How to build a tumor: An industry perspective

Highlights

• •The integration of complex 3D models can improve drug candidate selection and facilitate informed decision making from target identification until biomarker development.
• •The cell source has a significant impact on the translational relevance of the platform.
• •An increase in complexity usually comes with a decrease in scalability and robustness.
• •To drive adoption, standardization will be key and best achieved in the framework of PPPs.

Abstract

During the past 15 years, a plethora of innovative 3D in vitro systems has been developed. They offer the possibility of identifying crucial cellular and molecular contributors to the disease by permitting manipulation of each in isolation. However, improvements are needed particularly with respect to the predictivity and validity of those models. The major challenge now is to identify which assay and readout combination(s) best suits the current scientific question(s). A deep understanding of the different platforms along with their pros and cons is a prerequisite to make this decision. This review aims to give an overview of the most prominent systems with a focus on applications, translational relevance and adoption drivers from an industry perspective.

Technology evolution, translational gap in oncology and market forces

The drug development process specifically in oncology is a time- and cost-intensive process. Starting with target validation and lead selection an investigational new drug (IND) application can take between three to six years, and more than double the effort to apply for market approval. For every drug that is approved, up to 20 000 molecules could have been synthesized and iteratively tested in the earliest stages of drug discovery.¹ Only 4–6 % of compounds that enter Phase I clinical trials will reach the market and many more drugs meet their demise in Phase II studies when efficacy cannot be demonstrated.[2]^, [3] However, more-focused clinical trial design in combination with truly translational biomarkers including implementation of machine learning approaches has narrowed this gap in recent times.⁴ One of the factors contributing to the productivity gap is the lack of translation of preclinical models to patient pathology. This translational oncology dilemma is the driving force to use more-sophisticated methods to recapitulate human tumor biology in the drug development process. Primarily invented for tumor biology or precision medicine purposes, several platforms have found their way into the drug discovery programs of the pharmaceutical industry. The growing prevalence of these new platforms in the literature is being mirrored by their increasing acceptance by regulatory agencies. The five main technology drivers are: (i) improvements in cell culture tools from cell sourcing to reagent development⁵; (ii) the ability to recapitulate a disease by genetic modification including CRISPR/CAS9⁶; (iii) the increasing number of biomimetic materials and progress in microfabrication and microfluidics⁷; (iv) the availability of well-annotated, large patient datasets that lead to a better understanding of the molecular landscape of a disease⁸; and (v) the evolution of artificial intelligence and machine learning tools to interpret large datasets and develop new readouts.² These developments are being promoted by market forces driving the accelerated adoption of 3D models by the biomedical industry: (i) drug development strategies are shifting from broad to targeted treatment approaches; (ii) regulatory agencies and industry are working toward a decrease of animal use⁹; and (iii) an increase of precompetitive collaborative efforts among the pharmaceutical industry around model development and validation. These trends elevate the appeal of 3D cell technology systems offering the opportunity to build a human-relevant model with the advantages of in vitro repeatability, higher throughput and ease of manipulation.

Overview of advanced 3D in vitro platforms for drug screening in oncology

The triad of cell source, tissue architecture and technology provides the basis for the different applications in drug discovery (Fig. 1).

Figure 1.

Overview of advanced 3D in vitro platforms for drug screening in oncology. The triad of cell source, tissue architecture and technology provides the basis for the different applications in drug discovery. Abbreviations: iPSC, induced pluripotent stem cells.

Cell source

The cell source contributes significantly to the robustness and predictivity of 3D platforms. Immortalized cell lines are broadly used in in vitro assays mostly owing to their compelling advantages such as robust culturing features and straightforward accessibility. Usually, they are very well characterized and constantly quality controlled.[10]^, [11] The availability of phenotypic and molecular data enables the validation of targets across different disease types including the identification of potential new target-driven tumor dependencies. Prominent handicaps are their lack of heterogeneity and the non-physiological growth behavior; in particular, the selection of fast-growing cells. The latter has an impact on the regulation of different pathways which might influence their respective drug targets.¹² The missing heterogeneity relates to cells within one line as well as lines within one tumor type. Owing to the stringent selection process during the establishment phase, the composition of a cell line panel does not necessarily reflect the clinical reality.¹³ Induced pluripotent stem cells (iPSCs) and adult stem cells were first demonstrated to be useful for drug validation in several rare diseases that usually lack relevant animal models.¹⁴ Beside their increasing importance as a source for cell therapy, their high abundance and reproducibility enable drug screening approaches across a broad range of tumor types. The persistence of embryonic and/or fetal identity and some background genetic variability has led to a lack of standardization. To cover for this drawback an increased need for quality controls arises.¹⁵ Patient-derived xenografts (PDX) cannot only be used in an in vivo setting but can also serve as a valuable source for in vitro testing.¹⁶ Large PDX panels offer the advantage of combining a broad inter- and intra-tumoral heterogeneity with a large accompanying meta-dataset.[17]^, [18] The main pitfall in PDX is the chimeric nature of the tissue composed of human tumor cells and mouse noncancerous cells. Depending on the platform, this implies different steps of cell sorting or adaptions of the readouts.¹⁹ The direct use of patient material is the ultimate translational approach. Its use beyond precision medicine is mainly hampered by technical challenges like reduced viability owing to sampling procedures and restricted availability.²⁰ All those considerations are, to the same extent, applicable for the non-tumorous cells in the tumor tissue. Owing to the even lower scalability and availability of those cell types, it is essential to determine the necessary degree of genetic or phenotypic match between the different cellular components within one system.²¹.

Tissue architecture

Tumor spheroids are defined as cellular aggregates obtained in suspension or embedded within a 3D matrix. Based on cell origin and culture methods employed, several spheroid culture types can be distinguished: the multicellular tumor spheroid model using cell lines and non-adherent support; tumor spheres using cells obtained from solid tumor dissociation; tissue-derived tumor spheres using cells derived by partial dissociation of solid tumors; and, finally, organotypic multicellular spheroids which are generated without tissue dissociation. Multiple approaches producing spheroids are available but standardization across the industry is still lacking. Forced floating systems like ultra-low attachment plates or agarose underlayer,²² the hanging-drop method²³ or embedding in tissue-specific matrices²⁴ are some options. All those models aim to represent the in situ tumor more closely by growing it in 3D and by including multiple cell types.[25]^, [26] Thereby, these platforms enable drug screening for compounds affecting the non-malignant parts of the tumor as well as the tumor cells themselves. The increased heterogeneity when culturing tumor cells as spheroids enhances the translational value of the system but also the challenges to standardize screening efforts.²⁷ In most of the co-culture efforts it must be considered that the rebuilding of the tumor tissue is performed in an allogenic setting. Unless the platform is built on patient material this is an inevitable fact. However, the use of the latter comes with other drawbacks as described above.

Advances in stem cell biology have paved the way to grow tissues as 3D organotypic structures. Organoids are characterized by some specific key features including the ability to self-organize in 3D starting from adult stem or progenitor cells that differentiate into multiple cell lineages.²⁸ Patient-derived cancer organoids have been increasingly adopted in oncology because they cover a broad range of different tumor types; and they show partially promising predictive value when comparing drug response to the in vivo patient response.[24]^, [29] The possibility to create organoids from healthy as well as diseased tissue is another advantage of the technology. Although the conditions to grow organoids have been improved, there are still some obstacles such as low success rates for some organs, a limited standardization across different histologies and the need for sophisticated readouts.³⁰ In addition, the dependency on specific growth factors resulting in a pathway addiction affects the interpretation of drug screenings. Finally, one limitation in the tumor-derived organoid technology is that it applies mainly to cells of epithelial origin. Thus, stromal as well as immune components are underrepresented; and respective co-culture approaches are very limited.³¹ Besides the patent-restricted access to the technology, this is seen as a major obstacle for the broad implementation of the organoid technology in an industrial drug screening setting. Overall, organoids as well as spheroids are most used in a post-candidate setting on the premise that no microfluidics are applied. Ranging from a 96-well to a maximum of 1536-well format together with a broad range of readouts it is an ideal format to perform mechanistic studies and to screen for new indications or combinations.³².

Instead of assembling the tumor from its parts, tissue slice cultures use the intact tissue for ex vivo experiments.[33]^, [34] Developed in the early 20th century using rodent and human tumor tissue, the main focus of many studies was the prediction of an individual patient’s drug sensitivity.³⁵ The development of the tissue slicing technique enables generation of precision-cut tumor slices (PCTS) from mouse xenografts and human tumor tissues with defined thicknesses between 150 to 400 µm.³⁶ The PCTS maintain the 3D architecture and heterogeneity of the original tumors, as well as the TME comprising different cell types and the extracellular matrix (ECM). With the capability of producing PCTS in a more standardized manner, the application of this platform for the preclinical evaluation of new drugs has become more prominent.³⁷ There are a growing number of studies using different tumor types as well as compound classes including but not limited to immuno-oncology drugs, cell therapies and oncolytic viruses.[38]^, [39] A recent improvement toward industry application of the PCTS system is the addition of microfluidics.[40]^, [41] The applicability of PCTS for mid-to-high throughout drug testing is still in its infancy. Yet, besides the obvious benefits in the context of precision medicine, PCTS could gain more traction in the early phases of drug discovery when the focus is more on target expression and modulation as well as on biomarker discovery and validation.

At the interface between tissue architecture and technology there is the matrix component in every 3D platform. The increasing evidence that the dynamic reciprocity between the cell and the non-cellular part of the TME has an impact on the basic biology of a specific tissue has triggered the development of different types of hydrogels for cell culture use.[42]^, [43] Hydrogels, hydrophilic polymer networks of synthetic or biological origins, are unique because of their ability to mimic ECM while allowing soluble factors such as cytokines and growth factors to travel through tissue-like structures. The possibility of adjusting hydrogel stiffness by external stimuli such as pH, temperature, crosslinking, magnetic or light exposure allows the replication of diverging tumor tissues.⁴⁴ It is thereby possible to modulate migration, proliferation and drug sensitivity of the embedded cells. The synthetic hydrogels are superior to biological hydrogels with respect to reproducibility, scalability and tunability. However, beside technical challenges related to the polymerization process the evidence for predictivity toward clinical reality of those platforms remains to be provided.⁴⁵.

Technology

Currently, three types of bioprinters exist: droplet, extrusion and laser bioprinting. The former is the most used technique in the pharmaceutical industry owing to its applicability to higher throughput settings.[46]^, [47] When designing the bio-ink for a specific application many features such as biocompatibility, degradability, hydrophilicity or processability have to be taken into consideration.⁷ The outstanding advantage of any bioprinting approach is the possibility to settle multiple matrices, cells and nutrient factors in a highly controlled manner into a predefined architecture. Beyond 3D in vitro platforms there are possible applications (e.g., in regenerative medicine).⁶ Yet, a lack of scalability and robustness in most of the currently available systems hampers the broad application of 3D bioprinting.⁴⁷.

Another way of culturing 3D multicellular aggregates are agitation-based approaches like stirred suspension culture systems.⁴⁸ The addition of hydrodynamic forces adds a microfluidic component to the culture system. This enables the efficient local exchange of gases, nutrients and growth factors⁴⁹ but implies a high protocol specificity for the individual experimental setup.⁵⁰ Stirred bioreactor systems possess several characteristics that support their use for drug screening purposes: reproducibility, scalability, adaptability and feeding for high-throughput formats amenable to automation.³⁵ However, the phenotypic changes that the tumor cells undergo in those cultures still require more detailed analyses. Reports about epithelial–mesenchymal transition, increase in proliferation and resistance toward standard-of-care treatment are still fragmented and must be validated across different tumor types and platforms.[50]^, [51] Some of these concerns might be addressed by combining platforms. There are data suggesting the incorporation of hydrogel into stirred bioreactor systems can maintain appropriate tumor phenotype and response to therapy.⁵².

The advancement of microfluidic technologies led to the development of organ-on-a-chip systems that incorporate perfusion flow into cell culture systems, providing functions including oxygen and nutrient supply, fluid shear and gradient control that might be crucial to different aspects of cancer biology such as angiogenesis, metastasis and immune cell infiltration.⁵³ Applications range from toxicology to different disease models.[54]^, [55] Regarding oncology, the fluidic connection of different cell compartments mirrors closely the metastasis cascade, making this technology predestined to studying this complex biological process.⁵⁶ Current challenges for drug screening applications are the hydrophobic nature of polymer-based microstructures commonly used in those systems, defining the optimal media and flow conditions for multiple organs, and monitoring the system over extended periods. Although most of the reports around this technology are focused on aspects of tumor biology, the combination with in silico approaches might enable drug screening campaigns beyond oncology.⁵⁷.

The application of advanced readouts to deconvolute the efficacy of the tested compounds is another important piece of the puzzle. The demand for robustness and reproducibility is the same as for the platform itself.[58]^, [59] However, with the advent of new modalities the simplistic way of determination of cell death or viability is no longer sufficient. Live cell imaging, transcriptomic and proteomic analyses will be required to determine the most efficacious candidate.⁶⁰.

Applications of advanced 3D culture platforms to oncology drug discovery

Apart from the mutational landscape of a tumor, the cellular and non-cellular tumor microenvironments play essential parts in the response toward different treatments. Therefore, the analysis and modeling of the complexity of the microenvironment must be taken into consideration when identifying new drug candidates and early definition of predictive biomarkers. The implementation of 3D culture models reflecting the architecture and cellular composition of a tumor is crucial to future novel drug development strategies in oncology.

Those 3D culture models can be implemented at different stages of the drug discovery pipeline from target identification via efficacy testing through to biomarker identification and subsequently companion diagnostics development. Their use is mainly regulated by scalability and robustness. The ability to mimic specific aspects of tumor biology drives the application for certain modalities and targets. Lastly, technical prerequisites influence the adoption of the different platforms. As cancer immune modulation as a therapeutic concept becomes more important, there is a growing demand for complex 3D models that address questions around the crosstalk of the immune system and tumor cell. Beyond drug discovery, 3D models are becoming increasingly important in studies focusing on drug delivery, drug resistance and drug repositioning.[39]^, [41] Even in the highly regulated toxicology field, there are recent developments to use 3D platforms for non-GLP toxicology studies. Of note, outside of toxicology there is no need for implementing GLP standards on those platforms. A more detailed overview of the different applications per platform is given in Table 1 and Fig. 2.

Table 1.

Synopsis of 3D in vitro technologies and their applicability in oncology drug development.

Technology	Effort	Throughput	Reproducibility	General remarks	Application in the drug development process	Exemplary references (a selection only)
Spheroids	Low/medium	Medium	High	Biological relevance highly dependent on the cell source First reports on HTS applications available Applications outside oncology	Target identification Target validation Hit-to-lead	PMID: 30871626, PMID: 35306207, PMID: 35013156, PMID: 32897190, PMID: 31196138, PMID: 25247711, PMID: 33213473, PMID: 35551303
Organoids	Low/medium	Medium	Medium	Patent situation limits use and standardization efforts Highly specific culture conditions for each tumor type Often without cellular TME components Applications outside oncology	Target identification Target validation Hit-to-lead Identification of biomarkers	PMID: 31752287, PMID: 33749682, PMID: 34827570, PMID: 35187519, PMID: 29131160, PMID: 35425715, PMID: 34713317, PMID: 34663939
Precision cut slices	Medium	Low	Low	Multiple culture protocols & systems are available Slicing success rate varies between different tumors Influence of culture conditions on tumor slice culture is still poorly studied	Biomarker development PK/PD ADME	PMID: 25174503, PMID: 28670477, PMID: 26647838, PMID: 29233603 PMID: 23076396, PMID: 20404174 PMID: 21782910, PMID: 32175407 PMID: 31963500, PMID: 31741771
Bioprinting	Medium	Medium	High	3D microfabrication technique supporting other platforms Reproducibility better with synthetic than with natural sources Applications outside oncology	Target identification Target validation Hit-to-lead Non-GLP tox	PMID: 29935988, PMID: 30755456, PMID: 33774120, PMID: 29881724, PMID: 33741496, PMID: 34205767, PMID: 33572757, PMID: 35025130, PMID: 35171571, PMID: 34328185, PMID: 33440351, PMID: 32499560, PMID: 32297019
Bioreactor	Medium/high	Low	Medium	Main applications outside of oncology Numerous non-standardized systems available Complex systems to address crosstalk between various cell types	Biomarker development Target validation	PMID: 30912021, PMID: 32285369, PMID: 29476107, PMID: 32351161, PMID: 30912021, PMID: 29477032, PMID: 34761350
Organ-on-a-chip	High	Low	Low	Complex non-standardized system Validation data needed for specific applications mainly applied outside of oncology to model human (patho)physiology and address clinical questions	PK/PD ADME Non-GLP tox	PMID: 31726354, PMID: 22422217, PMID: 23636129, PMID: 20576885, PMID: 25524628, PMID: 31988459, PMID: 19865724, PMID: 35628214, PMID: 35694831, PMID: 35410642

Figure 2.

Applications of sophisticated in vitro 3D models in the discovery pipeline. 3D models can be applied at different stages of the drug development pipeline from target ID until biomarker development. Abbreviations: CADD, computer-aided drug discovery; CRISPR, clustered regularly interspaced short palindromic repeats; HCS, high-content screening; FBDD, fragment-based drug discovery; GLP, Good Laboratory Practice; PK, pharmacokinetics; PD, pharmacodynamics.

Translational relevance of advanced culture systems

As mentioned above, carcinomas are highly complex structures comprising genetically altered tumor cells, normal and cancer-associated fibroblasts, endothelial cells, pericytes and inflammatory cells embedded in the ECM. It is exactly this molecular heterogeneity that influences the way tumor cells migrate, proliferate and survive during tumor progression.[61]^, [62]^, [63] This dynamic evolving heterogeneity and its biological, pathological, clinical, diagnostic and, in the end, therapeutic implications have been not properly addressed in the past. A translational approach integrating gold-standard as well as innovative preclinical platforms with clinically relevant biomarker strategies is needed enabling a seamless transition into early clinical research.

It seems that in vivo models would be most appropriate to address TME modulatory questions. Several in vivo models such as PDX in humanized mice, transplantable syngeneic mouse tumor lines or genetically engineered mouse models do exist but they all have pros and cons. ⁶⁴ Mice and men differ substantially in many aspects of tumor biology especially in the regulation of the immune system. Thus, complex human 3D in vitro systems can be the models of choice for many translational aspects of the drug development process. The predictive value of individual preclinical platforms has yet to be determined and prospective clinical trials are currently under way (e.g., TUMOROID trial NL49002.031.14). Reports on successfully developed molecules in these platforms are still sparse. The cell source has a significant impact on the relevance of the model so has the ‘breadth and depth’ of the model collection. Access to primary patient material is indispensable as is the creation of large biobanks across tumor (sub)types. To define an exact threshold for the intended size of such a biobank is challenging because it depends on many different factors including but not limited to the nature of the test compound. In fact, the increase in significance follows a cumulative frequency curve. Finally, it is crucial to understand how heterogeneity is maintained in the different platforms. Studies like those performed on PDX or cell lines must be executed to understand the implication of culturing and passaging human tumor cells in 3D systems.⁶⁵ The direct correlation between translational relevance and microenvironmental characteristics like fluidics, pH or hypoxia has yet to be determined and it should be remembered that an increase in complexity usually comes with a decrease in scalability and robustness (Fig. 3). Nevertheless, the biology of these models places them between cell lines and PDX models. They maintain a broader cellular diversity and more closely represent the intratumor but also interpatient heterogeneity when compared with 2D screening platforms.

Figure 3.

Dependencies of scalability vs complexity in preclinical models of oncology. In most models, scalability and complexity display a negative correlation. The choice of the best suited model heavily relies on the scientific question and the need to reflect specific characteristics of the tumor in situ.

Taken together, novel 3D models unequivocally capture the in vivo situation better than standard 2D models and allow the validation and testing of novel therapeutic concepts. Over the past decade, complex 3D cancer models have started to be adopted for drug discovery purposes. This seems likely to increase in the future, given the current productivity gap and first promising candidates developed using 3D technologies.⁵⁸.

Factors driving adoption of complex 3D culture systems in drug discovery

In general, there is most interest for 3D systems in areas where predictive models are lacking completely or when new technologies provide clear translational or productivity benefits versus traditional 2D in vitro and/or in vivo models. This holds true for example for modeling the blood–brain barrier or immune–oncology strategies beyond T-cell modulation.

The integration of the 3R (refinement, replacement, reduction) principles¹⁰ into the drug discovery strategy in industry supports the incorporation of complex 3D models into the pipelines owing to their ability to mimic the in vivo situation very closely. The increasing reputation and influence of multiple 3R initiatives is another adoption driver for in vitro 3D models. A variety of governmental, non-governmental and industry consortia have supported the incorporation of 3D in vitro models into the drug development pipeline: the EU Reference Laboratory for alternatives to animal testing, EURL-ECVAM, the National Centre for the Replacement, Refinement and Reduction of Animals in Research and NC3Rs and the North American 3Rs Collaborative are just a few.

The new technologies must be translational in a sense that they should summarize the human disease and ideally capture the side effects as well as the efficacy of the drug candidates. Their deployment will improve preclinical datasets affecting decision making on early-stage molecules. A validation dataset proving equivalence in reproducibility, standardization and robustness is a necessity. Moreover, the demand on throughput depends on the development stage and ranges from a 24-well up to a 1536-well format. Finally, an indispensable feature of an innovative 3D drug screening platform is the applicability to all modalities including small molecules, biologics, cell and gene therapy.

Concluding remarks and future trends in complex culture systems

Although several gold-standard models exist in oncology, there are still gaps in the model landscape. Because individual studies and drug discovery in general are getting more complex, we envisage a toolkit of different platforms that can be combined in a way that best suits the scientific question and can cope with the increasing number of modalities including but not limited to cell therapy or cancer vaccines. Beyond target validation, efficacy and safety, future applications include long-term disease progression modeling and precision medicine offering early insights into responder populations. Furthermore, emerging goals of drug design strategies like epigenetic changes or modulating the tumor metabolome cannot yet be modeled with the available technologies.⁶¹.

For industrial purposes there are still some limitations to overcome. This includes standardization of protocols, upscaling of the respective culture types, miniaturization in multi-well formats including automated processes for establishment and testing and the stable integration of 3D bioprinting and fluidics into HTS. Some aspects have already been implemented mainly in personalized medicine.⁶⁶ As a first step in eliminating the above-described restrictions, intercommunicable standard protocols must be implemented, highly specialized media harmonized and commercially available, and culture and freezing protocols standardized. This will best be achieved in the framework of large public–private partnerships such as the IMI initiatives (https://www.imi.europa.eu/).

Finally, the key to success will be the optimal combination of new state-of-the-art platforms. This requires a deep understanding of the individual platforms and their output as well as data harmonization. Supported by the increasing use of artificial intelligence the latter is getting more and more awareness thereby increasing the quality of preclinical datasets in general.

Teaser: This review provides a synopsis of innovative 3D in vitro systems for oncology drug development with an emphasis on applications, translational relevance and adoption drivers from a drug development standpoint.