Unraveling the Complexity of 3D Cell-Based Models

Marc Ferrer, PhD, talks to contributing editor Tanuja Koppal, PhD, about his recent work using 3D cell culture models. Dr. Ferrer discusses some of the innovations taking place and some of the applications in which these 3D cells are being used. He also addresses some of the existing challenges in using 3D cell-based models and what can be done to overcome them. Dr. Ferrer is leader, biomolecular screening and probe development, Division of Pre-Clinical Innovation, National Center for Advancing Translational Sciences, at the National Institutes of Health.

Q: Can you talk about how and when you started working with 3D cell culture models?

A: About five years ago we started working with tumor spheroids for screening to try and see if the pharmacological responses between 2D and 3D cellular models were different. We now know that the responses are different, but we are still trying to figure out how predictive the responses are that we obtain using 3D cells. The 3D cell-based assays are technically very challenging to do in a screening format. Today, with round bottom plates and hanging drop models, it’s become technically more feasible to generate these spheroids in microtiter plates for screening. We are trying to generate tumor spheroids with nutrient and hypoxic environmental gradients and with all the relevant cell-cell interactions, screening them, and doing all the pharmacological benchmarking using data from in vivo testing and clinical trials to see how predictive the data is compared with animal models and humans.

Q: How have the 3D cell models changed over the years, and are there still areas for improvement?

A: Right now, most of our work is still focused on culturing cells in conditions that generate spheroids resembling the tumor in terms of the types of cells present, and we are also trying to increase the complexity and biology of the 3D models. 3D models are very broad, as they could refer to 3D spheres made up of a single cell type, or they could be heterogeneous, for instance—tumor cells along with fibroblasts or immune cells. Organoids, on the other hand, are also 3D cells, but they include the microhistology of the tissue, which adds another layer of complexity to the 3D model. Organoids are usually cells taken from an animal model like a mouse or a rat, but these organoids include the histological architecture of the tissue of origin, such as pancreatic models and mini-brains. It’s hard to build organoid models by reconstituting human-derived primary cells or induced pluripotent stem (iPS) cells, but there is definitely a lot of interest in doing that. The next level of complexity is to use 3D bioprinting techniques to build architecturally well-defined tissues. With organoids, you put the cells together and hope for the best, whereas with 3D bioprinting, you direct the organization of the tissue by printing layers of cells.

During the past two to three years, in our laboratory, we have been working on the production of bioprinted tissue models for use in screening and drug discovery. We have four programs using tissue printing—including retina, skin, blood vessels, and cancer metastasis. The goal is to develop these 3D printed models so that the tissue of interest has the right histology and function. The main challenge after creating these tissue models is to develop an assay that will measure some disease biology in a high-throughput screening format. The common techniques for assay readout that we use with 2D cells, such as fluorescence microscopy, cannot easily be used for 3D models. With 2D cells, it’s easy to get the light penetration for imaging, but 3D tissues are too thick. We are trying to develop technologies and 3D segmentation algorithms to quantitate the biology in these models for use in a high-throughput format. So it’s not only about developing these tissue models, but also about generating the infrastructure to use these models for screening. We are using mostly human-derived primary or non-patient iPS cells to first generate our wild-type models, to make these 3D models as predictive and as close to human natural tissues as possible. Once that is done, we will create disease models using patient-derived iPS or primary cells. The other aspect of this 3D work is the development of organ-on-a-chip systems. With 3D cellular models, as we increase complexity, we decrease the throughput. 3D spheres offer high throughput, whereas organ-on-a-chip systems that have tissues embedded together offer all low-throughput options. Hence, our vision is to offer a continuum of various cellular models that increase in complexity to test out different compounds. They are all complementary models, and if we can find a way to integrate them into this new 3D drug discovery pipeline, it will hopefully be more predictive.

Q: What are the main limitations or challenges when working with 3D cells?

A: One of the challenges is that the benchmarks we are using to test the 3D tumor models are all based on data generated using 2D systems. So we don’t know how the cellular pathways and the microenvironment differ in these two systems when the compounds are tested. What needs to happen is a systematic, rigorous assessment of what each of these 3D models capture[s]. Do we have hypoxia in these models? How does the cell signaling change? How does the presence of other cell types in these models affect tumor signaling? This is all a work in progress. The good news is that from a technical standpoint we are now able to generate these models, and all we need is a systematic harmonization of all the models to figure out the biology captured with each model.

The reproducibility is very good with 3D spheres. With organoids, it’s a little bit more complex to figure out what really constitutes reproducibility, whether it’s size, thickness, an architectural feature, or something else that needs to be defined.

In terms of throughput, 3D spheres and organoids can now be generated fairly easily using 384 round bottom plates. We are trying to grow them in 1536 well plates but with miniaturization there are some design challenges. If the wells in the plate get too round this makes good cell aggregates, but it gets hard to image the cells. The beauty of 3D models is the ability to differentiate the effects on specific cell populations or cellular signaling pathways within the tumor spheroid, and that is lost when you can’t get good imaging data in 1536 well plates. With 3D bioprinting the throughput is 24 or 96 well plates, and with organ-on-a-chip systems the experiments can only be done one at a time.

Q: Where do you see the biggest impact of 3D cellular models in terms of their applications?

A: It’s not that we are going to stop using 2D cell-based assays any time soon, but the goal is to create a platform of 3D assays that we can integrate into our discovery pipeline for testing new compounds. We are also using compounds that we have tested in 2D models and testing them again in 3D models to see how they perform and if they remain active. At the same time, we are comparing these results with those obtained from animal testing. The goal is to generate a lot of data to find out how predictive these 3D models really are, and if the predictions translate well in vivo we can make better decisions for compound prioritization. Hopefully, this will reduce the need to do some of the initial testing in animals, which remains a huge bottleneck in terms of throughput.

The 3D assay predictions can help prioritize the compounds for animal testing, which will be a huge savings in terms of time, costs, and resources for drug development. The other application is to use these 3D models, such as liver and kidney models, for testing drug toxicity. For the toxicity studies, these models have been found to be quite predictive, but for efficacy testing we still don’t have enough data.