Vibrissal location coding

• Prof. Ehud Ahissar, Department of Neurobiology, The Weizmann Institute, Rehovot

• Dr. Per M Knutsen, UC San Diego, La Jolla, CA, USA

Figure 1: Whisking behavior of freely-moving rats during two different tasks. Left and right C2 whiskers denoted in red and green, respectively. Top: Rhythmic whisking during whisking in air. Bottom: Whisking during an object localization task. Contact periods denoted by thick lines.

Figure 2: Object location in spherical coordinates. Elevation, ϕ; azimuth, θ; radial distance, r;. Blue circle is object.

Figure 3: Spatial coding. Two whiskers (A, B) scan regions of space that are determined by their positions on the whisker pad, lengths and movement patterns. An object positioned at different coordinates (circles) can be contacted (filled circles) if it is inside the field scanned by a whisker. Individual whiskers may contact an object at many locations and contact by itself does not encode location. Combining contact across whiskers (bottom panels) improves accuracy. The intersect A+B outlines the approximate location of an object if both whiskers contact. The difference B–A outlines the location if whisker B, but not A, contacts.

Figure 4: Classes of trigeminal ganglion (TG) cells that respond to movement and contact events.

Vibrissal location coding refers to the ways by which the location of external objects is coded (represented) in the vibrissal system of rodents. The vibrissal system contains the vibrissae (whiskers) and the follicles, neurons and muscles associated with them. Coding is traditionally sub-categorized to encoding, i.e., coding at the whisker-object interaction phase, and recoding, i.e., coding at processing stages that are remote from this direct interaction. The vibrissal system is an active-sensing system – the system acquires information about objects in its environment by moving its whiskers (“whisking”, see Vibrissal behavior and function, Whisking kinematics) and interpreting the resulting sensations (Figure 1). In recent years, this system has attracted interest from researchers that study the emergence of perception from motor-sensory interactions and from engineers who regard whisking as a useful model system for developing robotic touch platforms, such as whiskered robots. This article reviews recent progress and our current understanding of vibrissal object location coding in rodents.

Coordinates of object location

Object location can be specified in Cartesian head-centered coordinates by three spatially-orthogonal axes; rostro-caudal, dorso-ventral and medio-lateral. These axes are often also referred to as the horizontal, vertical and radial axes, respectively. Spherical object coordinates are specified in terms of azimuth, elevation and radial distance relative to a plane intersecting the eyes and nose ( Figure 2).

Possible coding dimensions

The location of an object could in principle be specified by information contained in four different encoding domains:

Space: Spatial encoding refers to the spatial distribution of activated sensory neurons. For instance, an encoding scheme whereby individual neurons are exclusively activated when objects are located at specific locations would constitute coding by a spatial parameter (neuron identity). An example of such encoding is retinotopic representation of object location. In the vibrissal system, individual primary sensory neurons have single-whisker receptive fields. Thus, these neurons encode the identity of contacting whiskers. The identities of contacting whiskers are, however, ambiguous with respect to the location of an object. Figure 3 illustrates how an object positioned at different coordinates may result in the same combination of contacting whiskers. By increasing the number of contacting whiskers the location of the object can be ascertained with higher accuracy.

Intensity: In addition to being activated by contact, primary sensory neurons are also tuned to contact parameters such as force, velocity and direction. During active sensing, the intensity of receptor activation depends on the location of the contacted object, through the dependency on object location of the rotational and translational forces applied on the follicle (Birdwell et al., 2007). The exact mapping of spatial coordinates onto different levels of neural activity depends on the details of the interactions between the whisker and the object.

Time: The whiskers provide a very sparse (~1/1000) snapshot of the environment. This is because the whiskers are thin (microns) and relatively widely spaced apart (millimeters), thus introducing ‘blind spots’ between whiskers that limit the resolution of spatial and intensity coding. During whisking, however, the entire spatial continuum reached by the fanned-out array of whiskers is scanned through time. Thus, space and location can additionally be encoded by the timing of contact events. Sensory neurons along the trigeminal pathway are sensitive to the timing of whisker deflections, accurately encoding events with sub-millisecond precision (Arabzadeh et al., 2005, Jones et al., 2004). During whisking, temporally encoded tactile events can therefore be encoded with higher angular precision than spatially encoded events.

Posture: Posture: Crucially, temporal encoding requires temporal reference signals, such as proprioceptive or other motor signals derived from corollary discharges or re-afference. Proprioceptive signals can also code object location directly when contact is controlled; for example, the angle of the whisker encodes the azimuth of the object in head-related coordinates when contact is controlled to be of minimal impingement. The postural dimension also includes morphological coding by which object location is coded by relationships between morphological variables of the whisker at any given moment of interaction with the object.

Neural encoding of object location by primary afferents

In awake animals, the behaviors generating sensory inputs can be highly dynamic, reflecting continuous updating of motor output in response to sensory inputs. This makes systematic investigation of sensory signals difficult, prompting the development of anesthetized preparations where naturalistic whisking patterns can be mimicked by electrical stimulation of central motor regions or nerves supplying the mystacial musculature (referred to as electrical, artificial or fictive whisking).

During artificial whisking, primary afferents of the trigeminal ganglion (TG; about 200-400 neurons per whisker) encode tactile events and object location (Szwed et al., 2003, Szwed et al., 2006). Based on the type(s) of event responded to, four principal classes of primary afferents have been characterized ( Figure 4). “Contact” cells respond briefly and with short latency when a whisker contacts an object. “Pressure” cells respond throughout contact. “Detach” cells respond when a whisker detaches from an object. “Whisking” cells respond only to whisker movement. “Contact”, “Pressure” and “Detach” cells are referred to as “Touch” cells. Additionally, a class of cells referred to as “Whisking/Touch” respond during whisking and increase their firing upon and during touch.

These experiments demonstrate that together, neurons of all these classes encode the 3D location of an object during whisking. This coding exhibits an orthogonal scheme, in which each spatial dimension is coded by an independent neuronal variable. “Contact” cells use a temporal code to encode the horizontal coordinate (azimuth) of an object. These cells are activated upon contact, and therefore encode the protraction angle at which the object was encountered by the timing of the first evoked spike. In order for this latency code to be correctly decoded, read-out circuits must compare the time of contact with a reference signal indicating whisker position. This reference signal is provided by “Whisking” cells, whose activity is locked to specific phases of the movement cycle. The radial coordinate is primarily encoded by an intensity code. Upon contact, “Touch” cells respond with spiking rates of varying intensity that depends on the radial location of the object. Typically, evoked responses drop in intensity with increasing radial distance. In some cases, this intensity coding is reduced to a binary code with some “Touch” cells responding only when an object is positioned very close to the whisker pad. Not every “Touch” cell, however, encodes radial position in a monotonic manner (Szwed et al., 2006). Thus, reliable decoding by read-out circuits should pool the activity of many “Touch” cells. Encoding of vertical object location is determined by anatomy. Because primary afferents have single-whisker receptive fields and because whiskers move along the axis of rows (Bermejo et al., 2002, Knutsen et al., 2008), any afferent is therefore activated only when an object is present at the elevation of its receptive whisker. Thus, the vertical coordinate of contact is encoded by neuron identity.

Behavioral aspects of object localization

Figure 5: Head and whisker movements during an object localization task (described in Knutsen et al., 2006). This rat was trained to detect which of the two vertical poles was closer to the home cage, using a single whisker on each side.

Behavioral studies of object localization in rats have confirmed the contribution of temporal, intensity and spatial information in object localization, and are consistent with the orthogonal scheme revealed using artificial whisking (Ahissar and Knutsen, 2008). Behavioral studies have isolated the optimal behavior for each spatial dimension by maintaining object location constant in two and varying location only along one dimension. The selective contribution of spatial information has been probed by removing sub-sets of whiskers, and the contribution of temporal cues facilitated by training animals to make relative comparisons that exceed the acuity limits of spatial coding ( Figure 5; Krupa et al., 2001, Shuler et al., 2002, Knutsen et al., 2006, Mehta et al., 2007).

The ability of rats to localize objects along the horizontal direction depends on both prior training conditions and kinematics of whisker movements. Two independent studies of horizontal object localization in rats (Knutsen et al., 2006, Mehta et al., 2007) agree on the following; 1) Rats can accurately localize with a single whisker on each side of the face, provided they first learn the same task using many whiskers, 2) the identity and number of whiskers contacting the objects does not determine horizontal acuity, 3) whisker movements are required to localize, 4) the energy of whisking correlates with acuity, and 5) relative localization (between co-existing objects) is more accurate then absolute (memory-guided) localization. Head-fixed studies have also confirmed that mice can localize along the horizontal dimension with a single whisker (O'Connor et al., 2010). Single-whisker localization precludes any dependency on spatial encoding of horizontal object location.

Horizontal acuity during relative localization (as fine as ~1° when comparing positions of two co-existing objects) is an order of magnitude better than the limit imposed by the spacing of adjacent same-row whiskers (~20°), a performance level referred to as vibrissal hyperacuity (Knutsen et al., 2006). During absolute (memory-guided) localization the behavioral resolution (~15°) is closer to the whisker spacing limit. Vibrissal hyperacuity is also achieved by rats with just one whisker remaining on each mystacial pad (i.e. one whisker contacting each object). This manipulation excludes the possibility that the identity of contacting whiskers is compared (see Figure 3). Instead, other kinematic variables afforded by whisker movements, such as the relative angles or contact-times, may be important for horizontal encoding. Horizontal location may also be encoded in part by a labeled-line code (e.g., by a set of neurons, each of which fires if and only if contact occurs at a certain horizontal coordinate) due to interactions between torsional whisker rotation and directional tuning of afferents (Knutsen et al., 2008).

Behavior during localization of objects along the radial axis differs from that during horizontal localization. Whisking is not always required during radial localization. Instead, the whiskers can be brought in contact with objects through slow adjustments in whisker set-point and head/body movements. In one paradigm of radial localization, radial acuity correlates with the number (but not the identities) of intact whiskers available to the rat, and has not been shown to improve with the presence of a reference object (Krupa et al., 2001; Shuler et al., 2002). The following conclusions about radial encoding have been inferred from these available observations. In freely moving rodents a simple labeled-line code can probably be ruled out since whisker identities used to localize do not determine performance (note that head fixed rodents may exhibit a different strategy (Pammer et al., 2009)). Suppression of whisking in freely moving rodents suggests that temporal cues are not important for radial localization. Rather, the observation that radial acuity falls proportionally to the number of removed whiskers, suggests that radial object location is encoded by a sensory cue accumulated across all whiskers, such as the firing intensity of populations of Touch neurons (see above).

Thus far, no behavioral test of vertical object localization has been reported.

Orthogonal coding of object location

Figure 6: Proposed orthogonal encoding scheme of object location. During exploration, whisker movements are mainly along the horizontal plane. Upon contact with an object (blue dot) the timing of the contact response (latency to spikes) encodes the horizontal dimension (red). The vertical dimension (blue) is encoded by the identity of contacting whiskers. The radial dimension (green) is encoded by the intensity of activation (e.g. rate of evoked spikes) due to bending and mechanical forces acting upon whisker shaft. These three codes for location are orthogonal and the spatial dimensions can thus be encoded independently of each other.

Behavioral studies of object localization agree with the observations that primary sensory afferents encode object coordinates by orthogonal neuronal codes. Primary afferents have been shown to be highly temporally tuned and encode horizontal object location on an individual basis, consistent with behavioral observations that the highest acuity can be achieved with a single whisker contacting an object. Radial location is poorly encoded by individual afferents and requires the pooled activity of multiple afferents to reach the precision exhibited by behaving animals. Consistent with this, behaving animals also need multiple whiskers to localize objects along the radial axis. These are consistent with a temporal and a population rate code for the horizontal and radial coordinates, respectively. No behavioral support has so far been found for the labeled-line spatial code of vertical coordinates ( Figure 6).

Orthogonal coding appears to be an efficient scheme for object localization as during natural behavior object location is encoded for all spatial dimensions simultaneously (Knutsen and Ahissar, 2009). By relying on an independent neuronal variable for each dimension, 3D location can be read out from the same afferents using orthogonal decoding circuits in parallel.

Morphological coding of object location

Figure 7: Morphological coding by whisker angle and curvature. Trajectories of whiskers in the θp - K phase plain (θp: push angle, maximal change in whisker angle from contact onset during contact; K: curvature at whisker’s base) are shown. Neither of the two variables provide unambiguous coding of object location by itself; for example, K around .02 mm-1 (at max protraction) codes for both ~[p2, 60%] and ~[p1, 90%]. In contrast, a motor-sensory contingency, between the motor (θp) and sensory (K) parameters, provides unique coding of both azimuthal and radial coordinates (Bagdasarian et al., 2013).

Object location in the horizontal plane, spanned by the azimuthal and radial coordinates, is also encoded by phase planes of whisker-related morphological variables (Bagdasarian et al., 2013). Phase planes spanned by angular and curvature variables encode object location reliably already 15 ms after object contact onset, and reliability increases as long as contact continues. Coding is based on motor-sensory contingencies rather on sensory cues alone (Figure 7). Whisker rigidity allows direct transformation of morphological coding to mechanical coding within the follicle (Bagdasarian et al., 2013;Quist and Hartmann, 2012).

Recoding

Eventually, proposed encoding schemes of object location must also take into account realistic read-out circuits that could possibly recode each variable in some common internal language. Such recoding is sometimes termed "decoding", although unlike engineered devices, no restoration of the original encoding signals is attempted in the brain. Recoding of temporal information requires an additional reference signal that signals whisker position. Such a reference signal can be generated internally as a corollary discharge, by proprioceptors or by mechanoreceptors (re-afference) responsive to whisker motion (e.g. as signaled by "Whisking" cells, Szwed et al., 2003). The comparison between Touch and reference signals could be implemented by phase-locked loops or phase detectors. Re-afferent signaling of whisker motion has been observed in both the somatosensory thalamus (Yu et al., 2006) and cortex (Fee et al., 1997; Brecht et al., 2006; Crochet and Petersen, 2006; Derdikman et al., 2006; Curtis and Kleinfeld, 2009). Evidence for recoding the horizontal coordinate of object location in firing rates, via phase detection in the thalamocortical network, was recently demonstrated (Curtis and Kleinfeld, 2009;Yu et al., 2013). Intensity coding of radial position is likely based on interpolating firing intensity across multiple whiskers. This could be implemented by neural integrators, peak detectors, attractor neural networks or synfire chains. The vertical coordinate, encoded by labeled-lines, can be read out by threshold detectors. Recoding may involve iterative processes within thalamocortical loops (Yu et al., 2013;Edelman, 1993) as well as across motor-sensory-motor loops involving multiple object contacts (Knutsen et al., 2006;Saig et al., 2012). During such iterative processes, various coding schemes may be used in parallel or in sequence (Horev et al., 2011).

Role of motor strategies

Motor control of head, pad and whiskers is crucial for the consistency, reliability and resolution of each of the localization codes: pad orientation, whisking velocity and contact force will directly affect the mapping of the spatial coordinates into neuronal variables. The closed-loop architecture of the vibrissal system allows efficient adaptive control of vibrissal touch, though it remains to be seen to what extent adaptive behavior is indeed manifested during perceptual tasks, which motor variables are controlled in which contexts, and how they are controlled.

A perceptual coding scheme, unlike a sensory coding scheme, must include motor variables in its definition. This is made clear by the fact that the sensory cues that serve perceptual comparisons are not necessarily the same cues that convey primary sensory data, where the latter refers to the manner of receptor activation. The former depend on motor strategies, which involve multiple levels of sensory-motor control loops encompassing body, head and whisker movements, while the latter depend only on the peripheral level of vibrissal interaction with the external world. For example, one rat may whisk synchronously on both sides and use the time difference between left and right contacts as a primary cue for discrimination (Knutsen et al., 2006). Another rat may instead aim to contact (rather than move) synchronously and use the bilateral angular difference as a primary cue. In these sensory-motor processes, both whisker angle and contact times (the sensory data) are crucial, but each serves as a perceptual cue in only one of the strategies, and as a sensory-motor coordinator in the other.

References

• Ahissar(2008). Object localization with whiskers. Biological Cybernetics 98(6): 449-458. doi:10.1007/s00422-008-0214-4.

• Arabzadeh, E; Zorzin, E and Diamond, M E (2005). Neuronal encoding of texture in the whisker sensory pathway. PLoS Biology 3(1): 155-165. doi:10.1371/journal.pbio.0030017.

• Bagdasarian, K et al. (2013). Pre-neuronal morphological processing of object location by individual whiskers. Nature Neuroscience 16: 622-631. doi:10.1038/nn.3378.

• Bermejo, R; Vyas, A and Zeigler, H P (2002). Two-dimensional monitoring of whisker movements. Somatosensory & Motor Research 19(4): 341-346. doi:10.1080/0899022021000037809.

• Birdwell, J A et al. (2007). Biomechanical models for radial distance determination by the rat vibrissal system. Journal of Neurophysiology 98(4): 2439-2455. doi:10.1152/jn.00707.2006.

• Brecht, M; Grinevich, T E; Jin, T E; Margrie, T and Osten, P (2006). Cellular mechanisms of motor control in the vibrissal system. European Journal of Neuroscience 453: 269-281. doi:10.1007/s00424-006-0101-6.

• Curtis(2009). Phase-to-rate transformations encode touch in cortical neurons of a scanning sensorimotor system. Nature Neuroscience 12: 492-501. doi:10.1038/nn.2283.

• Crochet(2006). Correlating whisker behavior with membrane potential in barrel cortex of awake mice. Nature Neuroscience 9(5): 608-610. doi:10.1038/nn1690.

• Derdikman, D; Haidarliu, S; Bagdasarian, K; Arieli, A and Ahissar, E (2006). Layer-specific touch-dependent facilitation and depression in the somatosensory cortex during active whisking. The Journal of Neuroscience 26(37): 9538-9547. doi:10.1523/jneurosci.0918-06.2006.

• Edelman, G M (1993). Neural Darwinism: Selection and reentrant signaling in higher brain function. Neuron 10: 115-125.

• Fee, M A; Mitra, P P and Kleinfeld, D (1997). Central versus peripheral determinants of patterned spike activity in rat vibrissa cortex during whisking. Journal of Neurophysiology 78(2): 1144-1149.

• Horev, G et al. (2011). Motor-sensory convergence in object localization: A comparative study in rats and humans. Philosophical Transactions of the Royal Society of London Series B, Biological Sciences 366(1581): 3070-3076. doi:10.1523/JNEUROSCI.2432-12.2012.

• Jones, L M; Depireux, D A; Simons, D J and Keller, A (2004). Robust temporal coding in the trigeminal system. Science 304: 1986-1989. doi:10.1126/science.1097779.

• Knutsen(2009). Orthogonal coding of object location. Trends in Neuroscience 32: 101-109. doi:10.1016/j.tins.2008.10.002.

• Knutsen, P M; Pietr, M and Ahissar, E (2006). Haptic object localization in the vibrissal system: behavior and performance. The Journal of Neuroscience 26(33): 8451-8464. doi:10.1523/jneurosci.1516-06.2006.

• Knutsen, P M; Biess, A and Ahissar, E (2008). Vibrissal kinematics in 3D: Tight coupling of azimuth, elevation, and torsion across different whisking modes. Neuron 59(1): 35-42. doi:10.1016/j.neuron.2008.05.013.

• Mehta, S M; Whitmer, D; Figueroa, R; Williams, B A and Kleinfeld, D (2007). Active spatial perception in the vibrissa scanning sensorimotor system. PLoS Biology 5(2): 309-322. doi:10.1371/journal.pbio.0050015.

• Quist(2012). Mechanical signals at the base of a rat vibrissa: The effect of intrinsic vibrissa curvature and implications for tactile exploration. Journal of Neurophysiology 107: 2298-2312. doi:10.1152/jn.00372.2011.

• Saig, A; Gordon, G; Assa, E; Arieli, E and Ahissar, E (2012). Motor-sensory confluence in tactile perception. The Journal of Neuroscience 32(40): 14022-14032. doi:10.1523/JNEUROSCI.2432-12.2012.

• Szwed, M et al. (2006). Responses of trigeminal ganglion neurons to the radial distance of contact during active vibrissal touch. Journal of Neurophysiology 95(2): 791-802. doi:10.1152/jn.00571.2005.

• Szwed, M; Bagdasarian, K and Ahissar, E (2003). Encoding of vibrissal active touch. Neuron 40(3): 621-630. doi:10.1016/s0896-6273(03)00671-8.

• Krupa, D; Matell, M S; Brisben, A J; Oliveira, L M and Nicolelis, M A L (2001). Behavioral properties of the trigeminal somatosensory system in rats performing whisker-dependent tactile discriminations. The Journal of Neuroscience 21(15): 5752-5763.

• Shuler, M G; Krupa, D and Nicolelis, M A L (2002). Integration of bilateral whisker stimuli in rats: Role of the whisker barrels cortices. Cerebral Cortex 12(1): 86-97. doi:10.1093/cercor/12.1.86.

• O'Connor, D et al. (2010). Vibrissa-based object localization in head-fixed mice. The Journal of Neuroscience 30: 1947-1967. doi:10.1523/jneurosci.3762-09.2010.

• Yu, C; Derdikman, D; Haidarliu, S and Ahissar, E (2006). Parallel thalamic pathways for whisking and touch signals in the rat. PLoS Biology 4: 819-825. doi:10.1371/journal.pbio.0040124.

• Yu, C et al. (2013). Coding of object location in the vibrissal thalamocortical system. Cerebral Cortex Epub: bht241. doi:10.1093/cercor/bht241.

Sponsored by: Prof. Tony J. Prescott, Dept Psychology, Univ of Sheffield, UK

Reviewed by: Anonymous

Accepted on: 2011-08-21 15:39:56 GMT