Drug Discovery
3D human microtissues for drug discovery, organ-on-a-chip solutions image
3D Cell Culture - Engineering Composable, Disease-Tunable 3D Human Microtissues for Drug Discovery

3D Cell Culture - Engineering Composable, Disease-Tunable 3D Human Microtissues for Drug Discovery

By Dr Patrick Guye, Dr Eva Thoma and Dr Olivier Frey
Winter 2018/19

In this article, we will share some of what we have learned over the past 10 years as pioneers in the 3D cell culture space and provide guidance for those who are just getting started with 3D models and end-point assays.

Life operates in three spatial dimensions (3D), as do the patients for whom we are developing therapies and cures. Earlier limitations in cell culture and analysis methods constrained us to 2D cell culture substrates, but the amount of valuable information we can extract from such systems is quite limited, and largely exhausted.

The human organism is composed of cells, tissues and organs with defined functions and a variable capacity for regeneration. Adult, mature cells perform the bulk of functions while regenerative capacity is mediated either by rare, specialised progenitor cells or adult, differentiated cells that can reactivate regenerative programmes. Primary, differentiated adult cells, isolated from donor material, are therefore the performance gold standard for building 3D cell culture models.

Even though not an inexhaustible source, cell stocks lasting months to years can be built up from a single donor for some organ systems as, for example, the liver. An alternative to working directly with adult cells is starting with stem or progenitor cells, such as human pluripotent stem cells-derived progenitors or Lgr5+ cells isolated from adult tissue. This permits for an in vitro expansion, at the cost of often questionable maturity as the degree of heterogeneity among microtissues increases as they grow.

Tissues are heterotypic in nature and composed of multiple different cell types arranged in 3D. While most cell types in solid organs are organ-specific (eg, liver hepatocytes, cholangiocytes and hepatic stellate cells), others are specialised versions of cell types also found in other organs, such as liver sinusoidal endothelial cells (LSECs) and Kupffer cells, the resident macrophages of the liver.

Organs are also permeated by many other immune cell types that patrol and, importantly, modulate inflammatory activation as well as tolerance and quiescence within tissues. These immune cell populations play a critical part in the development of many diseases and must be considered when establishing advanced disease models.

The isolation process of cells from solid organs is not without stress. In our experience, we have observed that tissues generated from adult cells require a few days of resting to downregulate stress as, for example, inflammation, extracellular matrix or adhesion markers. This brief resting period serves as a good indicator for the plasticity potential, which is also a hallmark of complex, organotypic tissues. Picture cells being ripped out of their in vivo environment and forced to adapt for their new in vitro environment.

While cells, in vivo, are embedded in complex, organ-specific extracellular matrices, conventional cell culture substrates, in comparison, are quite primitive, consisting only of flat, transparent plastic. The relocated cells must either secrete their own extracellular matrix and rebuild the tissue structure - which we propone - or one can provide artificial assistance through exogenous extracellular matrices, such as Matrigel or other, more advanced, formulations.

Equally important, blood needs to be replaced by a surrogate. Sophisticated medium compositions with defined supplements play a key role to support in vitro adaptation and further medium transfers may be needed to comply with in vitro endpoints.

Figure 1 Inducible fatty liver disease model

Human biology is complex. Effective models must be, too

Cell type composition determines the viability and limitations of in vitro models and usually stands in opposition to costs and quality controls (QCs). More complex models of multiple cell types cost more to create and validate, but they are usually worth the extra time and investment. For example, traditional hepatocyte toxicity assays rely on a simple, single-cell-type model of hepatocytes. A model composed only of hepatocytes, however, cannot detect inflammation-dependent toxicities.

Adding Kupffer cells to this model enables the detection of a wider range of inflammation-dependent toxicities and provides some further plasticity, such as tissue remodelling. The next level is to include liver sinusoidal endothelial cells (LSECs), which are important modulators of liver health and critical for both toxicity and inflammatory applications. As key regulators of tissue inflammation, homeostasis and repair, hepatic stellate cells should be incorporated in models used for any chronic inflammatory studies.

Validation, standardisation and translatability

3D models and their components, the cells, must be well validated to ensure cell-specific marker expression and functionality, both of which are prerequisites for robust model performance and assay readouts.

Using adult cell lots requires good QCs, welldefined pass/fail criteria parameters and often requires testing multiple lots. QC criteria may vary depending on the cell type and must ensure the qualified lot is fit for purpose for downstream applications when integrated in a tissue. This is especially challenging as cell functionality assessment in 2D does not always correlate with the same features in multicellular 3D models.

Critical parameters for hepatocytes include albumin secretion, inducibility and expression of a large panel of CYP genes, while for Kupffer cells the activation of inflammatory pathways upon exogenous inflammatory stimuli should be assessed. Hepatic stellate cells must fulfill stringent criteria for morphology, activation state and activation potential.

As with cell lots, the 3D model must perform according to well-defined QC metrics. Defined metrics are critical when changing models, whether to use a new cell lot, change the model cell composition or introduce new culture conditions.

It is crucial to establish a harmonised model definition. Such a definition should clearly state the minimum criteria required to answer defined scientific questions by specific experimental approaches. For example, high throughput screening approaches require a model with high scalability, while physiological cell composition and interaction are more relevant for mechanistic studies. In most situations, a model is intended for several applications requiring a compromise between model features and experimental feasibility. Knowing the limitations of your model is as important as the performance metrics.

3D Assay Guidelines and Considerations

‘Thinking in 3D’ is key to making the transition from monolayers to 3D tissues. Most in vitro bio-chemical cell-based assays currently in use were designed with monolayers or single cell suspension in mind. In 3D, many of the cells are not directly exposed to the cell culture medium and therefore do not or only slowly contribute to the detection of soluble factors.

Tissue diffusion becomes an important factor not only for nutrients, oxygen and cellular products, but also for therapeutic antibodies or staining agents. For terminal endpoints, such as nucleic acid and protein extractions or fixing the tissues for immunolabelling, the stability of the 3D tissue versus simple 2D monolayers must be taken into consideration when dissociating the material. Achieving a good and complete lysis of in vitro 3D tissues often requires the harsher conditions required for their in vivo counterparts.

This becomes especially challenging when single cell integrity needs to be preserved. Thus, single-cell isolations from tissues, regardless if in vivo or in vitro, necessitates specialised protocols. As the community of singlecell analytics users is growing rapidly, there is also an increasing number of protocols to perform such isolations.

Genetically modifying cells, transiently or stably, works most efficiently when applied to single cells. For 3D tissue models composed of adult cells, transfection or transduction is usually performed upstream (pre-aggregation) of tissue generation. Once aggregated, formed and matured, outermost cell layers or vascularised regions are usually still accessible to transduction or transfection methods.

Accessing cells buried deeper within the tissue is more difficult. This leads to a significant bias on which cells are being modified. With organoid models, having genetically-stable modified precursor cells enables the growth of identically-modified progenitor cells, albeit epigenetic silencing can still be a challenge for transgenes containing non-constitutive promoters.

Technologies for Fully Unlocking 3D

Now, let us consider a hypothetical gold standard: the ability to extract multi-omics information from a single 3D tissue cell with high spatial and temporal resolution. This would enable a researcher to understand, at the single cell level, what is happening within the tissue over time through multiple omics endpoints. While 3D cell-based technologies are evolving quickly, we are not quite there as all existing solutions have specific constraints. Do not be discouraged by that – current technologies can provide a wealth of meaningful data, so much so the bottleneck is usually the analysis and interpretation, rather than the amount of data.

Transcriptomics are routinely performed with 3D tissues. Ensuring a good lysis of the tissue is important, and methods or kits used for lysing tissues from biopsies are often best suited as a starting point. When working with smaller microtissues as, for example spheroids encompassing a few thousand cells, it is also possible to lyse these tissues directly in the cell culture well, using the lysate for downstream NGS library generation, bypassing classical RNA extraction.

3D Volume Imaging

Just as pathologists use thin tissue sections as clinical endpoints, the same is routinely used for in vitro 3D models. Classical histochemical stains, even though they can involve some processing, circumvent the challenge of tissue thickness and provide important information about the structure and composition of the tissue at a specific point in time. The trade-off (and an inherent limitation) is that it looks only at slices of the whole tissue model. If the tissues exhibit significant heterogeneity, then laborious acquisition, staining and analysis of many slices will be required.

Whole mount stainings, facilitated by the advent of optimised clearing reagents, offer significantly more flexibility and information, with the disadvantage of requiring upfront optimisations (clearing, antibody penetration in 3D) and more advanced microscopic technologies and solutions for data storage, processing and analysis. Especially for larger 3D models (>100um in diameter), imaging becomes a major challenge due to light scattering of the tissue. Advanced imaging technologies such as two-photon microscopy can solve this problem but often go with the cost of reduced throughput.

4D Live Imaging

While following cellular events in 3D over time is conceptually the top-tier endpoint, current challenges, limitations and required investments do not permit routine use of this technology. Lightsheet microscopy, with its fast acquisition and short exposure to strong light, is emerging as an approach to track 3D tissues or even small organisms over time. For doing so, one needs to have cells labelled with fluorescent markers compatible with little or no impact on their viability over long periods of time.

These are usually fluorescent proteins expressed by transduced cells within the tissue, or live chemical stains, permitting to distinguish single cells or even organelles. Current systems require laborious sample preparation and can only handle a few specimens in parallel due to tight space constrains to ensure high optical image quality.

These examples illustrate that culturing substrates for 3D tissue models not only needs to be adapted to preserve optimally cellular and tissue functions, but also allow accessibility to gain unparalleled information. This includes access to the supernatant, the cells themselves for downstream analysis and optical transparency for imaging. One need to consider scalability and compatibility with standard assays.

Patient-focused Disease Modelling in 3D

Central to drug discovery and development is the patient. When a patient has a condition that diverges from a pre-defined healthy state, the goal is to prevent further disease development or, even better, bring them back to the healthy state. Usually, some physiological functions are impaired, impacted or deficient in a way that significantly lessens quality of life. Thus, just having healthy, 3D tissues is of limited usefulness. One must also have the ability, knowledge and protocols for inducing specific disease states in 3D tissues within a reasonable window of time, to be able to rigorously test therapies for preventing or reversing disease states.

It is important to stress that in vitro disease modelling must be achieved in a reasonable time frame, as having to wait years to induce a disease is simply not acceptable. For example, it is thought that the average natural progression of each stage of nonalcoholic steatohepatitis (NASH) takes 7.7 years. Nobody wants to maintain their cell culture model over decades, so we need to accelerate this process significantly. Nevertheless, speeding the induction of disease states can be tied to compromises.

A diseased 3D model needs to be carefully validated to ensure it sufficiently recapitulates key pathological aspects of the disease and, importantly, responds to disease treatments in this accelerated situation. One of the newest developments in the field of liver disease is a screenable, in vitro 3D human liver disease model that faithfully recapitulates induction and progression of NASH. This model was bioengineered to include all primary liver cell types involved in NASH development. As in its in vivo counterpart, the progression of disease is induced over multiple steps, albeit condensed down to two weeks.

Sufficient translational correspondence is based on clinical biomarkers such as cytokines or collagen fibrils deposition. Importantly, this model responds to treatments of anti-NASH candidate drugs in clinical development. This ‘fit for purpose’ proof of principle is important for any 3D model intended for drug development and testing. Thus, testing larger panels of compounds from clinical stages that were successful or failed is a crucial step in determining if a model is robust and ready to translate to substantial savings in drug discovery and development.

Rules of Composition for Higher Order Systems

Modularity and the integration of functional parts into higher order systems is not only a fundamental ingredient for industrial success, it is also a basic design principle of biology. Our body is made of organs which are made up by tissues, which in turn are made up by cells. With the advent of 3D tissue models, we moved from cells to tissues. To get closer to the human patient, it can be necessary to go beyond tissues and explore organs and connected organ systems. Microphysiological Systems (MPS) or Body-on-a-Chip (BOC) systems provide platforms to connect tissue models in fluidic channels.

Figure 2 Higher order organ-on-a-chip solutions employ precisely-engineered microfluidics technology

Until recently, bioengineering of 3D tissues focused on cellular composition and structure. In higher order systems, the environment in which the tissues are cultured, becomes much more important. It has been widely recognised that physiological flow influencing mass transport and shear forces, as well as physio-mechanical cues, can have a critical impact on tissue function and drug-tissue interaction. Beyond the local tissue environment, organ function is decisively determined by interactions with other organs in the human body through endocrine signalling. Engineering multi-organ systems expands the challenges of model development to include micro-engineering, microfluidics and logistics.

Microfluidic systems can substantially increase experimental complexity not only by a more sophisticated set-up including tubing, pumps and microfluidic labware, but also through the requirement of intricate set-up protocols. A complete and reliable organ system can be generated only if it can be ensured that every element of the system (ie every single organ model as well as every technical part) functions properly and produces trustworthy results. Furthermore, modularity, simplicity and the introduction of QC steps on the biological and technical level are key for establishing multi-tissue experiments with high reproducibility and acceptable throughput.

As described above, the production and maturation of organ models after in vivo extraction needs to follow stringent protocols to preserve maximal viability and organ-specific functions. Protocols may vary in time and in most cases are incompatible with each other requiring isolated production and subsequent assembly. This poses challenges in timing, logistics and on-site, gentle multi-organ assembly and operation, for which we are developing novel, modular and robust solutions.

Despite these hurdles, higher organ systems open new doors in pre-clinical testing. While in the past, studying and modelling organ-organ interaction was only available in animal models or human trials, MPS enables transfer of systemic insights to the pre-clinical in vitro stage. More interesting is that, in contrast to animals, organ-organ interaction can be systematically built-up or dissected to elucidate on-target or off-target effects.


Recent technological and scientific developments now make it possible to build human in vitro 3D tissue models and even connected 3D tissue/organ models recapitulating human biology and pathophysiologies with a rapidly-increasing faithfulness.

While it is a continuous work in progress to improve these systems and make them even more reliable, the potential in risk reduction, increase in predictivity, time savings and cost savings are so significant that hardly any company in the drug discovery and development field can ignore these developments.

It is only a matter of time before we can truly engineer in vitro disease models that truly bring the bench closer to the bedside. When that happens, no one will want to be left behind. DDW


Dr Patrick Guye is Chief Scientific Officer at InSphero AG. Patrick has extensive experience in life sciences and bioengineering and has held leadership positions in both industry and academia. His research interests include cell therapy, 3D cellular models/organoids, synthetic biology, human stem cells, biologics/small molecules development and precision genomic engineering.

Dr Eva Thoma is Head of Liver Solutions at InSphero AG. A biomedical scientist with extensive experience in cell biology and the development of advanced cell models for drug discovery and development, Eva has held research management positions in pharmaceutical and biotech industries. Her team develops highly-sophisticated, inducible liver disease models.

Dr Olivier Frey is Head of Technology and Platforms at InSphero AG. Olivier is a bioengineering and microfluidics expert with a passion for microphysiological systems. His team is responsible for delivering the scalable 3D plate and flow technology that enables researchers to interrogate and interconnect advanced 3D models in organon- a-chip networks.