Brain Codes
Ron McKay, PhD, Chief Editor
Lieber Institute for Brain Development

Maltz Research Laboratories
The structure of DNA seems, with hindsight, to have led inexorably to the human genome project. Perhaps because we are captivated by the structural beauty of the double helix, we often think of the code as static but of course this is not the case. The code is a language and like all living languages, meaning is encoded in dynamic symbols. The readers and writers of the stable DNA code can now be studied genome-wide in the human brain. Through the development of powerful new tools we can access the inter-twined dynamics of gene expression and the interacting chromatin proteins that control this key aspect of the language of life. 
This edition of neuroDEVELOPMENTS is focused on two recent papers showing the extraordinary power that flows from defining the chromatin states of the developing and adult brain. One of these is formally part of a cooperative program, PsychENCODE, focused on discovering the DNA elements that regulate gene function in the brain. The other, from the laboratory of one of our Board members, dissects chromatin change in the developing human cerebral cortex. Both these papers illustrate the extraordinary cooperation between researchers that is revealing the dynamics of chromatin change in the nervous system.
Scientific progress is often abruptly accelerated by a technical advance. I can remember the excitement on first reading Fred Sanger’s report of the nucleic acid sequence of the bacterial virus PhiX174.  Now that we had genetic sequence, the world had been transformed. The last decades have seen several explosions in our understanding of DNA, RNA and the nuclear proteins that comprise chromatin. The newest tools allow systematic analysis of the specialized proteins that control gene function in the developing and adult brain. They promise a future where we have many human epigenomes to explore as we seek fluency in the dialects that bring our nervous system into being. For a range of perspectives on this topic, please read the comments from the Board. Thanks to all the Board members for their contributions. Thanks to the Lieber and Maltz families for their support of this project. And thanks to you for your interest.
neuroDEVELOPMENTS Editorial Board
Fred 'Rusty' Gage, PhD
President, The Salk Institute for Biological Studies

Daniel Geschwind, MD, PhD
Professor, UCLA School of Medicine

Yukiko Gotoh, PhD
Professor, University of Tokyo

Elizabeth Grove, PhD
Professor, University of Chicago
Jürgen Knoblich, PhD
Director, Institute of Molecular Biotechnology, Austrian Academy of Sciences 

Arnold Kriegstein, MD
Professor, UCSF

Pat Levitt, PhD
Professor, Keck School of Medicine of USC

Mu-Ming Poo, PhD
Director, Institute of Neuroscience, Chinese Academy of Sciences

John Rubenstein, MD, PhD
Professor of Psychiatry, UCSF

Nenad Sestan, MD, PhD
Professor, Yale University 

Chris Walsh, MD, PhD
Chief, Division of Genetics & Genomics, Boston Children's Hospital


Venkata S. Mattay, MD
Managing Editor

Michele Solis, PhD
Science Writer 

Caught in the act:  

Brain tissue studies reveal essential genetic machinery for building the human brain
The human genome has been breathlessly called a blueprint for human life, but in reality, it is even more than that. Like a blueprint, it depicts all the protein parts a human needs, in the form of genetic material. But these parts make up only 5% of the three billion base pairs of the genome. The rest is thought to harbor controls on whether, when, and how a gene should be transcribed into RNA and then translated into a protein, akin to a thick instruction manual. Importantly, a significant amount of genetic risk for neuropsychiatric disorders like autism, schizophrenia, or depression resides in this instruction manual portion of the genome.

To understand how the genome guides construction of the brain, and how this process is perturbed in human neuropsychiatric disease, we need to go to the source, human brain tissue. Even though all cells contain the same genes and largely the same genetic material, the precise itineraries in use differ across cell types, tissues, and even species. Human brains themselves provide the gold standard for elucidating development and function. Moreover, because the human brain is both immensely complex and extraordinarily dynamic during development, it is essential that analyses are conducted at multiple time points across development and into adulthood. Only by doing so can we identify the specific sets of instructions that control an epigenetic itinerary.

Researchers around the world are now measuring the genes and gene regulatory, or epigenetic signals that determine the constellation of RNAs and proteins in each cell of the brain. The two papers we discuss tackle the question of gene regulation in the developing and adult brain. They are part of a multi-national program to capture how psychiatric risk alters the epigenetic regulation of gene expression in human brain cells.

"Both studies (featured in this issue) are creating much needed maps of regulatory and transcriptomic landscapes in the developing and adult human brain." – Nenad Sestan  

"Both of these papers are tours de force and models to learn from, but neither is easily validated." – Fred 'Rusty' Gage


"Lineage tracing studies that link development of a given cell type and its mature functions after differentiation will be essential for coupling genomics (and other -omics studies) to phenotypes..." – Yukiko Gotoh  

Figure 1. Human prenatal brain. Studies of human prenatal brain are essential for understanding early brain development. Figure 1A shows, on the left, a coronal slice of human brain from post-conception week 15. The cortex (ctx), germinal zone (GZ), and cortical plate (CP) are labeled; the boxed region on the left is enlarged on the right, where the green stain shows cells. The researchers contrasted chromatin accessibility between the GZ, where progenitor cells are born, and the CP, where newborn cells become neurons. This comparison begins to delineate the different types of gene regulation underpinning different stages of brain development and disease risk (Figure 2).

"...these disorders appear to be caused not by deletion of a gene, but rather inactivation of distant regulatory elements." – Elizabeth Grove

"The understanding of diseases such as autism and schizophrenia will almost certainly have to involve the study of developing human tissue." – Arnold Kriegstein

"The stage appears to be set for studying the molecular mechanism underlying neuronal subtype differentiation in different cortical layers." – Mu-ming Poo

Chromatin unwound

DNA’s double helix is packed away for safekeeping, spooled around other proteins in a dense structure called “chromatin.” When a new protein needs to be made, the appropriate section of chromatin unwinds to expose the gene as well as the regulatory regions necessary for turning a gene into a protein. Led by Dan Geschwind, an editorial board member of neuroDEVELOPMENTS at the University of California, Los Angeles, researchers used a technique that pinpointed spots of unwound chromatin in cells from human prenatal brain tissue (Figure 1). This approach identified which genes are called upon during brain construction and also revealed non-coding regulatory regions of the genome, the promoters and enhancers that implement bursts of transcription. Promoter sequences are closely associated with the coding sequence of the gene they regulate but enhancers can be distant. Nearly half of all human enhancers lie far away from their target gene, and chromatin forms loops to bring these far off enhancers and their associated activating proteins in contact with the target gene. This makes the linear proximity in the genome a misleading clue to the interaction between genes and enhancers.

Across different tissues and diseases, new techniques, such as the ones in this paper, are being deployed to map in three dimensions the physical interactions between enhancers and the genes they regulate. Geschwind likens the findings to aligning different kinds of maps, such as adding airline flight routes or highway systems or political boundaries onto geographical maps. “There are different kinds of maps that have different functional information. This study brings together multiple maps into one showing how key elements interact,” he says.

Using these new maps, researchers linked recently evolved human-specific enhancers to their target genes in the developing cerebral cortex, ground zero for human cognition and risk for psychiatric disorders. These genes were highly expressed in outer radial glia, cells that divide early in development to become neurons in nascent brains. Outer radial glia are especially numerous in the human forebrain and these new data provide further evidence about their roles in powering the cortical expansion that distinguishes human brain evolution.

Out of the hundreds of thousands of possible enhancers, this study focuses on two that play key roles in human forebrain development. One enhancer controls a transcription factor EOMESODERMIN (Gene name: TBR2) that is central to the specification of neuronal fates and the other regulates a growth factor receptor, fibroblast growth factor receptor 2 (FGFR2). FGFR2 plays a role in intellectual disability. In both cases, the researchers confirmed that disabling the human-specific enhancers decreased the levels of protein in brain precursor cells and disrupted the production of neurons. They also show that other genetic variants previously linked to brain size, level of education, and neuropsychiatric disorders were enriched for functions in the early developing prenatal brain (Figure 2). This work shows how subtle differences in gene regulation in early development reverberates into adulthood, emphasizing that many brain disorders thought to arise late in life have early developmental components that require understanding their origins in prenatal brain tissue.

Figure 2. Brain-related traits can arise early. Upon evaluating genetic variants already associated with different phenotypes (listed on the left), the researchers found these variants enriched in regions of accessible chromatin, particularly those found in the germinal zone (GZ). For example, variants associated with schizophrenia showed a significant enrichment in germinal zone (GZ) cells as indicated by blue bars that surpass the red threshold line; in contrast, enrichment in cortical plate (CP) cells, which mark a later stage of development, was not observed as often. This suggests that risk for disorders like schizophrenia acts during the early proliferative phase of brain development rather than during subsequent differentiation and maturation of neurons.

DNA dashboard  

To bring together a growing number of new techniques aimed at deciphering gene regulation in the developing and adult brain, the PsychENCODE Consortium formed. Here we highlight this ongoing effort by focusing on one new study using an unprecedented number of 1866 adult human brain samples to reveal a wide range of variation in gene regulation across individuals. The team measured signals in brain cells that flag regulatory elements like promoters and enhancers, sites of action for transcription factors. When they integrated the data to link these regulatory elements to their target genes, they found over 80,000 elements that regulated 11,573 protein-encoding genes. This reveals a veritable DNA dashboard of controls that command gene expression in the brain (Figure 3).

The researchers consulted their new map of regulatory elements to ascribe function to 142 non-coding genetic variants associated with schizophrenia. These variants fell within regulatory elements involved with 321 genes, including calcium channels, cholinergic receptors, and synapses; these genes were highly expressed in excitatory neurons, which suggests these cells are of particular interest for schizophrenia research. How are we to make sense of this deluge of new data sets? Here, the researchers turned to machine learning approaches to incorporate the multiple types of information obtained from the brain tissue to predict how epigenetic signals control brain function. They developed a “Deep Structured Phenotype Network” that was six times better at predicting brain-related phenotypes than models based solely on DNA variation. This indicates that the data about gene regulation from brain tissue can add substantial power to results from genetic studies that have relied primarily on increasing the sample size of individuals studied, rather than adding these new data types.

Both papers use powerful new methods to catalog a vast regulatory landscape within the human brain. Multiple regulatory elements can act on a single gene to orchestrate its expression, and the nature of these interactions are being refined in a deluge of new data on human brain cells.

"Epigenetic signatures are a proxy for proof that a given locus is a regulatory element."-John Rubenstein

"A lot of geneticists haven’t fully appreciated how useful these functional data can be." – Daniel Geschwind

"Machine learning approaches are key to recognizing patterns in large datasets like the ones analyzed in these papers."-Jürgen Knoblich 

Figure 3. An interconnected genome. A circular network graph depicts the epigenetic regulation occurring within excitatory neurons in the brain. The outside circle represents all 24 chromosomes, aligned end-to-end; genes of note are labeled within the circle. Lines connect enhancers to promoters and their adjacent target genes. Enhancers can act from afar; for example, an enhancer on one chromosome can affect transcription of a gene on an entirely different chromosome. The pattern of interactions is different for different cell types, both within the brain and between tissue types.
The full commentary from our Editorial Board on the two papers highlighted in this issue of neuroDEVELOPMENTS.
 1. From Arnold Kriegstein, MD, University of California, San Francisco:
An important aspect of the de la Torre-Ubieta et al. study is the finding that many of the target genes of human-gained enhancers are enriched in outer radial glia cells. This cell type helps expand neural production and is particularly abundant in the developing human neocortex, but not in small brain mammals such as the mouse. Thus, the discoveries reported here could not have been made using a mouse model. To the extent that these gene changes are associated with neurodevelopmental diseases, it also points to a limitation in our ability to study these disease mechanisms in non-human animal models. The understanding of diseases such as autism and schizophrenia will almost certainly have to involve the study of developing human tissue. For this reason, the use of federal funding to study human brain tissue across the lifespan must be maintained. The importance of this precious resource, donated with appropriate informed consent and in accordance with ethical guidelines, is highlighted here.

2. From Fred “Rusty” Gage, PhD, Salk Institute, San Diego, California:
Both of these papers are tours de force and models to learn from, but neither is easily validated. The same strategies are being applied to Alzheimer’s disease and to cancer.

A feature worth noting is the difficulty in linking a non-coding variant (single nucleotide polymorphism, SNP) to a gene. This is usually done by "closest gene" characterization, but one of the great things in the de la Torre-Ubieta paper is that they show this is not necessarily the way things work. When they disabled a SNP faraway from a gene called FGFR2, they saw a decrease in FGFR2 expression. This was good validation of a relationship between the SNP as an enhancer and FGFR2, but multiple techniques are needed to confirm this, or even to make a prediction about a potential enhancer and the gene it acts on in the first place.

3. From Elizabeth Grove, PhD, University of Chicago:
Luis de la Torre-Ubieta and colleagues used methods that could associate distant enhancers with the genes they regulate. For example, they found five distant regulators of a gene associated with microcephaly in humans (EOMES), and indicated that these disorders appear to be caused not by deletion of a gene, but rather inactivation of distant regulatory elements. The bigger picture is that many human brain disorders could be caused in this way, making it vital to identify distant regulatory elements of genes implicated in these disorders.

4. From Mu-ming Poo, PhD, University of California, Berkeley, and Chinese Academy of Sciences, Shanghai, China:
In the de la Torre-Ubieta paper, I am mostly intrigued by the finding of differential enrichment of genes regulated by human gained enhancers within various prenatal cortical laminae and the identification of differentially expressed transcription factors in the germinal zone versus the cortical plate. The stage appears to be set for studying the molecular mechanism underlying neuronal subtype differentiation in different cortical layers. How does the genetic program unfold during radial migration of differentiating neurons? Is there layer-specific differentiation of cortical neuron subtypes? Does such subtype differentiation endow the layer-specific input/output connectivity of cortical neurons or vice versa? The results in this paper call for further layer-specific single-cell transcriptome analysis, an approach discussed in the first issue of neuroDEVELOPMENTS, together with mapping of a single cell connectome in the developing cortex.

5. From Jürgen Knoblich, PhD, Institute of Molecular Biotechnology, Vienna, Austria:
Machine learning approaches are key to recognizing patterns in large datasets like the ones analyzed in these papers. The ever-growing amounts of data available in genomics and transcriptomics cannot feasibly be analyzed in any other way. I would predict that this will soon also be true for the extensive diagnostic datasets available for each patient even in a standard clinical setting. Analysis of those data will shift from classical differential diagnosis performed by humans to computer-assisted development of therapeutic strategies. Importantly, such analysis will involve genome and epigenetic data such as the ones discussed in the papers as improving technology will make it feasible to obtain those data routinely from each patient.
In the times of modern human genetics, the Wang et al., study provides a roadmap for future genetic analysis of neurodevelopmental disorders. As such, it is as important for genomics as the annotation of open reading frames was after complete sequencing of the human genome.

6. From Daniel Geschwind, MD, PhD, University of California, Los Angeles:
A lot of geneticists haven’t fully appreciated how useful these functional data can be. But here, in Wang et al., we show how those intermediate levels of data can substantially increase the power and ability to call the relationship between genotype and phenotype.

7. From John Rubenstein, MD, PhD, University of Carlifornia, San Francisco:
Epigenetic signatures are a proxy for proof that a given locus is a regulatory element. Functional assays in the relevant cell types, preferably in vivo, but alternatively in suitable cell culture systems, are needed to identify when and in what cell types, and under what conditions the regulatory element is active. Furthermore, it is possible that regulatory elements that are relevant to human neuropsychiatric disease are functioning in an activity-dependent basis in the postnatal brain. Whether or not a regulatory element is active does not mean that it necessarily has an essential function, as there is evidence that regulatory element duplications lead to functional redundancy.

8. From Yukiko Gotoh, PhD, University of Tokyo, Japan:
Given the promise of the types of studies exemplified in both papers, the identification of cell types in the developing brain and characterization of their developmental behavior (when, where, and how they are born and what adult cell types they become or generate) are obviously important. That said, the identification of cell types in the adult brain and the lineage connection between embryonic and adult cell types are also important. Lineage tracing studies that link development of a given cell type and its mature functions after differentiation will be essential for coupling genomics (and other -omics studies) to phenotypes and should greatly increase our understanding of brain development and developmental psychiatric disorders.
Also, some non-coding RNAs do not regulate gene expression per se but instead play other roles such as providing scaffolds for proteins at nonmembranous organelles. In addition to multiomics platforms based on gene expression, we may also need other platforms to analyze the enormous number of non-coding RNAs expressed in the brain.

9. From Nenad Sestan, MD, PhD, Yale University, New Haven:
Both studies are creating much needed maps of regulatory and transcriptomic landscapes in the developing and adult human brain. Crucially, these advances have only been possible through the direct study of the human brain, and continued work on both developmental and adult tissue is essential for the further elucidation of the mechanisms underlying brain development, evolution, function, and disease. 
Copyright © 2019 Lieber Institute for Brain Development, All rights reserved.

Want to change how you receive these emails?
You can update your preferences or unsubscribe from this list.

Email Marketing Powered by Mailchimp