The future of disease modeling

Personalized medicine and the need for better models are the main driving forces behind the cur-rently rising interest in disease modeling. The numerous possibilities in disease modeling opened up by the invention of iPSC technology was mentioned by Takahashi et al. in the first hiPSC study [Takahashi et al., 2007]. Even if genetic alteration makes the implantation of hiPSC back into the patient too risky, the patients can benefit from drug discovery based on cell models of genetic dis-eases with the exact same genetic background as the living patient, combined with the patient’s medical records [Penttinen et al., 2015]. In addition, another trending approach is to correct the dis-ease-causing mutation with CRISPR-Cas9 gene editing technology and to compare how the patient-specific cells behave as a corrected, isogenic control cell line in the same conditions [Doudna, Char-pentier, 2014]. Finding drugs via this route is the main point of personalized medicine. As disease modeling is currently in a steadily advancing phase, with almost all somatic cell types being produced in vitro from the hiPSC, the reports of in vitro recapitulated disease-phenotypes are advancing at the same speed. [Robinton, Daley, 2012, Fang, Eglen, 2017] However, the next steps needed for major future breakthroughs are the transition from 2D culture on the well plate bottom into a biomimicking

105

3D culture and the transition from single cell studies to organ- and tissue-scale studies. These can better take into account the systemic effects and metabolic pathways because studying a single drug molecule on a single cell can only solve a very limited number of disease cases. [Langley et al., 2016, Fang, Eglen, 2017, Cossu et al., 2018, B. Zhang et al., 2018]

Studying the pathways and better controlling how the molecules of interest reach particular cells is the main point of the microfluidic devices used for organ-on-chip studies. When further combining multiple organs on the same chip, the system can be called body-on-chip or human-on-chip. [Chung et al., 2011, Huh et al., 2011, Nam et al., 2015, B. Zhang et al., 2018] For example, combining endothelial cell membrane with neurons and vasculature could be used to study the barrier proper-ties of the blood-brain-barrier. Likewise, combining hepatocytes via vasculature to other tissue-spe-cific cells, such as cardiomyocytes, could be used to study drug metabolism in the liver and the effect of the metabolites on the final application site. [Langley et al., 2016, B. Zhang et al., 2018] To produce these larger tissue blocks and organoids from single cells, ECM is needed [Shah, Singh, 2017].

There are 3D culture systems, such as the hanging drop method, which do not incorporate an exter-nal scaffold but instead rely only on the ECM produced by the cells [Fang, Eglen, 2017]. However, the scale reachable by these methods is limited because the diffusion of nutrients through a dense cell mass is often not sufficient. Thus, a more appealing method is to produce a tissue block with ingrown vasculature, all inside an ECM biomimicking hydrogel scaffold [Ikonen et al., 2013, Fang, Eglen, 2017]. In the ideal organ-on-chip systems, all these culture conditions are controlled and independently changeable by the researcher, so that in addition to just cellular interactions, we can also study, for example, the effect of oxygen concentration or temperature on cell functioning [Ka-tipparambil Rajan et al., 2018, B. Zhang et al., 2018].

Including the cell-cell and cell-ECM interactions into the models will also make them more complex to handle. Here, the rational design principles come into play again, i.e., the complexity should be controlled and too many unknown variables should not be added, which could be seen as one reason for the very slow translation of TE into clinical use [D. F. Williams, 2017]. Even more, the disease model systems should be designed so that the cells not only behave correctly, but that they can also be monitored in various ways throughout the culture period. As many of the current methods to study cells were originally developed with 2D culture in mind, the transitions of the methods simultaneously with the cell culture systems to 3D is on its own a challenging task. [Appel et al., 2013, Nam et al., 2015, Caicedo et al., 2017] For example, when comparing the microscopy methods used in Publi-cation I and PubliPubli-cation III, it is clear that OPT is a useful method for mesoscale imaging and larger volumes, while confocal microscopy functions very well when the distances are not more than roughly one hundred micrometers [L. E. Smith et al., 2010]. In addition to the rather static immuno-cytochemistry, the study of the functions of dynamic cells using electrophysiological methods is highly required from a successful disease model [Langley et al., 2016, Wallis et al., 2018]. Out of these, calcium imaging has been done the longest for 3D cultures [O'Connor et al., 2000]. Other

106

methods commonly used in 2D electrophysiological recordings include patch clamp, microelectrode array, traction force microscopy, nanoscale indentation or probing, and video microscopy, either combining some of these or using them separately [Laurila et al., 2016, Björk et al., 2017, C. Pra-japati et al., 2018]. Of these, we only used video recordings for 3D cardiomyocyte cultures in Publi-cation III, but the functional studies of 3D neuronal cultures are also coming in the future. Further-more, for disease models to be actually useful for the pharmaceutical industry, they need to be HTS compatible, as the numbers of drug molecules screened can easily be too high for manual screening [Nam et al., 2015, Mathur et al., 2015, Laurila et al., 2016, Kopljar et al., 2018, Mäkinen et al., 2018].

Once the complexity of combining 3D cell culture systems with microfluidic platforms and measure-ment methods is solved, yet another aspect to consider will be the maturity of the differentiated cells.

There is agreement in both the neural and cardiac fields that differentiated cells need to be in a sufficient maturation state before they can be useful in disease modeling. [Quadrato et al., 2016, Feric, Radisic, 2016, Tan, Ye, 2018] In addition, the cell characterization needs to be accurate enough to distinguish between various cell subtypes and to know what cells are being used [Quad-rato et al., 2016, Madonna et al., 2016]. Currently, the differentiated cells, such as the ones used in Publications I and III, are mature enough to have some spontaneous electrophysiological functions.

However, they still do not resemble adult cells, but rather a fetal phenotype, and not extensively screened for subtype [Ylä-Outinen et al., 2012, Vuorenpää et al., 2017]. As many aspects of the cells, such as morphology, electrophysiology, and metabolism, do not fully represent the mature patient cells, the discovered drug effects can also differ between cell studies and clinical studies [Quadrato et al., 2016, Feric, Radisic, 2016]. To solve this problem, multiple stimulation methods have been studied to speed up the cell maturation process, and one of the suggested options is 3D culture and its physical cues, such as mechanotransduction and morphology orientation along a topography [Quadrato et al., 2016, Feric, Radisic, 2016, Tan, Ye, 2018]. This is one of the main reasons why the transition of disease modeling into 3D is so important.

One further trend that is extending from material design to disease modeling is computational dis-ease modeling. Once the amount of data and variables becomes too large to handle manually, the bioinformatics and computational methods come in very handy. They can be used, for example, for the formulation of complex genetic and proteomic pathway analysis and for distinguishing the most important parameter changes. [Nam et al., 2015, A. S. Vasilevich et al., 2017] These methods have already been reported for both the analysis of hiPSC-derived neuronal network maturation [Lenk et al., 2016] as well as for the calcium handling of hiPSC-derived cardiomyocytes [Paci et al., 2018] in the 2D case. Computational modeling is also recognized as one of the key objectives of CiPA for the development of drugs without risks for cardiotoxicity [Wallis et al., 2018]. The obvious next step to advance these models is to apply 3D cell culture data and compare it with 2D culture.

107

The aim of this thesis was to develop and produce hydrogels for 3D cell culturing of neuronal and cardiac cells. Hydrogel biomaterials were developed based on the gellan gum polysaccharide, using both physical and chemical crosslinking strategies. The cy-tocompatibility of the hydrogel as well as of the crosslinking reaction was verified and functionalization with ECM proteins was deemed necessary for enhanced, positive cell response. The developed hydrogels are compatible with cell encapsulation in 3D inside the hydrogels, with the possible future aim towards disease modeling or clinical thera-peutic tissue engineering in soft tissue applications.

The main findings and conclusions of each Publication are listed briefly below:

Publication I:

1. The bioamines SPD and SPM are suitable and cytocompatible crosslinkers for GG.

2. The mechanical properties of bioamine crosslinked GG hydrogels are in the relevant range for brain tissue.

3. hPSC-derived neuronal cells survive the crosslinking process and encapsula-tion inside bioamine-GG hydrogel.

4. Laminin functionalization, even just via simple mixing, is required for the de-sired neuronal cell response, the neurite spreading. However, too high a lam-inin concentration is needed for the routine use of the material to be feasible.

Publication II:

1. OPT is both a suitable and a highly valuable method for studying hydrogels.

The transmission OPT of plain hydrogels can be used with Haralick’s textural features and MDA to distinguish between different hydrogel formulations,