Transcriptome-Guided Design of Physiological Multilineage Liver Organoids

Abstract

Despite several technical challenges, human induced pluripotent stem cell (hiPSC)-derived organoids enable biologically and clinically relevant functional study of physiology and disease. In a recent Cell Systems article, Velazquez et al. report a novel strategy to identify regulators of multilineage organoid maturation by reverse-engineering from the global transcriptome of human tissues.

The organoid model has been widely utilized as a tractable and convenient in vitro experimental system that constitutes a morphological architecture mimicking tissue type(s) of interest and offers the possibility to model multiple cell types simultaneously, as well as study physiology and disease mechanisms [1]. Use of hiPSCs for the generation of organoids is an attractive approach to ascertain clinical relevance of the model, in which the response to therapeutic intervention can be assessed as an ‘avatar’ that carries molecular features unique to each patient [2]. Despite such great promise, there are still several technical limitations to meet the expected goals. First, the modality to induce desired cell lineages is limited. Differentiation of source stem cells is typically induced biochemically by using an empirically determined cocktail of growth factors and cytokines/chemokines, or genetically by inducing expression of transcription factors and/or other regulatory factors. However, the former strategy is limited by an availability of recombinant proteins, and the latter is compromised by their spatiotemporal involvement, which is often difficult to model. Second, cellular differentiation in the organoids is often incomplete, and the cells still carry features of their progenitor state and/or other tissues. These limitations obscure the validity of organoids as a reliable experimental system to study disease pathogenesis.

To overcome these challenges, Velazquez, et al. took a sophisticated approach to develop a designer liver organoid (DesLO) by integrating transcriptome-based cell lineage modeling and genetic engineering [3]. In their prior work, multicell-lineage organoids were derived from fetal liver cells (fetal liver organoid; FeLO) by heterogeneously overexpressing a transcription factor, GATA6, although maturation of each cell type was incomplete [4]. To identify clues that induce further differentiation of the cells, tissue-specific gene regulatory networks (GRNs) were defined, based on a collection of transcriptome data sets from 97 studies, including profiles of 107 adult human livers. From a computational analysis, four transcription factors, including prospero-related homeobox 1 (PROX1) and activating transcription factor 5 (ATF5), were inferred as potential regulators of the liver-like GRN. As predicted, overexpression of these transcription factors, along with CRISPR-based transcriptional activation of a key hepatic enzyme, cytochrome P450 3A4 (CYP3A4), led to enhanced induction of molecular pathways involved in liver physiology, such as synthesis of coagulation factors, bile secretion, and lipid metabolism in the DesLO. The liver-like GRN was also enhanced, whereas GRNs of other tissue types were eliminated, supporting the acquisition of a global transcriptomic phenotype unique to adult liver tissue.

Interestingly, assay for transposase-accessible chromatin using sequencing (ATAC-Seq) revealed increased accessibility of promoter regions of genes involved in various liver physiologies, such as xenobiotic metabolism and bile acid synthesis. This suggests that the transcriptional regulators identified by the GRN analysis influence the regulation of chromatin remodeling to shape a liver-like global transcriptional program. Hepatocyte functions, such as albumin production, were substantially enhanced to a level comparable with that of primary human hepatocytes, making the DesLO outperform other models, such as hiPSC-derived hepatocytes (iHEPs). Furthermore, therapeutic targets, such as farnesoid X receptor (FXR) and its pharmacological response, were recapitulated, supporting its use to evaluate experimental therapeutic strategies.

Organoids have proven more effective over traditional cell culture systems for studying morphogenetic processes, although the level of resemblance to the real tissue is still limited. Vascular structure is one of the challenging architectures to model in organoids, which lack blood flow. Of note, the sparsely distributed clusters of endothelial cells in the FeLO were reorganized to form elongated lumen-like structures, which resembled hepatic sinusoids, in the DesLO. This is an appealing feature of this model, given the importance of sinusoidal endothelial cells in a range of pathophysio-logical processes in the liver, such as inflammation, fibrogenesis, carcinogenesis, and portal hypertension [5]. Immunostaining and single-cell RNA-Seq revealed that the DesLO comprises four groups of cells that resemble the major cell types in the liver, namely hepatocytes, cholangiocytes, stellate cells, and endothelial cells. When implanted in mice, the DesLO showed lasting high-level albumin production, even longer than that seen with primary human hepatocytes.

This study demonstrates a proof of concept for a new strategy in reverse-engineering multilineage human organoids with enhanced physiological resemblance to the target tissue. The use of a global transcriptome, together with gain-of-function genetic interventions, enabled an unbiased search of regulatory factors, which was not restricted by either prior knowledge or the availability of growth factors and/or cytokines/chemokines to induce maturation of organoids. Moreover, the global transcriptome-based modeling of tissue types (i.e., tissue-specific GRNs) allowed more reliable monitoring of the terminal differentiation of each cell lineage compared with traditional monitoring by expression of a few cell surface markers. The use of hiPSCs as the starting material may enable toxicology and/or drug response assessment according to genetic polymorphisms and/or environmental exposures unique to the donors. These advances represent a significant step toward establishing in vitro experimental models to study physiology and disease, with there being further room for technical refinement. For example, cell lineage-specific temporal gain/loss-of-function genetic intervention may optimize tissue maturation and maintenance. The sinusoid-like vascular structure may be further improved to reproduce microscopic structures, such as the space of Disse, for more precise modeling of interactions between parenchymal and nonparenchymal liver cells, possibly with incorporation of other techniques, such as 3D scaffolds, microfluidic devices and/or bioprinting [6–8].

Strategies to improve the chance of successful development will enhance the clinical utility of such organoids for individualized assessment of toxicology and therapeutic responses, given the generally low success rate of establishing patient-derived experimental models, such as patient-derived xenografts (PDXs) [9].

In conclusion, the study by Velazquez, et al. represents both a novel strategy to substantially improve biological and clinical relevance of organoid-based tissue modeling and a versatile platform with which to facilitate translational research in liver and other diseases.