Human brain gene regulatory network inference using single-cell multi-omics

da | Giu 27, 2025 | Biologia Molecolare, Brain, Organoids, Organs

Figure 1- Inference of the gene regulatory network involved in human brain development and regionalization performed with the use of human brain organoids, single-cell multi-omics, and the Pando algorithm.

Abstract

Due to the complexity of the interactions between different genes, understanding the gene regulatory network behind human brain development is a major challenge. Fleck et al. [1] addressed this challenge by combining the use of brain organoids, single-cell genomics methods, and an in-house developed network inference algorithm called Pando, and investigated the role of region-specific transcription factors involved in shaping cell fate in the brain.

Review

 

Genes do not work independently: a gene expression level is determined by a complex interplay of interactions with other genes and small molecules that represent a gene regulatory network (GRN)[2]. The GRN involved in cell fate decision during human brain development has not been fully understood yet. Previous studies provided valuable insights about the GRN involved in animal brain development, revealing that brain regionalization starts from the formation of an anterior-posterior axis, caused by the activity of a series of transcription factors and mechanical stimuli [3]. Human brain gene regulatory network inference requires advanced techniques, allowing to analyze the effects of interactions between different cells into the  complex architecture of the brain tissue. Hence, the use of conventional cell cultures is suboptimal for this analysis and more complex models, such as brain organoids, are required. Moreover, traditional bulk omics techniques are limited  since they only provide an average of cell state across a population of cells. In recent times, the advent of loss-of-function gene editing techniques, coupled with single cell transcriptomics readout, enabled the analysis of gene regulatory programs during development at unprecedented resolution.

With this review, we want to underline the potential of innovative techniques – such as single-cell multi-omics and brain organoids – that, in combination, can greatly enhance reconstruction of the gene regulatory network controlling human brain development. We also intend to focus on the limitations of these techniques and suggest some future perspectives.

Discussion

The techniques that represent the strength of Fleck et. al. experiments mainly include:

  • The use of human brain organoids as a model system.
  • Multi-omics single-cell methods for the analysis of gene expression and chromatin accessibility.
  • An algorithm capable to infer the gene regulatory network created by the authors themself and named Pando.
  • The use of CROP-seq, a technique that combines CRISP-Cas9 gene editing and droplet-based single-cell RNA-seq.

With the use of brain organoids, we can easily reproduce small tissues environments and test the effects of different factors in a three-dimensional model. Wang et al. underlined the advantages that the use of patient-derived organoids can have to study neurodevelopmental diseases. Although, since these models generate a broad spectrum of cell types, coupling brain organoids with single-cell multi-omics can be a game changer [4].

Single-cell omics propose a way to better appreciate the cell heterogeneity and to improve the reconstruction of developmental trajectories, avoiding the exclusion of cellular subpopulations which could have a role in defining cell fate. Importantly, Fleck et. al. used metacells as a computational methodology to integrate multimodal data from gene expression and chromatin accessibility and adopted Uniform Approximation and Projection method (UMAP) for dimensionality reduction. These techniques allow to simplify the data and visually appreciate them. [5,6] For what concerns the Pando algorithm, it uses a regression model to obtain sets of positively or negatively regulated genes for each transcription factor, using the data obtained from multi-omics single cell methods (single-cell RNA-seq and ATAC-seq). By resolving the interaction of such factors, the method enable to infer which of them is active in a certain brain region and at a given developmental stage.

Pando has been released as a software package to be distributed in the community. Therefore, it could be avaluable resource to infer the gene regulatory network in future experiments. Finally, the effectiveness of CROP-seq derives from using a pooled library of gRNAs – each cell receiving a single perturbation – that targets different transcription factors at once, and therefore the obtained “mosaic organoids” present multiple perturbed phenotypes all together.

Fleck et. al exploited this technique to screen a pool of 20 transcription factors with key roles in determining cell fate and then further analyzed some of them with single-perturbation experiments. The combinatorial use of the above techniques allowed Fleck. et. al to confirm that the programs previously identified in mouse and other non-human model systems are well conserved in human brain development [7].

Among all the resulting factors involved in the development of the human brain, researchers particularly focused the analysis on 2 genes, HES1 and GLI3, whose mutations are associated with developmental disorders, revealing that they have opposite effects on dorsal telencephalon commitment. Further analysis disclosed that GLI3 is necessary for dorsal telencephalon development, since its absence results in dorsal telencephalon depletion, with a concomitant enrichment of the ventral telencephalon.

Results indicate that GLI3 regulates different genes in different developmental stages – for example, HES4 and HES5 – and binds specific regulatory regions near genes that play a key role in the development of telencephalic regions.

This explains why GLI3 knockout organoids fail to differentiate the telencephalic region at all.

Future Perspectives

The techniques we described so far represent a noteworthy resource. They can be combined and leveraged to uncover regulatory interactions and infer developmental trajectories, both in physiological and perturbed conditions. 

Indeed, the paper provides strong evidence that human brain organoids can be used as a “predictive” model system: an option could be investigating the impact of specific risk factors, allowing the assessment of the resulting epigenetic alterations.

Just imagine a study which aim is to discover the gene regulatory network that brings an individual to develop a certain disease: it would be possible to treat healthy organoids with a series of risk factors, analyze the transcriptomic and epigenomic profile with sequencing techniques such scRNA-seq and scATAC-seq, and then compare the data obtained with Pando. This would provide us an overview of the possible transcription factors involved in the occurrence of the disease. An example could be treating human brain organoids with poly- and perfluoroalkyl substances (PFAs), compounds at which humans are daily exposed, since they are present in a wide range of products such as non-stick cookware, water and stain repellent fabrics, and fire-fighting foam. Nowadays, these substances are widely investigated across multiple research projects, considering that exposure to them is involved in the development of immunotoxicity and metabolic disorders. Recent studies suggest that PFAs could be involved in neurodevelopmental disorders, but additional research is needed [8].

It would be also possible to design experiments that reproduce certain neuronal states or mimic neurodevelopmental disorders by developing organoids from induced pluripotent stem cells derived from patients. These organoids can then be analyzed with single-cell multi-omics to uncover their pattern of gene expression and chromatin accessibility. Finally, the obtained data can be unified with Pando to investigate how the GRN is perturbed in disease conditions and predict diseases evolution processes. This workflow can be applied not only to understand the changes in the gene expression that are characterizing a certain disease, but also to assess the effects of potential therapeutic drugs. For instance, in 2023, Chong Li et al. published a paper in which the use of brain organoids, combined with single-cell genomic methods, has been adopted to identify autism spectrum disorderassociated regulatory modules [9].

Also, we think that the cited techniques could have a positive impact in investigating the role of long noncoding RNAs (lncRNAs), since their function in human brain remains mostly unknown, even if they seem to be involved in differentiation and development [10].  Indeed, previous studies demonstrated that the loss of some lncRNAs can impair the progression of cell-cycle and drive apoptosis [11].

Limitations

Despite what discussed until now, the use of Pando to analyze multi-omics data presents some limitations. Indeed, the algorithm assigns candidate regulatory regions to genes at which transcription factors probably bind only on the base of their vicinity.  Even though the analysis with Cleavage Under Target and Tagmentation (CUT&Tag) found that 94% of the candidate regulatory regions found by Pando have accessible peaks intersecting with the acetylation of the lysin 27 on histone 3 (H3K27ac), this mechanism based on proximity of the genes could be risky and results in a loss of information.

Furthermore, a limitation on future experiments concerns the lack of information about comprehensive active and repressive histone modification and chromatin conformation status across the organoid development, as well as the lack of a complete transcription factor motif database available. 

Conclusions

Fleck et. al suggest a broad range of application for single-cell multi-omics profiling techniques in the context of brain development, and present the innovative algorithm Pando as a precious tool to integrate sequencing data from transcriptome and chromatin accessibility analysis to infer GRNs. We think that this work could be an important starting point for the development of models that predict developmental processes with the use of organoids, both for testing risk factors in healthy organoids and for mimicking diseases conditions. 

References

  1. Fleck, J.S., Jansen, S.M.J., Wollny, D. et al. Inferring and perturbing cell fate regulomes in human brain organoids. Nature 621, 365–372(2023). https://doi.org/10.1038/s41586–022–05279–8
  2. Jang, Hyunsoo & Kim, Seo & Koh, Youmin & Yoon, Ki-Jun. (2022). Engineering Brain Organoids: Toward Mature Neural Circuitry with an Intact Cytoarchitecture. International Journal of Stem Cells. 15. 41-59. 10.15283/ijsc22004.
  3. Matsuo I, Hiramatsu R. Mechanical perspectives on the anterior-posterior axis polarization of mouse implanted embryos. Mech Dev. 2017 Apr;144(Pt A):62-70. doi: 10.1016/j.mod.2016.09.002. Epub 2016 Sep 30. PMID: 27697519.
  4. Wang L, Owusu-Hammond C, Sievert D, Gleeson JG. Stem Cell-Based Organoid Models of Neurodevelopmental Disorders. Biol Psychiatry. 2023 Apr 1;93(7):622-631. doi: 10.1016/j.biopsych.2023.01.012. Epub 2023 Jan, PMID: 36759260; PMCID: PMC10022535.
  5. Luecken, M. D., & Theis, F. J. (2019). Current best practices in single‐cell RNA‐seq analysis: a tutorial. Molecular Systems Biology, 15(6). 88746 https://doi.org/10.15252/msb.201
  6. Bilous M, Hérault L, Gabriel AA, Teleman M, Gfeller D. Building and analyzing metacells in single-cell genomics data. Mol Syst Biol. 2024 Jul;20(7):744-766. doi: 10.1038/s44320-02400045-6. Epub 2024 May 29. PMID: 38811801; PMCID: PMC11220014.
  7. Molnár Z, Clowry GJ, Šestan N, Alzu’bi A, Bakken T, Hevner RF, Hüppi PS, Kostović I, Rakic P, Anton ES, Edwards D, Garcez P, HoerderSuabedissen A, Kriegstein A. New insights into the development of the human cerebral cortex. J Anat. 2019 Sep;235(3):432-451. doi: 10.1111/joa.13055. Epub 2019 Aug 2. PMID: 31373394; PMCID: PMC6704245.
  8. Sunderland EM, Hu XC, Dassuncao C, Tokranov AK, Wagner CC, Allen JG. A review of the pathways of human exposure to poly- and perfluoroalkyl substances (PFASs) and present understanding of health effects. J Expo Sci Environ Epidemiol. 2019 Mar;29(2):131-147. doi: 10.1038/s41370-018-0094-1. Epub 2018 Nov 23. PMID: 30470793; PMCID: PMC6380916.
  9. Li, C., Fleck, J.S., Martins-Costa, C. et al. Single-cell brain organoid screening identifies developmental defects in autism. Nature 621, 373–380 (2023). https://doi.org/10.1038/s41586–023–06473–y
  10. Morris KV, Mattick JS. The rise of regulatory RNA. Nat Rev Genet. 2014 Jun;15(6):423-37. doi: 10.1038/nrg3722. Epub 2014 Apr 29. PMID: 24776770; PMCID: PMC4314111.
  11. Liang WW, Müller S, Hart SK, Wessels HH, Méndez-Mancilla A, Sookdeo A, Choi O, Caragine CM, Corman A, Lu L, Kolumba O, Williams B, Sanjana NE. Transcriptome-scale RNA-targeting CRISPR screens reveal essential lncRNAs in human cells. Cell. 2024 Dec 26;187(26):7637-7654.e29. doi: 10.1016/j.cell.2024.10.021. Epub 2024 Nov PMID: 39532094; PMCID: PMC11682925.

Elisa Russo

Cellular and Molecular Biology Master student

Maria Dessi

Cellular and Molecular Biology Master student

Apri un sito e guadagna con Altervista - Disclaimer - Segnala abuso - Privacy Policy - Gestisci preferenze di tracciamento