An Interesting Problem: Neural Wiring

The question I decided to pursue is how hard-wired neural circuits can be created using genetic instructions. How does nature build a brain that can completely self-assemble and function as soon as an animal is born? Over the past few decades, we have learned much about how neural circuits refine through activity-dependent plasticity, such as for learning and memory. However, we know little about how innate behaviors, such as breathing or reflexes, are prespecified in hard-wired circuits. Even a “simple” neural circuit with hundreds of neurons can contain millions of synaptic connections between them. This synaptic connectivity must also be extremely precise for any hard-wired circuits that underlie behaviors that affect the fitness of the animal, such as an escape response. The assembly instructions for hard-wired circuits must be written molecularly within the genome, and my research focuses on deciphering these instructions to uncover the molecular codes and strategies that neurons use to connect with each other. I chose to investigate this question to help us understand the basic rules governing how genetic information translates into the synaptic targeting decisions of neurons to produce innate behaviors.

Dscam, Mother Nature’s Innovation

My introduction into how molecular information can specify the synaptic targeting decisions of a neuron came from investigating a gene called Down Syndrome Cell Adhesion Molecule (Dscam) in the fruit fly genetic model organism Drosophila melanogaster. DSCAM is was discovered on human chromosome band 21q22, within a~4-Mb region called the Down syndrome critical region.Reference Yamakawa, Huot and Haendelt ⁵ Because the human gene was predicted to encode a large receptor with multiple immunoglobulin domains, it was named the Down Syndrome Cell Adhesion Molecule. The Drosophila Dscam protein is identical in domain structure to human DSCAM, placing Dscam’s origin at more than 600 million years ago. However, Drosophila Dscam is a very complex molecule with a unique feature: through alternative splicing, the gene can produce more than 100,000 different protein isoforms—the record holder for alternative splicing.Reference Schmucker, Clemens and Shu ^{6 ,} Reference Schmucker and Chen ⁷ Different RNA isoforms are created by splicing together different alternate exons within the gene, and each of these RNA isoforms is translated into protein isoforms that all maintain the same overall Dscam protein architecture. This extensive isoform diversity has been proposed as a synapse-specific molecular label through different binding affinities of Dscam isoforms.Reference Schmucker and Flanagan ⁸ To determine whether Dscam plays a role in synapse specification, I used a stereotyped axonal branching pattern as a readout for errors in synaptic targeting. By combining this with single-neuron genetic mosaic analysis and exon-specific Dscam deletions, I showed that Dscam’s extensive isoform diversity is required for proper synaptic targeting at the level of single axonal branches and that different isoforms are used to specify different synaptic contacts.Reference Chen, Kondo and Garnier ⁹ This study from my postdoctoral research opened the possibility that complex neural circuits can be prespecified genetically using diverse molecular labels generated by alternative splicing rather than being refined through the experience of the animal. This work was the foundation for the research in my laboratory investigating how hard-wired neural circuits wire up.

Dscam has now been shown to be an important molecule involved in many aspects of neural development in invertebrates and vertebrates including axon guidance, axonal and dendritic branching, synapse targeting, synaptic plasticity, and cell death (reviewed elsewhereReference Schmucker and Chen ^{7 ,} Reference Montesinos ¹⁰ ). However, vertebrate DSCAM is not extensively alternatively spliced, and my long-term goal is to understand how hard-wired neural circuits are molecularly prespecified, particularly in the mammalian brain. I use the Drosophila fruit fly as a tractable model system and investigate Drosophila Dscam hoping to reveal general strategies about how neural circuits use a diversity of cell-surface receptors to specify precise synaptic contacts. These insights also help guide my laboratory’s investigations into vertebrate mechanisms of neural wiring. Other laboratories have discovered parallel mechanisms—for example, in neurite self-avoidance, Drosophila Dscam isoform diversity or mammalian Protocadherin isoform diversity is used by neuronal branches to avoid self-crossing (reviewed elsewhereReference Kise and Schmucker ^{11 ,} Reference Zipursky and Grueber ¹² ). For me, this has left more questions than answers about the need for isoform diversity because not all complex neurons and neural circuits require it. Why does the Pancrustacea clade (comprising crustaceans and hexapods, the majority of the animal biomass on the planet) all require such extraordinarily large isoform diversity of Dscam?Reference Schmucker and Chen ^{7 ,} Reference Brites, Brena, Ebert and Du Pasquier ^{13 ,} Reference Chipman, Ferrier and Brena ¹⁴ What, then, are the genetic instructions to prespecify a hard-wired complex neural circuit in mammals? In the end, why investigate Drosophila Dscam if there are no human genes (mammalian protocadherins generate around 50 isoforms) with remotely similar extensive alternative splicing?

Dscam is the most complicated gene ever discovered. Here I provide brief examples of how curiosity-driven research on Dscam has pushed innovation in my laboratory and produced unexpected and far-reaching outcomes, often beyond neuroscience. Much of these advances arose simply because of the size and complexity of the Dscam gene. The Drosophila Dscam gene is 75 kb with 25 exons, but within variable exons 4, 6, 9, and 17, there exist tandem arrays of nearly 100 alternate exons.Reference Schmucker and Chen ⁷ Thus, performing or proposing experiments on Dscam forces you to think big. No meaningful experiments can be performed on the standard scale because of the hundreds of thousands of isoforms confounding the interpretation. Standard protein structure-function analyses are useless because of the absurd number of protein isoforms to manipulate. Loss of function or necessary and sufficiency analyses are equally difficult because of the broad range of phenotypes that may result from compensating or misregulated isoforms. Thus, it is nearly impossible to ascribe any specific function to specific isoforms. These theoretical and conceptual complexities become real-world challenges when there are limited funds for newly starting, unproven investigators studying fruit flies. Here it becomes even more critical to understand what impactful and interpretable result will come and what significant question will be answered at what monetary and time cost. But the riddle of Dscam will always be a fascinating problem to solve.

Down Syndrome Cell Adhesion Molecule and Neural Wiring: Implications for Down Syndrome and Fragile X Syndrome

Dscam has been shown to be involved in several phenotypes associated with Down syndrome.Reference Alves-Sampaio, Troca-Marin and Montesinos ^{15 -} Reference Hildmann, Kong and O’Brien ¹⁷ Down syndrome is the result of triplication for human chromosome 21, which causes an overexpression of the thousands of genes located there. Similarly, Fragile X syndrome is the result of an overexpression of thousands of genes resulting from the silencing of the Fragile X Mental Retardation gene. This gene encodes Fragile X Mental Retardation Protein (FMRP), a suppressor of protein translation; excessive protein synthesis of FMRP’s thousands of normally suppressed messenger RNA (mRNA) targets is thought to underlie the neuronal defects in Fragile X syndrome. Interestingly, DSCAM mRNA was found to be selectively associated with FMRP (mouse and human) in high-throughput RNA sequencing and microarray screens identifying molecules involved in Fragile X syndrome and autism spectrum disorders.Reference Ascano, Mukherjee and Bandaru ^{18 -} Reference Darnell, Van Driesche and Zhang ²⁰ Thus, my laboratory sought to determine whether overexpression of Drosophila Dscam protein through gene triplication (reflecting the Down syndrome trisomy 21 case) or through loss of translational suppression by FMRP impairs synaptic targeting precision and neural circuit function.Reference Cvetkovska, Hibbert, Emran and Chen ²¹

To examine how excessive Dscam protein levels affect neural wiring decisions, we used a hard-wired mechanosensory circuit in Drosophila that allows for quantitative analysis of synaptic targeting errors.Reference Chen, Kondo and Garnier ^{9 ,} Reference Neufeld, Hibbert and Chen ^{22 ,} Reference Kays, Cvetkovska and Chen ²³ This mechanosensory system also allows for a functional assessment of synaptic connectivity through the animal’s reflex behavior by mechanically stimulating the neuron.Reference Kays, Cvetkovska and Chen ²³ Using Drosophila genetics, we removed FMRP solely within four mechanosensory neurons, leaving the rest of the animal and nervous system normal. We found that this resulted in reproducible synaptic targeting errors within just these mechanosensory neurons and disrupted the animal’s mechanosensory perception. Animals with three copies of the Dscam gene had similar synaptic targeting errors and impaired touch responses as those with loss of FMRP. Reducing the Dscam gene dosage in the FMRP mutant animals reduced the synaptic targeting errors and rescued the behavioral responses, indicating that much of the errors in neural wiring and synaptic function were the result of excess Dscam protein in these neurons. We were also curious as to whether FMRP regulated all Dscam RNA isoforms. We used next-generation sequencing to obtain long read lengths of nucleic acids and deep coverage of more than 1 million reads and found that all possible isoforms were equally associated with FMRP, with no significant bias in the Dscam isoforms bound. This study demonstrated that excess Dscam protein might be a common molecular mechanism underlying altered neural wiring in two of the most common causes of intellectual disability: Down syndrome and Fragile X syndrome.Reference Cvetkovska, Hibbert, Emran and Chen ²¹ Because Dscam function is evolutionarily conserved across vertebrates and invertebrates in both cardiac and neural development,Reference Schmucker and Chen ^{7 ,} Reference Grossman, Gamliel and Wessells ²⁴ we speculate that Dscam protein expression is tightly regulated in cells, and dysregulation in neurons may underlie a wide variety of neural developmental disorders.

How Does Protein Concentration Affect Neural Phenotypes?

To examine how dysregulation of mammalian DSCAM protein might affect cortical development, we sought to overexpress DSCAM in mouse neocortical neurons. It would be beneficial to understand how excess DSCAM protein concentration produces different abnormal neural phenotypes and, ideally, how the variations in concentration of endogenous DSCAM protein might correlate with the variations in these same phenotypes in normal neurons. Although DSCAM is expressed at high levels in the developing neocortex,Reference Barlow, Micales, Chen, Lyons and Korenberg ²⁵ the natural heterogeneity of cortical neurons and their large variability in phenotypes—for example, in dendritic branching structure and ion channel propertiesReference Spruston ^{26 ,} Reference Somogyi and Klausberger ²⁷ —precludes any meaningful analysis of DSCAM expression with neural phenotype at the single-neuron level, unless the effect size is large or the neurons are as stereotyped as Drosophila neurons.

Thus, we needed a method to correlate the variation in neural phenotypes with concentration of DSCAM protein at the single-cell level. Accurate quantification of protein amounts is very difficult at this level. The only current ways to quantify protein levels are using crude techniques such as protein assays and immunoblots. Essentially, these entail dissecting and homogenizing heterogeneous tissue from large numbers of animals, sometimes from irrelevant developmental stages or tissues because of widely accepted dissection practicalities. Because of poor sensitivity and detection capabilities, these techniques become laborious and time-consuming, but also less relevant because of their poor cellular resolution. All current techniques are also necessarily destructive to cells and animals. To overcome these limitations, we developed a technique to quantitate protein amounts in single living cells, which we call “protein quantitation ratioing.” This technique uses modified virus sequences that allow for an equimolar separation of an upstream protein of interest and a downstream protein of interest, all contained within a single strand of RNA. When a fluorescent reporter is separated from the protein of interest, the number of fluorescent molecules produced is stoichiometric with the number of molecules of the protein of interest produced, and thus the fluorescence output (i.e. brightness of the cell) can be used as a readout for the number of molecules of interest produced (i.e. its relative protein concentration).Reference Lo, Kays, Emran, Lin, Cvetkovska and Chen ²⁸ We invented this technique to correlate DSCAM protein concentration with neural phenotype in single neurons in the living animal. By simply using the cell’s brightness as a readout for the protein expression level of a gene, we hope that our protein quantitation ratioing technique will have a wide range of creative applications in cell biology to quantitatively measure relationships between phenotypes and protein levels, with broad expansion possibilities. Almost every cell biologist requires the quantification of protein amounts at some point; thus, our technique could have a large impact across biology.

Visualizing and Browsing the Most Complicated Gene Ever Discovered

Performing experiments on Drosophila Dscam is difficult because of the complexity of the alternative splicing. Even visualizing the gene itself is a challenge, whether when designing experiments or explaining the gene’s alternative splicing to laboratory trainees. To ease this burden, we created a web application for online visualization and browsing of the different alternatively spliced RNA and protein isoforms (brianchenlab.mcgill.ca/dscam). This web app allows for visual selection of isoforms by moving the cursor over the desired alternate exon, which will then dynamically update and display the corresponding mRNA and amino acid sequences. An interactive protein schematic of Dscam is also displayed above the mRNA and amino acid sequences for structure-function experiments, so that specific protein domains (e.g. immunoglobulin domain 7) can be selected and the corresponding mRNA and amino acid sequences are highlighted. These are only a few of the major features that are included in our web app, designed and deployed within one summer. Seeing this as inspiration, I realized that if this could be accomplished for the most complex Drosophila gene, then it could be accomplished for the entire Drosophila genome and, by extension, all sequenced genomes available; soon afterwards GeneDig.org was created. Both of these web tools were developed because of the striking lack of resources available for accessing genomic and bioinformatics data for standard biology applications.

GeneDig is a web application for user-friendly access to genomics and bioinformatics knowledge,Reference Suciu, Aydin and Chen ²⁹ and is designed to be comprehensive yet intuitive, allowing for searching and browsing of all sequenced genes and genomes with ease and efficiency. Our dynamic navigator displays genomic, RNA, and protein information simultaneously for co-navigation, with disease and traits overlaid on associated genes. GeneDig is freely available in more than 15 languages at GeneDig.org. I created GeneDig to benefit any general user that requires information about specific genes or diseases, for scientists, health care practitioners, patients, or science education. I believe that access to genomic information is important, and my goal with GeneDig is to make all genomic information easily accessible and useful.

Disclosures

BC is a researcher for and has received research support from the Canadian Institutes of Health Research and Natural Sciences and Engineering Research Council of Canada.