Improving efficiency of human pluripotent stem cell differentiation platforms using an integrated experimental and computational approach

Abstract

Human pluripotent stem cells (hPSCs) have an unparalleled potential for tissue engineering applications including regenerative therapies and in vitro cell-based models for studying normal and diseased tissue morphogenesis, or drug and toxicological screens. While numerous hPSC differentiation methods have been developed to generate various somatic cell types, the potential of hPSC-based technologies is hinged on the ability to translate these established lab-scale differentiation systems to large-scale processes to meet the industrial and clinical demands for these somatic cell types. Here, we demonstrate a strategy for investigating the efficiency and scalability of hPSC differentiation platforms. Using two previously reported epithelial differentiation systems as models, we fit an ODE-based kinetic model to data representing dynamics of various cell subpopulations present in our culture. This fit was performed by estimating rate constants of each cell subpopulation's cell fate decisions (self-renewal, differentiation, death). Sensitivity analyses on predicted rate constants indicated which cell fate decisions had the greatest impact on overall epithelial cell yield in each differentiation process. In addition, we found that the final cell yield was limited by the self-renewal rate of either the progenitor state or the final differentiated state, depending on the differentiation protocol. Also, the relative impact of these cell fate decision rates was highly dependent on the maximum capacity of the cell culture system. Overall, we outline a novel approach for quantitative analysis of established laboratory-scale hPSC differentiation systems and this approach may ease development to produce large quantities of cells for tissue engineering applications.

Keywords: cell fate self-renewal; differentiation; expansion; model; parameter estimation; pluripotent stem cell; scale-up.

Figure 1

Establishment of cell states and compartmentalization of differentiation processes. A) Schematic representing the four cell states present during epithelial differentiation and the protein markers used to distinguish each cell state. Arrows indicate self-renewal or differentiation events among the various subpopulations. Rate constants representing either self-renewal or differentiation rates are also indicated. B) Schematic illustrating the compartmentalization of the differentiation process. Each phase is defined by a specific microenvironment and the primary differentiation or expansion event for each phase is indicated.

Figure 2

Analysis of hPSC expansion (Phase 0). A) Total cell number as a function of time in Phase 0. B) Flow cytometry data representing population breakdown as a function of time in terms of percentage cells expressing Nanog. C) Total apoptosis levels in culture as a function of time as measured by percentage of cells expressing the active form of caspase3. D) Model fit to data where data points represent the total number of Nanog+ H9 hESCs in culture as a function of time in the Phase 0 culture conditions. Line represents model fit to data by estimating the self-renewal rate of Nanog+ cell growth (k₀₀). k₀₀ value (in hr⁻¹) and quality of fit, indicated by the residual sum of squares (RSS) value, are denoted. Error bars represent standard deviation (N=3).

Figure 3

Analysis of Phase I of epithelial differentiation. A) Total cell number as a function of time in Phase I for Protocol 1 (left) and Protocol 2 (right). B) Flow cytometry data representing population breakdown as a function of time in terms of the percentage of cells that express exclusively Nanog (grey), express K18, but not Nanog (black), or express neither marker (white) for Protocol 1 (left) and Protocol 2 (right). C) Apoptosis levels in culture as a function of time as measured by the percentage of the total number of cells expressing caspase3 in Protocol 1 (left) or the percentage of K18+/Nanog− cells expressing caspase3 in Protocol 2 (right). D) Model fits to data where data points represent cell subpopulation dynamics in Phase I. The total number of either Nanog+, K18+/Nanog−, or K18−/Nanog− cells in culture are plotted as a function of time during Phase I of Protocol 1 (left) or Protocol 2 (right). Lines represent model fits to data by estimating the various self-renewal and differentiation rates pertinent to this differentiation phase for each protocol. For Protocol 2, a death rate of the K18+/Nanog− cell population (k₂₄) was incorporated into the model and estimated with the other cell fate decision rate constants. Values for each estimated rate constant (in hr⁻¹) in each system, as well as the RSS value, are denoted. Error bars indicate standard deviation (N=3).

Figure 4

Analysis of Phase II of epithelial differentiation. A) Total cell number as a function of time in Phase II for Protocol 1 (left), Protocol 2 (middle), and in Phase IIIa for Protocol 2 (right). B) Flow cytometry data representing population breakdown as a function of time in terms of the percentage of cells that express K18, but not Nanog or K14 for Phase II in Protocol 1 (left), Protocol 2 (middle), and Phase IIIa in Protocol 2 (right). C) Model fits to data where data points represent the total number of K18+/Nanog− cells in culture as a function of time during Phase II of either Protocol 1 (left) or Protocol 2 (middle). In addition, the number of K18+/Nanog− cells in Phase IIIa, an additional phase specific to Protocol 2, was analyzed (right). Lines represent model fits to data by estimating the self-renewal rates of the K18+/Nanog− cells in Phase II (k_55a) or Phase IIIa (k_55b). Values for the estimated rate constants (in hr⁻¹), as well as the RSS value, are denoted. Error bars indicate standard deviation (N=3).

Figure 5

Analysis of Phase III of epithelial differentiation. A) Total cell number as a function of time in Phase I for Protocol 1 (left) and Protocol 2 (right). B) Flow cytometry data representing population breakdown as a function of time in terms of the percentage of cells that express exclusively K18, but not Nanog or K14 (black), cells that express K14, but not K18 (dark grey), or cells that express none of these markers (white) for Protocol 1 (left) and Protocol 2 (right). C) Apoptosis levels in culture as a function of time as measured by the percentage of the total number of cells expressing caspase3 in Protocol 1 (left) or in Protocol 2 (right). D) Model fit to data where data points represent cell subpopulation dynamics in Phase III. The total number of either K18+/Nanog−, K14+/K18−, or K14−/K18− cells in culture are plotted as a function of time during Phase III of either Protocol 1 (left) or Protocol 2 (right). Lines represent model fits to data by estimating the various self-renewal and differentiation rates pertinent to this differentiation phase for each protocol. Values for each estimated rate constant (in hr⁻¹) in each system, as well as the RSS value, are denoted in the table. Error bars indicate standard deviation (N=3).

Figure 6

Sensitivity analysis of epithelial differentiation protocols. Sensitivity values were calculated to determine the impact of a 10% change in an individual parameter to the resulting K14+/K18− cell yield as predicted by the model. A comparison between the sensitivity values for each of the two protocols is shown in either A) the lab-scale system analyzed with a maximum capacity or B) a hypothetical unconstrained system lacking a maximum capacity assumption.

Figure 7

Effect of microenvironmental changes on estimated cell fate parameters and overall epithelial cell yield. A Synthemax substrate was used in lieu of a gelatin substrate and data was collected to show A) total cell number in Phase II (left) and Phase III (right) as a function of time. B) Flow cytometry data demonstrating high purity K18+/Nanog− cells during Phase II (left) and the population breakdown as a function of time in terms of the percentage of cells that express exclusively K18, but not Nanog or K14 (black), cells that express K14, but not K18 (dark grey), or cells that express none of these markers (white) during Phase III (right). Cell subpopulation dynamics data were collected and fit to corresponding models in Phase II (left) and Phase III (right) of Protocol 1. Estimated rate constant values (in hr⁻¹) in each phase are denoted as well as the RSS value for the model fit. Error bars indicate standard deviation (N=3).

Figure 8

Schematic representing an approach to investigating efficiency and scalability of laboratory-scale hPSC differentiation protocols. Gray boxes represent the strategy outlined throughout this study to bridge the gap between laboratory differentiation systems and bioprocess design and improvements.