All in a sniff: olfaction as a model for active sensing

Abstract

Sensation is an active process involving the sampling and central processing of external stimuli selectively in space and time. Olfaction in particular depends strongly on active sensing due to the fact that-at least in mammals-inhalation of air into the nasal cavity is required for odor detection. This seemingly simple first step in odor sensation profoundly shapes nearly all aspects of olfactory system function, from the distribution of odorant receptors to the functional organization of central processing to the perception of odors. The dependence of olfaction on inhalation also allows for profound modulation of olfactory processing by changes in odor sampling strategies in coordination with attentional state and sensory demands. This review discusses the role of active sensing in shaping olfactory system function at multiple levels and draws parallels with other sensory modalities to highlight the importance of an active sensing perspective in understanding how sensory systems work in the behaving animal.

Figure 1. Sniffing as a dynamic, precisely-controlled and context-dependent behavior

A. Sample of respiratory signals (measured as intranasal pressure) taken from an awake, freely-moving rat. Arrow indicates direction of inspiration. Shaded areas demarcate basal respiration (left) and a bout of sustained exploratory sniffing (right). Numerous other sniffing patterns involving changes in frequency, amplitude and duration of individual ‘sniffs’ - as well as pauses in breathing - can be seen within this ~ 20 second span. B. Percent time spent sniffing at a given frequency for stationary rats, active rats and active mice. While sniffing behavior is highly variable and dynamic, active rodents spend the majority of time sniffing at frequencies above 4 Hz. C. Sniff timing can be precisely controlled. Histogram shows inhalations measured in a head-fixed rat performing a two-odor discrimination task; trace at top shows intranasal pressure from a single trial. This rat displays a brief bout of high-frequency sniffing occurring just after odorant onset, which is repeated reliably over hundreds of trials. Modified with permission from (Wesson et al., 2009). D. Animals modulate sniffing independent of any olfactory context. Plot shows increase in respiratory frequency evoked by self-administered electrical stimulation of the lateral hypothalamus in an unrestrained rat. Sniff frequency increases just prior to the rat beginning self-stimulation and persists throughout the stimulation period. Modified with permission from (Clarke, 1971). E. Histogram of inhalations relative to time of tone presentation; same paradigm as in (C), but measured in a rat performing an auditory discrimination task. This rat displays a brief bout of high-frequency sniffing just after tone onset. Modified with permission from (Wesson et al., 2009).

Figure 2. Inhalation controls the dynamics of odorant-evoked activity throughout the olfactory system

A. Inhalation is required to activate sensory inputs to the olfactory system. Traces show activation of olfactory receptor neuron (ORN) inputs to the olfactory bulb (OB), averaged across multiple trials and aligned relative to odorant presentation (top) or relative to the time of the first sniff after odor presentation. For the latter, odorant is present before the dashed line. Responses occur only within a fixed time after inhalation. Shaded region around each trace indicates the variance across trials. Modified with permission from (Wesson et al., 2008a). B. ORN input signals (bottom) and sniffing (top, gray) measured during odor presentation in an awake rat. Each sniff elicits a transient burst of ORN activity. Modified with permission from (Verhagen et al., 2007). C. Current clamp recording from a mitral cell in an anesthetized, freely-breathing mouse. The cell shows oscillations in membrane potential synchronized with respiration (top, gray); odorant presentation elicits spike bursts after each inhalation. Modified with permission from (Schaefer et al., 2006). D. Voltage clamp recording from a piriform cortex neuron in an anesthetized rat. Each inhalation (gray) in the presence of odor elicits transient excitatory as well as stronger inhibitory currents. Same time-scale as (B). Modified with permission from (Poo and Isaacson, 2009).

Figure 3. Dynamics of inhalation-driven olfactory inputs may set limits for the temporal integration of odor information

A. Responses of ORN inputs evoked by an inhalation of odorant to five different glomeruli, imaged in an awake rat. Traces are ‘sniff-triggered averages’ of presynaptic calcium signals (shaded areas indicate variance around the mean response). Vertical lines below trace indicate time to half-max for input to each glomerulus. Input to different glomeruli occurs with different latencies and rise-times, over a range of ~200 msec. Modified with permission from (Carey et al, 2009). The time-scale is the same in all panels (A – D). B. Spike histogram showing inhalation-driven response across a population of 37 MT cells recorded during sniff ‘playback’ in an anesthetized rat; the rise-time of the population response is ~ 150 msec. Modified with permission from (Carey and Wachowiak, 2011). C. Postsynaptic currents recorded from two neuron types in a piriform cortex slice preparation evoked by a train of stimuli (5 pulses at 20 Hz) delivered to the lateral olfactory tract, which contains the axons of MT cells. Recordings are from an inhibitory layer I interneuron (L1 INT, top trace) and a pyramidal neuron (PYR, lower trace). Blue dashed trace (bottom) shows the envelope of the in vivo MT cell population response evoked by inhalation from (B). MT inputs evoked over a time-course matching the inhalation-evoked response cause depression in the interneuron and facilitation in the pyramidal neuron. Modified with permission from (Stokes and Isaacson, 2010). D. Schematic illustrating differences in odor discrimination times for mice performing a two-odor discrimination that is ‘easy’ (pair of dissimilar monomolecular odorants) or ‘difficult’ (pair of highly similar odor mixtures). Mice take 100 – 200 msec longer to perform the ‘difficult’ discrimination, similar to the amount of time needed for inhalation-driven odor response patterns to develop fully. Modified with permission from (Abraham et al., 2004).

Figure 4. Adaptive filtering of sensory inputs controlled by sniffing

A. Rapid attenuation of receptor neuron activation during sustained high-frequency sniffing of an odorant. Top traces show sniffing in a head-fixed rat, lower traces show ORN responses (estimated from presynaptic calcium signals using temporal deconvolution). Responses are reliable across inhalations when sampled at low frequency (left), but attenuate rapidly during high-frequency sniffing of the same odorant. Modified with permission from (Verhagen et al., 2007). B. Schematic illustrating effects of high-frequency sniffing on odor representations in a changing odor landscape. Top, representative sniffing pattern including a bout of high-frequency sniffing. Solid lines indicate preence of odorA, odor B or both A and B. Lower graphics (‘OB map’) represent patterns of activation across five glomeruli in the olfactory bulb at different times during the sniff bout. When sampled in isolation, A and B have overlapping representations (1 – 2). High-frequency sniffing of odor A causes an attenuation of the response map (3 – 4); when odor B is encountered against the background of odor A, only those glomeruli that differ from those activated by odor A are strongly activated, so that the representation resembles the difference between the two odor maps (5). A return to low-frequency sniffing removes the attenuation and the representation changes to resemble the sum of the two maps (6 – 7). Thus, changes in the odor landscape are represented in a subtractive or additive mode depending on sniff frequency.

Figure 5. Potential effects of sniff frequency on odor processing in the olfactory bulb

A. Principal cell types and canonical circuits of the olfactory bulb. Excitatory elements are ORN, olfactory receptor neurons (ORN), external tufted cells (ET), and mitral cells (MC). Inhibitory elements are periglomerular cells (PG), short axon cells (SA) and granule cells (GC). ET cells also provide excitatory input to MT cells (not shown). Schematic taken from (Wachowiak and Shipley, 2006). B. Schematics illustrating predicted effects of increasing sniff frequency on the strength and coherence of activity in each cell type. Coherence reflects time-locking to sniffing and synchrony with other neurons of the same type. Experimental data supporting each prediction are cited in the text. The ultimate effect on MT cell output depends on emerging details of the OB circuitry. C. Increasing input frequency leads to an increase in MT cell spike output in an OB slice preparation. (i) Voltage recordings from a MT cell showing responses evoked by four brief current injections simulating synaptic inputs (arrows) at 1 and 2.5 Hz. (ii) Plot of total spike output among MT cells as a function of input frequency. Precision of spike timing also increases (not shown). Modified with permission from (Balu et al., 2004). D. Increasing sniff frequency shortens inhalation-evoked MT spike bursts in vivo. (i) Spike histogram of a MT cell during odorant sampling controlled by playback of sniffing recorded previously in an awake rat. This cell responds phasically to each sniff during both low-frequency respiration and a high-frequency sniff bout. Contrast with ORN responses in Figure 4A. (ii) Plots of inhalation-triggered spiking patterns for an MT cell evoked by inhalations repeated at 1 – 5 Hz. The cell shows a shortening of its spike burst, along with a modest decrease in peak firing rate and increase in response latency, all of which are consistent with increased inhibition. Modified with permission from (Carey and Wachowiak, 2011).

Figure 6. Sensorimotor integration during odor sensing

A. Humans show rapid modulation of sniff magnitude as a function of odorant concentration. Blue and purple traces show mean intranasal flow rate for low and high concentrations of an odorant; black trace shows the p-value of the difference between the flow rates at each concentration. Sniff flow rates diverge and remain different at 160 msec after inhalation onset. Modified with permission from (Johnson et al., 2003). B. Rats show similarly rapid changes in sniffing in response to a novel odorant. Red and blue plots show cumulative inhalation count over time during presentations of novel versus familiar (‘learned’) odorants. Black plot shows p-value of the difference between novel and learned trials. Sniff counts diverges significantly at 140 msec after the first inhalation. Modified with permission from (Wesson et al., 2008a). C. Potential substrates for cortical control of sniffing behavior. Electrical stimulation of infralimbic cortex in an anesthetized rat elicits respiratory changes that closely resemble exploratory sniffing (compare with Figure 4A). Modified with permission from Aleksandrov et al. (2007). D. Schematic illustrating the potential neural pathways involved in attentional modulation and sensorimotor integration during sniffing. Gray paths indicate the movement of odorant molecules through the nasal cavity; blue paths indicate sensory afferent pathways; red paths indicate centrifugal pathways targeting OB and PC. Dashed lines indicate more speculative connections. See text for references. IL, Infralimbic cortex; Ins, Insular cortex; LC, locus coeruleus; RN, Raphe nuclei; Cb, cerebellum; NTS, nucleus of the solitary tract. Brain template courtesy of A. Puche, Univ. of Maryland.