Research
I can write down a recipe on an index card using far less information than it takes to completely describe a finished dish. The recipe gives me more information about universal, fundamental characteristics of the dish than does an image of one instance of the cooked product. My lab seeks to understand how neural circuits are organized in the brain to interpret information in a variety of ways. We think the best way to do this is to study the recipes (or algorithms) that development uses to construct neurons and circuits. Rather than on an index card, these recipes are written in DNA. We focus on chemosensory circuits, as we know that there are different kinds of neural circuits that use chemosensory information to different ends–for instinctual behaviors versus for learning. Our experimental work is conducted in fruit flies due to the pleasing characteristics of their brain organization and development. We hypothesize that the algorithms used to generate learning versus instinctual circuits during development are very different from one another, and require very different amounts of genomic information.


Project 1: The patterning of sex-specific circuits for sexually dimorphic mating behaviors
“Innate behaviors” are things animals do without needing to be taught and which are universal across members of a species. When animals use instinct to respond to signals from the environment, the meanings of those signals must somehow be encoded in their genomes and be unpacked during development into the hardwired structures of neuronal circuits. We think of these instinctual circuits as “bespoke”–built to purpose to connect specific sensory stimuli to specific behavioral outputs. We believe that generating such tailored circuits is the most information-intensive process in neural circuit development. As if that were not bad enough, the existence of sex-differentiated circuits and sexually mature versus juvenile circuits means that the genome must be able to produce multiple different brains. To study formation of instinctual circuits, we focus on a neural circuit that allows male fruit flies to instinctually identify other individuals as “good mates” and initiate courtship displays. This circuit forms only in males, due to the action of a male-specific transcription factor, called Fruitless. We are using a combination of molecular dissection techniques, including proteomics and single cell ATAC-seq and RNA-seq; imaging; genetics; circuit mapping; and behavior to determine how diverse neurons that work together in this circuit acquire their connections, functions, and sex differences during development.
Project 2: Randomized wiring in the mushroom body.
While we have these very special circuits dedicated to instinctual behaviors, as in Project 1, most of the objects we encounter on a day-to-day basis were not present during evolution. How does the brain represent these evolutionarily unpredicted objects? How can development efficiently produce circuits that are capable of learning to interpret *pretty much anything* our senses can detect? The majority of neurons in the human brain are likely to receive combinatorial sensory input to allow us to perceive arbitrary objects as combinations of sensory features. Non-deterministic patterning of the inputs to these neurons would be a genetically efficient way to maximize the objects that can be perceived and discriminated from one another without assigning meanings to them a priori. In this project, we are studying the development of general-purpose learning circuits, using the fruit fly mushroom body as our model. Individual cells of the mushroom body receive discrete and unpredictable combinations of sensory inputs, with different cells receiving different inputs. The size of this brain area, and the sensory inputs its cells receive, vary across members of species and over evolutionary time.
We think that such general-purpose learning circuits develop according to a very simple algorithm and that their variation and scalability are a “feature not a bug.” If the bespoke instinctual circuit is a sculptural croquembouche that needs to be made only once to the precise specifications of a multi-page recipe, a general-purpose learning circuit might be more like a sandwich, made in vast numbers with innumerable variations, built by a recipe that just says, “put some things in between two pieces of bread.”
In this project, we use anatomic approaches, like connectomics, to study the features of the finished, adult circuit; histology and longitudinal imaging to study how neurites grow and interact with each other during development; and transcriptional approaches to study the molecular “ingredients” used in cellular development. We hypothesize that randomizing connectivity in learning circuits is a compact and scalable recipe to generate the potential for a vast set of perceptions. We are also asking how these learning system neurons develop so as to make the right kind of computation (i.e. input-output transformation) to allow flexible learning.


Project 3: Engineering development to test form-function relationships in olfactory coding
Theoretical models can predict the potential utility of circuit motifs, but these models are generally not tested via experimental perturbation. In the mushroom body, our developmental studies of how neurons get their computational structure, in Project 2, has produced a very useful side effect, of allowing us to change their computations during development. We are now using these methods to manipulate neural circuit development and test how changes in circuit architecture affect sensory coding. To do this, we combine genetic, environmental, and chemical perturbations during development with functional imaging and behavioral analyses in the adult.
Project 4: The relationship between organization of genes in the vertebrate genome and their transcriptional regulation
In microbes, genes are organized in the genome according to their functions. This is not generally thought to be the case in eukaryotes. Nevertheless, genes are organized along the animal chromosome in ways that reflect their evolutionary history and that, we have shown, can predict their transcriptional regulation. Moreover, different kinds of genes are surrounded by radically different amounts of genomic space. In this project, we are focused on two extreme forms of genome organization in humans: one in which individual genes are packed together and seem to require very little regulatory information, and another in which genes are spaced extremely far apart from one another and potentially require much larger amounts of regulatory information. Curiously, both these extreme kinds of genomic regions share very high AT sequence content, while genes with “garden variety” regulation and organization have higher GC content.
Genes who are tightly spaced and share compact gene regulatory programs with each other are rapidly evolving tandem arrays of paralogues, whose protein products often directly interact with the external world. Because these genes share regulatory elements, their transcription can be mutually exclusive across different cells. In contrast, we find that genes surrounded with the largest amount of genomic space are evolutionarily ancient and conserved. We are now asking where this genomic organization came from, and what it does for these genes or the genome as a whole.
