Broadly speaking, our lab studies Recognition; that is, we study cortical processing at the intersection of perception and memory. Nowhere is this relationship between memory and perception more powerful than in the sense of smell. The mouse piriform cortex provides an exceptionally tractable model cortical circuit to probe this question. Piriform, or primary olfactory cortex, is thought to be the locus for the perception of odor objects. It is also thought to be the locus for the formation of odor memories. Research in the Franks lab therefore breaks down into two related general questions: (1) how is odor information represented in the brain, and (2) how do these representations change with experience to ultimately drive specific behaviors?

We employ an arsenal of cutting-edge methods to address these questions, including in vivo and in vitro electrophysiology and optical imaging of neural activity to examine function; molecular genetic tools to label and control subsets of odor-responsive cells; various optogenetic and chemogenetic tools to perturb defined elements of the piriform circuit; computational tools that deepen our intuitions and allow us to extract general principles from high-dimensional datasets; and, ultimately, behavioral analyses to reveal what the mouse is smelling (or thinks its smelling!).

Decoding cortical odor representations:

Odors activate ensembles of neurons distributed across piriform cortex neurons. Somehow, their concerted activity allows us to identify thousands of different odors. However, odor identity is only one of the relevant features of an odor stimulus. In addition to identity, information about odor intensity and odor valence must also be represented. Ideally, these representations should be independent or non-interfering so, for example that the representation of identity is stable over a large range of odor concentrations. We are actively studying how different features of an odor are encoded in the firing rates of large populations of piriform cortex neurons in awake, behaving mice. Preliminary work suggests that odor identity may be encoded using a spatial or membership code while odor intensity may be represented using spike timing information within the ensemble. In the extreme, this could support a fully multiplexed coding strategy to independently represent these distinct features of the odor.

A major part of our research is not only to identify different types of neural codes used to represent odor information but to also reveal the specific neural circuit operations that generate them. For this we combine our in vivo studies with molecular genetic and pharmacological perturbations of defined elements of the pirfirom circuit in vivo, and with in vitro patch-clamp recordings that allow us to determine the specific wiring diagrams for the circuits that implement these transformations.

Moving forward, we intend to examine how different classes of piriform neurons represent odor information. Inputs from bulb to piriform are, as far as we can tell, random. However, we are beginning to resolve that individual cells exhibit specific and functionally distinct types of odor responses. We intend to determine whether different classes of piriform neurons –  defined genetically, morphologically, by their local connectivity, or by their specific target projections – differentially encode different features of an odor stimulus (e.g. identity vs. intensity vs. valence). Crucially, we will combine in vivo and in vitro physiology and computational modeling to determine how this information is extracted and represented from seemingly undifferentiated input. 

 

Neural substrates of odor learning and memory:

Individual piriform neurons are interconnected through extensive recurrent collateral connections. It has long been thought that a major role of this recurrent circuit is to specifically interconnect piriform neurons that routinely respond to similar odorants, forming multiple odor-specific assemblies of interconnected cells, called cell assemblies. In this way, partial activation of a subset of the assembly, as might happen in a turbulent environment, could reactivate the entire assembly, generate a stable representation, and therefore allow the recognition of a familiar odor. This process, often called "pattern completion," is thought to occur in multiple circuits thorughout the brain and to underlie learning and memory. However compelling, this theory remains untested.

Powerful new tools that allow us to label and drive gene expression in transiently activated subsets of piriform cortex neurons now provide us with the ability to probe and mechanistically dissect the process of 'pattern completion' directly. By labeling and controlling the activity of odor-responsive neurons, we are examining how these cells participate in different odor ensembles, and we are determining the role of coordinated spiking within the ensemble in stabilizing odor representations. Similarly, we are using behavioral assays to determine the conditions in which we can create fictive odor memories, or change the identity of a "familiar" odor. Moreover, because optogenetics allows us to selectively recreate odor-responsive neurons in vitro, we are beginning to hone in on the specific changes in cellular and synaptic properties that underlie the formation of odor-specific cell assemblies.