Zhisong Jeff HE
何志嵩

Transcriptomic profiling of Ngn2-induced iPSC-derived neurons

​

The directly-induced neurons from human iPSCs through Neurogenin 2 (Ngn2) overexpression are widely used to study neuronal differentiation mechanisms and to model neurological diseases. However, the differentiation path and the heterogeneity of the emerged neurons is not yet clear. This project aims to dissect the cell states that emerge during Ngn2 overexpression across a time course from pluripotency to neuron functional maturation using scRNA-seq technology.

​

The manuscript has been preprinted in biorxiv in 2020, and published by Stem Cell Report in 2021.

​

Autistic changes of laminar transcriptome in human prefrontal cortex 

​​

Mammalian neocortex is characterized by its six-layer structure. These cortical layers consist of different cell type compositions and neuronal connectivity, therefore present varied functions. Their harmonious integration is critical for brain functions. In neural disorders, e.g. autism, physiological alterations including changes of connectivity among brain regions, as well as molecular alterations including gene expression changes, have been observed. However, no systematic and comprehensive research has been reported to study whether the local laminar structure of neocortex is disrupted in these diseases. This project aims to answer this question, by applying the recently developed unsupervised sectioning experimental procedure to study the transcriptome alterations of cortical layers in prefrontal cortex in autistic patients.

​

The manuscript is now in revision.

​

Temporal and pathological lipidome changes in human prefrontal cortex 

​​

Lipids, the important structural and functional molecules which composes more than 50% of the dry weight in brain. Lipids is not only the essential building material of membranes and bioenergetic fuel, but can also act as signaling messagers in neural tissues, as well as influence vesicle fusion and fission processes, ion flux and leteral diffusion of membrane proteins by regulating the chemical and mechanical properties. The evolutionary study of lipidome in primate brains also implied the relationship between lipids and cognitive ability. While the temporal changes of varied molecular levels in human brains have been observed, there is still lack of comprehensive investigation on the lipidome level. Furthermore, it is known for a long time that the disruption of the brain developmental trajectories could lead to multiple neural disorders including autism and schizophrenia. However, it largely remains unclear that how the brain lipidome responds to the disorders and what is the role of the brain lipid alteration there. Therefore, this project in designed to investigate the temporal lipidome profiles of human prefrontal cortex as well as its behavior in the two neural disorders, namely autism and schizophrenia, across lifespan.

​

The manuscript has been published in Molecular Psychiatry in 2018.

​

Transcriptome changes during human neuron maturation

​

Neuronal maturation is the process linking differentiation from neural progenitor cells and fully matured neurons. Huge physiological changes, e.g. synaptogenesis and synaptic maturation, have been observed during this process. However, the molecular dynamics in behind remains largely unclear. To figure out the molecular changes of this process, we developed the methodology to analyze gene expression dynamics in the context of protein-protein interaction network. Based on the identified functional modules relevant to neuron maturation, we developed the Neuron Maturity Index (NMI) model as a new metric of neuron maturity state. By using the NMI measurement, we found that the transcriptome transition during neuron maturation is conserved in mouse. We also found the in vitro human neuron models likely encountered maturation arrest due to the lack of proper environment.

​

The manuscript has been published in BMC Genomics in 2018. Please refer to the Data page for the used data.

​

Investigation of composition-dependent and independent component of the human temporal and autistic brain transcriptome changes

​

The development and aging of human brains are complex processes which are shaped by anatomical and molecular changes. Neurodevelopmental disorders, e.g. autism, disrupt the temporal trajectories of brains and result in disruption of cognitive ability. Dozens of studies have been conducted to character the temporal and pathological molecular changes in human brains. However, the human brain is a highly complex and heterogeneous organ comprised of different cell types including neurons, glias and vascular cells with distinct molecular signature. Thus, it remains largely unknown that whether the age-related and/or pathological molecular changes are primarily caused by changes of cell type compositions. Furthermore, the biological meaning of the rest changes is also unclear. To investigate such question, we dissected the temporal and autism pathological gene expression profiles into composition-dependent and independent components, and analyzed the two components separately for their expression patterns, functions and regulators.

 

We found that while cell type compositions explained the majority of transcriptome changes in both scenarios (from immature neurons to mature neurons in early postnatal development; decreased RNA contributions of neurons in autism patients), the composition-independent transcriptome changes, which enriched for cell-cell communications including synaptic functions in both cases, also significantly contributed to the observed changes. Interestingly, the genes with their composition-independent expression components changes in autism patients, but not those with only composition-dependent components changed, enriched for autism-associated genetic variations, suggesting that the composition-independent changes may represent more important biological influences.

​

The paper has been published in Scientific Report in 2017.

​

Molecular signature of pathological change at the fear neural circuit in PTSD

​

PTSD, which is short for post-traumatic stress disorder, may develop after a person is exposed to one or more traumatic events. PTSD is thought to be strongly linked with changes happened in the fear neural circuit. However, the detailed molecular mechanism of such change remains unknown. In this work, our collaborators in MPI of Psychiatry in Munich, Germany developped the mouse model of PTSD. By generating the transcriptomic, proteomic and metabolomic data using high-throughput sequencing and mass spectrometry for six brain regions involved in the fear neural circuit, we compare the abundance of RNAs, proteins and metabolites in normal mice and mice sufferring PTSD, to identify the molecular signature of PTSD, and try to elucidate the possible mechanism of PTSD in molecular level.

​

This project resulted in two papers, both published in 2016. The first paper publishing in Journal of Psychiatric Research talks about the transcriptome investigation, and the second one publishing in Molecular Neuropsychiatry talks about the proteomics and metabolomics studies.

​

Conservation and functional implication of lincRNA in primate brains

​

LincRNA is short for long intervening non-coding RNAs. Different from the typical RNAs involved in the central dogma including mRNA, tRNA and rRNA, as well as the short ncRNAs such as microRNA, siRNA and piRNA, lincRNA is a large category of RNAs which is long but encodes no protein. Since the dicovery of the first lincRNA XIST, there are quite a few functional lincRNA that have been identified, but the functions and evolutionary characteristics of a lot more of them remain unknown. In this study, we characterized and compared the expression level trajectories of lincRNA in primate lifespan, with the result implying the general functionality of lincRNAs. The over-expression of one candidate lincR-0003 further suggested the conserved functionality of at least some lincRNAs in primate brains.

​

The paper describing the result has been published in RNA in 2014.

​

Applications of machine learning in biology

​

Machine learning approaches are a set of algorithms and statistical methods that can learn from known data to make predictions. When the detailed mechanisms of biology are so complex that most of them are still under the darkness of unknown, machine learning approaches can help us to make predictions without fully understand everything, meanwhile to figure out the potential pattern of features that related to the biological prediction.

 

During the preparation of my bachelor thesis under the supervision of Prof. Dr. CHEN Ming in Zhejiang University, as well as my two-month internship and the following collaboration with Prof. CAI Yudong (now in Shanghai University), I contributed myself to this field, and successfully applied the machine learning approaches to solve several biological problems, such as microRNA target prediction, protein phosphorylated site prediction and drug-target prediction.