Trends in Neurosciences
ReviewNeurons and circuits for odor processing in the piriform cortex
Introduction
The primary senses have long been used as portals into the workings of the brain, a strategy that has facilitated major advances in our understanding of how information is processed by neural circuits to form a coherent picture of the outside world. The olfactory system has been less prominent in this effort than other sensory modalities, perhaps in part because the sense of smell is less important to humans. However, olfaction offers significant advantages for exploring the basic science of sensory processing. For instance, the olfactory system is anatomically shallow and remarkably stereotyped across different species [1], suggesting that it is both tractable to study and likely to reveal fundamental principles about optimal coding strategies that have persisted through evolution. However, olfaction has a number of features that make it uniquely challenging: odor space is multi-dimensional and poorly defined; odor objects (e.g., the zest of lemon and the stench of sewage) are complex syntheses of many chemical components; and the sense of smell is densely interwoven with memories and emotion 2, 3.
Here, we review recent developments in just one area of olfaction, the cellular physiology of the piriform cortex (PC) of mammals. The PC is the largest cortical region that receives direct synaptic input from the olfactory bulb (OB), which in turn receives direct input from the olfactory epithelium at the back of the nose. Hence, the PC is only two synapses removed from the outside world and, uniquely for a sensory cortex, does not receive its sensory input via the thalamus. Much classic work has been done on the PC 1, 3, but more recent research on mammalian olfaction has tended to focus on the epithelium and OB. Now, with a growing understanding of its inputs, fresh attention is being directed to the PC. There have been several excellent reviews of the PC in recent years, although these mainly focus on its higher-level functions 2, 3. Here we take a more reductionist slant and specifically review recent papers on the neuronal hardware – the cells and circuits – in which the processing functions of the PC are implemented.
Section snippets
Basic architecture of the PC
The PC is a trilaminar paleocortex located (in rodents) on the ventrolateral surface of the brain close to the lateral olfactory tract (LOT), which is a myelinated fiber tract conveying output from the OB (Figure 1A). In brief, the PC comprises a sparsely populated superficial layer (layer 1), a main input layer (2) containing the densely packed somata of glutamate-releasing principal neurons, and a deep layer (3) containing principal neurons at lower density (Figure 1B). The input fibers of
What the OB tells the PC
A potential benefit of studying the PC is that its main input, the OB, is increasingly understood. The broad picture of bulbar structure and function is well established 26, 27. Activation of dispersed classes of receptor neurons in the olfactory epithelium is transformed into a punctate map of excited glomeruli in the OB, the odotopic map (Figure 2A). The outputs of the several dozen mitral and tufted cells forming each glomerulus are further refined by local interneuron circuits. Feedback
OB → PC transformation of odor representations
As noted above, there is a remarkable transformation from an odotopic map in the OB to a distributed representation in the PC (Figure 2A). This transformation presumably allows the PC to perceive a complex odor mixture as a unique odor object distinct from its components [19]. How is this remapping achieved? One aim of neurophysiological studies of the PC is to answer this question in terms of underlying circuits. First, however, we set the scene by mentioning several recent papers that report
Afferent circuits
Afferent inputs from the OB to the PC are anatomically diffuse 34, 35, 36, but these findings give no information about the identity of targeted cells in the PC or the functional properties of the connections. Recent patch-clamp studies have sought to address these issues, but they have reached different conclusions in some cases.
When applying minimal extracellular stimulation in PC slices, a substantial number of layer 2/3 principal cells received strong single-fiber connections from the OB
Associational circuits
It has long been thought that the profuse associational connections in the PC may lie at the heart of its computational power [22]. As well as being abundant, associational connections are electrotonically closer to the soma (and hence to the spike initiation zone) and are more plastic and more affected by neuromodulators 74, 75. Thus, associational fibers seem better equipped than afferent fibers for implementing complex olfactory processing (while keeping in mind, of course, that the whole PC
Inhibitory circuits
Synaptic inhibition is ubiquitous in the cortex [81]. Excitation and inhibition typically act together in a balanced way to maintain sparse firing, which may have computational and energetic advantages [82]. Although the roles of particular interneuron classes may be uncertain, it is generally thought that two types of canonical inhibitory circuit predominate in the cortex: feedforward inhibition and feedback inhibition (Figure 1D) [81].
Until recently, information about inhibitory neurons in
Plasticity
Olfaction is a highly plastic sense [97]. The apparently random connectivity from the OB to the PC immediately suggests that the representation of odors in the PC is not hard-wired but must be learned from experience. Indeed, the PC is in some ways an archetypal associative memory device [25]. Inevitably, there are complications. For example, different plasticity-related functions seem to be partitioned into different parts of the PC (aPC vs pPC), and important kinds of olfactory plasticity
Coding
Ultimately we seek to understand how information is encoded in the brain. Despite the complexities touched on above, the PC is an interesting subject for studying coding because it seems to be a compact and tractable circuit for implementing combinatorial representations that are robust to degradation, background, and natural variations in stimuli [47]. We are still very far from articulating a bottom-up neurophysiological theory of how this encoding is achieved in mammals. Nevertheless, some
Conclusions
Olfaction has long been regarded as a mysterious sense, tasked with decoding a complex olfactory world of hard-to-describe smells. Some of this mystery has been laid to rest by new paradigms built on receptor genes and odotopic maps. However, the diffuseness of the olfactory representation at higher levels in the brain remains a puzzle (Box 1). Cellular neurophysiology is establishing some ground rules for the mechanics of this higher-level olfactory processing; for example, recent work is
Acknowledgments
Our research is supported by Project Grants 585462, 1009382 and 1050832 from the National Health and Medical Research Council of Australia.
References (106)
- et al.
Cortical processing of odor objects
Neuron
(2011) Olfaction as a model system for the neurobiology of mammalian short-term habituation
Neurobiol. Learn. Mem.
(2009)Dissociable codes of odor quality and odorant structure in human piriform cortex
Neuron
(2006)Olfactory predictive codes and stimulus templates in piriform cortex
Neuron
(2011)Cortical feedback control of olfactory bulb circuits
Neuron
(2012)Functional properties of cortical feedback projections to the olfactory bulb
Neuron
(2012)From the top down: flexible reading of a fragmented odor map
Trends Neurosci.
(2009)Olfactory perception: receptors, cells, and circuits
Cell
(2009)Early transformations in odor representation
Trends Neurosci.
(2010)- et al.
Slow-waves in the olfactory system: an olfactory perspective on cortical rhythms
Trends Neurosci.
(2006)