations in the complex scaffold of organoids. In a study published in 2021, Fang et al. [31] used PALM as an image-acquisition tool to study the organization of signaling proteins within the stem-cell-derived organoids, with a focus on the Notch signaling pathway. One of the most important pathways in development, Notch controls cell proliferation and cell death. The study revealed the Notch signaling receptors’ and their associated ligands’ nanoscale clustering within organoids, thus providing new insights into how the signaling protein molecules are organized and what role they play in regulating stem cell behavior and tissue formation—essentially, how organoids grow. Visualizing signaling protein organizations in organoids via PALM also highlights the importance of this imaging technique in mapping molecular architecture of signaling pathways within 3D cellular scaffolds.
Another localization-based super-resolution technique, called STORM, has similarly been used to study the dynamics of signaling complexes in organoids. Like PALM, STORM achieves super-resolution by the stochastic activation and subsequent localization of individual fluorophores, but the signal for free fluorophore emission comes from the random opening of a closure at the fluorophore. STORM has been widely used to study the precise localization and interaction dynamics of signaling proteins that are involved in key developmental pathways. Using STORM, Keshara and colleagues [32] were able to study the Wnt signaling pathway in organoids, which controls cell proliferation, differentiation and migration during development. Their map of the Wnt receptor distribution and between Wnt receptor and its downstream effectors provided the first view of how signal transduction is organized within cellular membranes of organoids. Such high-resolution insights into protein interactions allow us to unravel the complicated signaling networks involved during organoid development and differentiation and gain insights into how these processes are orchestrated at the molecular level [32].
Super-resolution microscopy is playing an important role in the study of organoids, shedding light on the crucial mechanisms that underlie their development, disease modelling and regenerative medicine. Specifically, these techniques allow us to visualize molecular interactions and cell signaling, which are essential for understanding how these cells assemble into functioning organoids. For example, Ormel et al. [33] used super-resolution microscopy to study the way in which neurons form synaptic connections in brain organoids, providing new insights into how these cells form functional networks with each other. This is crucial for our understanding of various neurodevelopmental disorders such as autism and schizophrenia. Moreover, these models can be used to engineer brain organoids that more faithfully represent human brain function for disease modelling. Super-resolution microscopy is also important for the study of aggregates and other features in brain organoids and other neural tissue models [33].
High-throughput screening (HTS) and high-content screening (HCS) are two essential aspects of drug discovery these days. These technologies, when applied to organoid models, enable the generation of large and statistically robust pharmacological data to help us identify drugs for various human disorders. The use of HTS and HCS technology in combination with patient-derived organoids have emerged as a powerful approach for the rapid and comprehensive analysis of vast compound libraries in a large number of diverse biological contexts (Fig. 3).
Figure 3. Comparison of high-throughput screening (HTS) and high-content screening (HCS) in the analysis of patient-derived organoids. HTS enables rapid, large-scale screening of compounds, providing broad quantitative data such as viability or growth reduction. In contrast, HCS captures detailed phenotypic changes, revealing alterations in cellular morphology, protein localization, and organoid architecture. This complementary approach highlights the trade-off between throughput and resolution in drug discovery and mechanistic studies using organoids. iPSC, induced pluripotent stem cell.
Walsh et al. [34] investigated the use of HTS in pancreatic cancer organoids. Pancreatic cancer is one of the most hard to treat and has the worst prognosis of any cancer. Current treatment options are limited and novel therapies to improve outcomes for patients are urgently needed. In this study, compound screening was performed using automated brightfield and fluorescence microscopy to assess cultured organoids. The imaging information was then used to analyses drug effects on organoid size, shape, and viability.
One of the study’s major contributions was the discovery of a drug cocktail that substantially reduced the viability of pancreatic cancer organoids. This was particularly impactful because pancreatic cancer has notoriously resisted many standard therapies. The drug cocktail was further validated in in vivo models, highlighting it as another strong candidate for pancreatic cancer treatment. One conclusion from this study is the strength of combining HTS with the organoid models, particularly in the identification of drug combinations that can overcome the resistance mechanisms that can make effective treatment of pancreatic cancer so challenging. The power of this approach of screening thousands of compounds in a much more relevant and complex model is a significant step in the fight against this devastating disease.
Czerniecki et al. [35] also applied HTS to organoids derived from a variety of cancers such as breast cancer. The authors tracked the growth of organoids over time, using a variety of imaging modalities, including live-cell imaging and endpoint assays. HTS was particularly important in this study because high-throughput confocal microscopy enabled real-time visualization of the impact of different drug treatments on organoid growth dynamics. With live-cell imaging, the timing of how drugs act on cells to alter growth can also be tracked. This provides information on the mechanisms of action and the likelihood of resistance pathways.
In fact, a large part of this work was identifying drug leads that specifically targeted the cancer stem cells within the organoids. Cancer stem cells are a subpopulation of tumor cells that are thought to drive tumor initiation, progression and relapse, and which are often resistant to conventional therapies, so the ability to identify drugs that specifically target them is a major step to harnessing these cells as a target for novel therapies. Such targeted therapies could lead to treatments that can kill tumors at their source, and reduce the risk of relapse and metastasis.
Gunasekara et al. [36] used this orientation with a focus on applying HTS to gastrointestinal (GI) organoids in order to understand the impact of a compound on the integrity and function of the GI barrier. To image structural and functional changes in the organoids induced by treatment with different drugs, the study combined various 3D imaging technologies such as OCT and multiphoton microscopy, which provided high-resolution, volumetric images of structural and functional changes. In this study, high-resolution imaging was critical because the barrier function of the GI is influenced by the underlying structural complexity of the GI epithelial tissue. 3D imaging of intact epithelial tissue was essential for understanding both structural and functional changes in the 3D epithelial tissue following drug treatment.
The finding that these compounds modulated GI organoid barrier function was particularly intriguing for researchers as many diseases like inflammatory bowel disease and colorectal cancer are characterized by a loss in GI barrier integrity. Therapeutically, identifying compounds that restore or enhance barrier function is thus very relevant. The advanced imaging technologies used in this study, provided for the first time a 3D view of how different barrier promote the growth of intestinal cells in the GI organoids. This platform has untapped potential for future studies on the genetic and molecular basis of gut health and disease, and holds the promise to revolutionize the drug-screening process.
In a study by Brandenberg et al. [37], the researchers employed confocal to visualize specific cellular processes like proliferation, apoptosis and differentiation in response to chemotherapeutic agents. Because confocal microscopy allows for the visualization of dynamic cellular events, these types of processes would be difficult, if not impossible to observe at a single depth in 2D cultures. Whole-mount staining of an organoid from the LGR5-positive progenitor cells differentiated into an intestinal organoid model was used to quantify chemotherapy-induced toxicity and response. The 3D rendering shows cell death appearing in the differentiated villus structure. The authors found that organoids were better at predicting patient-specific responses, which is a clear advantage compared with more traditional models. In the era of precision medicine, where treating patients based on their own tumor biology has proven to lead to better clinical outcomes, this technique could have a big impact.
Beyond providing a new human cancer model, the study’s use of HTS showed how organoid models might be used to streamline the discovery of effective cancer drugs. Given a sufficiently large library, screening can be used to discover combinations of compounds and doses that will kill tumor cell populations while leaving healthy ones untouched. In theory, this approach will lead to more effective drugs with fewer side-effects, as well as increasing the speed of the drug discovery process. The ability to screen and profile the impact of thousands of compounds on complex organoid models represents yet another leap ahead for a field of research that is, arguably, the most important in medical science today.
Choo et al. [38] also used HTS to determine whether this new drug candidate was capable of killing colorectal cancer organoids. Using an automated fluorescence microscopy that allows the simultaneous quantification of several cellular parameters—a well-suited high-content imaging technique to assess the effects of drugs that affect multiple cellular processes—the researchers quantitatively distinguished the different mechanisms of action and time points of drug-induced changes in organoid morphology and function. For example, colorectal organoids can be quantified by their branching-tree-like formations and the number of lumens (the central cavity of the cells), and the number of cells per structure is an additional factor in characterizing their shape. Multi-parametric analysis revealed changes in the number of lumens, the surface of the lumens, and the number of cells per lumen, indicating the degree to which the organoids underwent cell death, differentiation and were structurally intact. The ability to quantitatively distinguish between the different mechanisms of action and time points of drug-induced changes in organoid morphology and function is essential for understanding the more complex cellular responses in heterogeneous tumors such as colorectal cancer.
Most notably, given the heterogeneity of colorectal cancer, the researchers were able to find specific drug combinations that led to apoptosis in a subset of the colorectal cancer organoids. The development of therapies that target the right subpopulation can obviously be a significant step forward. What’s also important about this study is that the team used HCS to demonstrate that therapeutic effects were not manifested by a single cellular parameter but by a combination of parameters. It’s not at all uncommon for a drug that has an inhibitory effect on a single measure in a single-endpoint assay, such as 3-(4,5-dimethylthiazol-2-yl)-5-(3-carboxymethoxyphenyl)-2-(4-sulfophenyl)-2H-tetrazolium measurement of metabolic activity or the rate of cell growth, to do so in a way that produces a diverse set of effects that, when taken together, are fundamentally incompatible with cell vitality [39]. Traditional single-endpoint assays have a limited ability to capture such complexity because the single endpoint is often a misleading readout when considered in isolation. Hence, it’s important to ascertain the proper measure. When they do, they reveal drug efficacy in an entirely novel light.
In a recent study, Li et al. [40] catalogues the HCS techniques used to study liver organoids, emphasizing their application to drug metabolism and toxicity. By combining fluorescence imaging with mass spectrometry-based imaging, their study assessed cellular and molecular-level injury upon drug treatment. Specifically, through fluorescence imaging, changes in cell viability, cell morphology, and apoptosis were evaluated in real-time. Mass spectrometry-based imaging provided direct information on the metabolic changes driven by drugs, paving the way for identification of the mechanisms of drug-induced toxicity.
Equally noteworthy, the study led to the discovery of biomarkers for drug-induced liver injury—a common problem during drug development and a major reason for the failure of promising drug candidates. The ability to predict this type of toxicity using liver organoids would be a major advancement in the field since it could dramatically reduce the incidence of adverse drug reactions in phase I trials. The ability to better identify drug toxicity using HCS is an excellent example of the potential of this technique to provide more nuanced readouts of drug toxicity, which is critical for the development of safer pharmaceuticals. The second example from the same paper is the multi-modal imaging that was used—both fluorescence and mass spectrometry-based imaging. This study underscores the added value of using different imaging modalities to enhance the understanding of drug effects in organoids. While single-modality imaging techniques have proven valuable for basic cell and organoid biology, multi-modal imaging can provide a better, more comprehensive understanding of pharmacological actions in a dish.
HTS of organoids is finally facilitating the identification of new therapeutic agents and advancing our knowledge of diseases in fields such as cancer drug discovery. For example, HTS using organoids derived from patient tumors has sped up the discovery of personalized therapies against cancers. A striking example of this is a study involving colorectal cancer organoids that used HTS to identify drug combinations that could overcome chemotherapy resistance in clinical practice and provide patients with personalized cancer therapy by developing precision medicine strategies tailored to individual tumors. In the context of cystic fibrosis (CF) therapy development, HTS of CF organoids can screen hundreds of small molecules identified by high-throughput compound screening to improve drug discovery. VX-770 (Ivacaftor; Vertex Pharmaceuticals), a pioneering drug developed in 2012 for CF, works on the CFTR protein, which is defective in CF patients. To discover VX-770, researchers were able to focus on components specific to the disease, using HTS of organoids derived from CF patients. VX-770 is now under clinical assessment [41]. The application of organoids in drug testing enabled more efficient testing of drug responses in the context of patients, and VX-770 was approved for clinical use around 10 years earlier than most drugs of this novel class, which is a major step forward in precision medicine for CF.
Besides its use in drug discovery, HTS of organoids has also played a crucial role in virology. In the case of the Zika virus, brain organoids were used to mimic fetal brain development and, using HTS, small molecules that protected neural cells from Zika virus infection were screened [42]. This approach not only provided an understanding of how Zika virus interrupts the development of the brain causing microcephaly, but also identified potential antiviral drugs that could prevent these effects.
HTS of organoids for neurodegenerative diseases has also advanced precision medicine, including important breakthroughs in Alzheimer’s disease. Researchers generated organoid models to reproduce the phenotypic characteristics of neurodegenerative diseases, and screened for those drugs that could reverse toxic protein aggregation, a hallmark of Alzheimer’s pathology [43]. This provided critical insights into disease mechanisms and provided promising new therapeutic targets for combating neurodegeneration. Taken together, these case studies exemplify the predictive nature of HTS in organoid systems, and the impact of this technology on drug discovery and precision medicine. Additionally, they highlight how organoids provide more physiologically relevant biological models for studying human disease, and hence HTS in organoid systems provides a transformative tool in biomedical research. The use of organoids has accelerated the development of personalized therapeutics by providing more physiologically relevant biological models for studying diseases.
Combined with the use of HTS and HCS platforms, organoid technology can provide novel opportunities in drug discovery to develop more efficacious and personalized therapeutics. Finally, a few sophisticated imaging technologies—including confocal microscopy, fluorescence imaging, OCT, and mass spectrometry-based imaging and sensing—have been used for visualizing the complex, spatially organized and densely packed drug responses in organoids—providing unprecedented insights into cellular and molecular mechanisms of drug responses, and improved sensitivity and specificity for drug screening approaches. The profound discoveries emerged from the studies using organoids could provide direction for precision medicine, where many drug targets and cancer drugs have shown little improvement due to the inadequate recapitulation of human physiology in existing models. Organoids have the potential to provide a more physiologic setting for drug screening approaches, which might be the central player in the next generation of drug discovery and development and could potentially lead to more effective therapeutics for patients.
The integration of artificial intelligence (AI) with imaging technologies has transformed the study of organoids, offering unprecedented insights into the complex biology of these 3D culture systems. The complexity of organoid systems, however, presents significant challenges in data analysis, particularly when it comes to interpreting the vast amounts of imaging data generated in HCS experiments. AI, particularly machine learning (ML) and deep learning (DL) techniques, has emerged as a critical tool in overcoming these challenges, enabling the efficient and accurate analysis of organoid imaging data and facilitating the translation of experimental findings into clinical insights. With these computational and modelling capabilities, it has been possible to share a threshold where research on AI-powered organoids became real, and the study of organoid biology became astronomical. For example, the spatial organization of cells within the 3D culture system remains one of the most challenging aspects to assess experimentally. This is where AI, including ML and DL, can really shine. These computational and modelling methods have been instrumental in overcoming the challenges related to the vast amount of imaging data collected in the context of HCS and how such information is being converted into actionable insights for mechanistic investigations of organoid biology.
AI has been applied to organoid research in a number of ways that powerfully serve the more general goal of improving the resolution, accuracy and thus interpretability of data generated. A particularly useful application takes advantage of the inherent parallelism of image data and combines AI and advanced imaging techniques such as lattice light sheet and volumetric imaging. Schöneberg’s [44] work on pyLattice, a combination of adaptive optics lattice light-sheet microscopy (AO-LLSM) and ML algorithms—to study organoids, especially human ESC (hESC)-derived organoids. AO-LLSM offered high-resolution, real-time, minimally phototoxic imaging of hESC-derived 3D organoids, enabling long-term tracking of dynamic processes such as cell migration, division and differentiation in complex, thick 3D structures. On top of this, the ML algorithms on the pyLattice platform allowed for automated segmentation and tracking of individual cells in organoids, which revealed that these complex 3D collective structures are formed by coordinated cell migrations and the assembly of distinct cell populations, all crucial for organoid development and previously impossible to discover with manual analysis.
In addition, it enabled the combination of AO-LLSM data with that acquired from fluorescence or confocal microscopy, allowing researchers to access cellular processes at work in the organoids. This was ideal for the study of clathrin-mediated endocytosis (CME) in hESC-derived organoids—a process central to cellular signaling and nutrient uptake. The researchers found that the CME process was fundamentally different at different stages of organoid maturation, demonstrating clear differences in patterns of clathrin-coated pit formation and vesicle trafficking that correlated with major developmental landmarks.
DL methods based on so-called convolutional neural networks (CNNs) have been successfully applied for segmentation and cell-type classification tasks. CNNs are especially built to treat complex image datasets and have thus been widely used for image segmentation, where the aim is to automatically define the borders between cells, and cell-type classification, where the goal is to associate true identity labels (morphological and functional properties) to these cells. For example, Gritti et al. [45] applied CNNs to high-resolution images of organoids to discriminate between different types of cells and accurately map their spatial organization. He was able to identify very subtle phenotypic changes that appeared in cells treated with drugs. CNNs have been used to automatically segment pancreatic organoids based on their respective locations within the 3D tissue. Applying CNN-based methods to organoid imaging data also allows for unbiased discovery of relevant cell characteristics, especially when they interact across multiple and interwoven scales of organization and complexity.
Other integral use cases of AI have been in the integration of multi-modal imaging data, which is another scenario where AI is highly beneficial. As mentioned, organoids are often studied using several different imaging techniques to obtain far more information about their structure and function than any one of these techniques alone: confocal microscopy yields 3D volumes, while techniques such as fluorescence imaging and super-resolution microscopy can be used to assess specific proteins or to image special structures at a higher resolution. Integrating these diverse data streams using AI’s algorithms has great potential to improve our analysis of organoids, including those based on DL. For instance, Beghin and colleagues [46] built an AI that combines data from confocal microscopy images and live-cell imaging to permit continued monitoring of organoid development and drug responses in real time. Integrating imaging data from multiple different modalities allows the analysis of organoids at higher resolution, as well as with far greater depth because it combines these diverse data streams of different types of information on the same 3D structure. This provides a more holistic view on how different regions and cell types of the organoid interact and respond to stimuli.
AI tools are not only used for analysis and interpretation but, increasingly, to model predictions. Now, predictive models are being employed to simulate potential therapeutic responses to various interventions before they are actually carried out. For example, models employing generative adversarial networks and variational autoencoders have been used to simulate future imaging outputs for organoid data. These models learn from existing imaging data captured across various conditions and can then be harnessed to predict organoid behavior under different treatment conditions. Though current applications of this technology focus on organoid images, one day it may be possible to apply this approach to simulate and predict organ physiology. As well as being applied to simulating in vitro data, AI-based predictive models can be harnessed to forecast the response of an individual patient to a treatment. In the future, it should be possible to construct models of personalized medicine in which AI receives an input in the form of imaging data from an organ considered to be diseased, and then predicts the patient-specific response to a given therapeutic intervention based on a trained model derived from the individual patient.
Furthermore, the AI program MOrgAna (ML-based Organoid Analysis) is an excellent example of automation of the image analysis of organoids. MOrgAna is based on the application of different ML methods, such as support vector machine and random forests, to automate the estimation of the number of cellular phenotypes as well as morphological features occurring in organoids. To deal specifically with the particularities of 3D organoid cultures, MOrgAna uses DL to automatically segment the organoids as well as classify the phenotypes. The results of studies such as Gritti et al. [45] have shown that MOrgAna can produce very reliable results while processing large datasets at an incredible speed, greatly facilitating the process of HTS to quickly identify the most promising drug candidates. AI-based analysis platforms, such as deep-learning-based systems, can perform automated segmentation of an imaging data to extract features that emphasize drug-induced phenotypic changes. This is especially important for drug discovery because the goal of screening compound libraries is to find drugs that produce a robust response within a complex organoid system. AI can help minimize the time and labor involved in HCS, increasing the sensitivity and specificity of the screening process.
The combination of photonics-based imaging and organoid technology is enhancing biomedical research, allowing for visualization of complex biological processes in high resolution and in real time within organoids. In the future, trending advances in imaging methods, such as LSFM, super-resolution microscopy, Fourier light-field imaging and new label-free methods, will provide a powerful tool to observe the dynamic events of cell and molecules in detail. Moreover, alongside the integration of HTS with organoid models, AI is already being applied to process the big data for drug discovery and the establishment of personalized medicine. The growth of organoid models in the future will lead to the improvement of imaging technologies and the integration of multi-modal approaches, which will increase our understanding of organoid biology and accelerate the development of better and more targeted therapeutics, solidifying organoid technology as a key technology in future disease modelling, regenerative medicine and personalized treatment.
None.
None.
All work was done by NTC.
No potential conflict of interest relevant to this article was reported.
None.
None.
CrossRef (0) Download (28)
Author ORCID Information Nathaniel Tarectecan Carpena |
 |
