Abstract
Cardiac pacemakers serve as a necessary device for millions of Americans suffering from sinoatrial node dysfunction. Frequent and unavoidable complications with electronic pacemakers have motivated research into biological pacemakers (bio-pacemakers) as a potential alternative for the medical device. Human induced pluripotent stem cells (hiPSCs) are a potential de-novo source for acquiring pacemaking cardiomyocytes, the cells necessary for the development of bio-pacemakers. Although protocols exist for differentiating hiPSCs into pacemaking cardiomyocytes, the protocols require refinement to increase yield. The proposed research aimed to delineate the connection between hiPSC morphology and potentiation of canonical Wnt/β-catenin signaling pathway, an essential regulator of cardiomyogenesis. Understanding this connection will enable improved modulation of the Wnt pathway during the cardiomyogenic differentiation process and ideally lead to the development of individualized hiPSC differentiation protocols contingent upon their morphology. Established protocols for directing hiPSC differentiation into pacemaking cardiomyocytes (PCMs) typically involve the biphasic manipulation of the Wnt/β-catenin signaling pathway—early-stage activation, followed by late-stage inhibition. These protocols can widely vary in terms of the duration of the Wnt/β-catenin pathway activation required to promote cardiomyogenesis across different hiPSC lines. This inconsistency may be attributed to the imposition of these deterministic protocols on the differentiation of hiPSCs that is a multistage, stochastic process of cell type transitions. We hypothesized that the differences in cardiac differentiation efficiency involve variations in hiPSC state and β-catenin expression and localization across different hiPSC lines and morphologies. This research investigated morphologically directed hiPSC cardiomyogenesis, emphasizing an individualized protocol for each batch by adapting the timing of the established protocol to align with cell state transitions as assessed by cell morphology.
We examined the role of hiPSC state, pluripotency marker expression, and β-catenin expression and localization within the different hiPSC populations to identify a morphologically adaptive approach to generate PCMs. Through use of the hiPSC line, β-catenin-mEGFP-hiPSC, that has been genetically engineered to express mEGFP-tagged β-catenin, alongside a transgene-free hiPSC line, we monitored the expression of pluripotency markers and β-catenin localization across various hiPSC morphologies. hiPSCs were induced to undergo the differentiation process using a variety of parameters, including variation in concentration and duration of exposure to a Wnt pathway activator (CHIR99021), a Wnt pathway inhibitor (IWR1), and small molecule Nodal inhibitor (SB-431542, also called SB). During the differentiation process, we used immunocytochemistry and confocal microscopy to measure the expression of pluripotency markers, mesodermal markers, and early cardiomyocyte markers. We also used immunostaining and fluorescence-activated cell sorting to analyze the downstream yield of PCMs produced by each differentiation treatment condition. This analysis revealed a consistent, reliable differentiation protocol for PCMs using 6 M CHIR for 24 hours to induce mesoderm induction, followed by 48 hours of 5 M IWR1 + 5 M SB promoting cardiac differentiation. However, due to the intrinsic variability in hiPSC state, this defined protocol cannot be blindly applied—hiPSC state must first be evaluated. In alignment with other studies, our data emphasizes the inconsistencies in hiPSC cardiac differentiation, while simultaneously acknowledging the ability to produce reliable protocols contingent upon the hiPSC state.
Finally, we also aimed to develop a system using artificial intelligence (AI) and machine learning (ML) tools to instruct morphologically adaptive differentiation protocols entirely based upon the hiPSC state prior to differentiation. To do so, we integrated our experimental data into an early-stage ML model capable of predicting the differentiation capabilities of hiPSCs under a defined protocol. This foundational model achieves a 63% success rate in predicting differentiation capability and serves as a vital foundation for the development of a more sophisticated ML model capable of guiding each dynamic step of the differentiation process.