Neural mechanisms for the localization of unexpected external motion

Animals actively move their sensory organs (e.g., eyes, whiskers) to explore the world. Accurate object localization requires distinguishing expected sensory consequences of one’s own movements from changes caused by external object motion. The superior colliculus (SC) contains egocentric sensorimotor maps and is implicated in spatial behaviors and stabilizing sensory representations during movement. Prior visual studies show SC selectivity for externally generated motion and predictive mechanisms that mitigate self-induced input. Far less is known about SC’s somatosensory function during active touch. The authors hypothesized that SC neurons are specialized to localize external motion during whisker-based active sensing by comparing incoming tactile input to sensorimotor predictions built from recent self-generated history.

• SC supports spatial behaviors (pursuit, escape, orienting) and stabilizes visual input during eye movements; it shows selectivity to externally generated visual motion. - In somatosensation, mouse SC responds to whisker stimulation and exhibits a somatotopic whisker map with large multi-whisker receptive fields, largely established under anesthesia. - Extrasensory and motor signals (e.g., eye position, corollary discharge) modulate SC responses, supporting suppression of self-generated sensory consequences. - Human studies show perceptual attenuation of self-generated touch, attributed to sensorimotor predictions. - These findings suggest general mechanisms for subtracting expected self-induced input across modalities and motivate probing analogous computations in tactile SC during active whisking.

• Subjects: CD-1 mice (both sexes, 9–15 weeks). Headplates implanted; mice trained to run on a circular treadmill. - Whisker-guided virtual reality: A cylindrical tactile surface rotated in closed-loop to match locomotion, creating a tactile flow-field. After 200 cm of running, the surface stochastically translated 1 cm rostrally, caudally, or stayed centered (equal probability; in some mice, an added outside-whisker-field position). Visual/auditory cues were obscured. High-speed IR imaging at 500 fps tracked whisker kinematics; encoders recorded surface and run speed. - Neural recordings: 3-shank, 128-channel silicon probe targeted intermediate and deep layers of lateral SC (~300–1000 µm below SC surface); 12 mice, 873 neurons. Sensory responsiveness assessed with one-way ANOVA and post hoc tests. - GoW/LoW analysis: External translations classified as Gain-of-Whisker (GoW) or Loss-of-Whisker (LoW) when whisker-surface contact was gained or lost. Manual video inspection and kinematic analysis (DeepLabCut) verified contact events, whisker angle and curvature. - Self vs external GoW experiments: (1) Surface moved between rostral space and outside the whisking field to compare external-GoW (after free whisking) with self-GoW (after quiescence-to-whisking transitions), matching touch phase and kinematics. (2) One-whisker, pneumatically controlled object: object inserted rapidly after free whisking to evoke external-GoW, then remained to allow repeated self-GoW bouts; locomotion/whisking tightly controlled. - Controls: Post-trimming controls (whiskers removed) to confirm whisker mediation. Velocity of external motion and number of contacting whiskers assessed for influence on responses. - Repetition/habituation: External-GoW events repeated over minutes; transient and sustained windows defined per mouse to quantify linear changes in firing with repetition; spike waveform stability checked to rule out recording drift. - Decoding: Population SVM classifier (Neural Decoding Toolbox) decoded surface location (rostral/caudal/center) over time using z-scored spike rates (150 ms bins, 20 ms step), 10-fold CV with 50 resamples. - Spike sorting and kinematics: Kilosort and Phy2 for unit isolation; DeepLabCut for whisker tracking (four labels per whisker); whisker angle bandpass-filtered (1–30 Hz) and curvature via Menger curvature. - Statistics: ANOVA, Tukey post hoc, Wilcoxon signed-rank, Mann–Whitney tests; linear regression for repetition slopes; modulation indices computed as difference/sum across pre/post movement windows.

• Broad tactile responsiveness: About two-thirds of recorded SC neurons responded to surface movement (67 ± 6%, 12 mice, 578/873 neurons, α < 0.05). - Transient response tied to contact changes: Transient firing increases emerged when surface motion caused GoW or LoW; pushing without gaining new whisker contact did not evoke transients. Many neurons preferred GoW (55 ± 4%, 8 mice, 385 neurons). - Whisker mediation: Trimming whiskers abolished responses to external motion (responsive neurons: 59 ± 4% pre-trim vs 6 ± 1% post-trim; 156 neurons, 2 mice). - External > self-generated touch: SC neurons strongly preferred external-GoW over self-GoW despite matched touch kinematics (curvature/angle). External-GoW evoked significantly larger responses (p = 6e−13, Wilcoxon, 4 mice, 139 neurons); replicated with stricter control (one-whisker paradigm, p = 2e−11, 2 mice, 74 neurons). Differences were not explained by stimulus strength, rate, or locomotion/whisking state. - Sensorimotor prediction effect: Self-generated stimulus history attenuated responses; external-GoW following free whisking (mismatch) produced larger transients than self-GoW following active touch (match). - Repetition-driven habituation: Transient external-GoW responses linearly decayed with repetitions across minutes (223 neurons, 8 mice, p = 2e−12), whereas sustained self-generated activity was stable (98 neurons, p = 0.64). Inter-event intervals averaged ~67 ± 5 s; spike waveform amplitudes were stable, ruling out low-level adaptation or recording drift. - Encoding of motion direction/location: Population decoding accurately classified surface location, peaking during external motion. External-selective (transient) neurons yielded high accuracy specifically during movement. Decoding was best when translations caused GoW/LoW across locations (GoW/LoW in all movements: 94 ± 5% accuracy, 2 mice; in half: 71 ± 7%, 6 mice; in none: 56 ± 3%, 4 mice). In one dataset, somatotopy aligned with different whiskers engaged by rostral/caudal motion.

The study shows that SC neurons contextualize tactile input with motor-derived predictions, enabling discrimination between externally induced object motion and self-generated touch during active whisking. A rapidly adapting transient response emerges specifically at externally generated gains/losses of whisker contact and is attenuated when touch matches self-generated stimulus history, consistent with predictive sensorimotor attenuation. Over longer timescales, transient responses habituate with repeated external motion, indicating integration of external history. Population activity briefly but accurately encodes the direction/location of external motion, especially when different whiskers are engaged, supporting a role in localizing moving objects. Potential mechanisms include convergent cortical (layer 5 cortico-collicular) and brainstem inputs, rapid loss of excitation or inhibitory filtering (e.g., divisive normalization), and cerebellar/motor efference copies for prediction. The observed whisker-specific adaptation suggests a labeled-line organization within SC that, combined with the dense whisker array, favors detection of novel external motion across adjacent whiskers during active exploration.

This work identifies a neural mechanism in mouse superior colliculus for localizing unexpected external tactile motion during active sensing. SC neurons exhibit a rapidly adapting, whisker-specific transient response that prefers externally generated gains in whisker contact, is attenuated by matching self-generated histories, and habituates with repetition. Population responses encode motion direction/location most effectively when external motion differentially engages whiskers. These findings position the SC as a hub that multiplexes tactile signals with sensorimotor predictions across timescales to guide orienting. Future research should dissect circuit mechanisms and cell types (e.g., L5 cortico-collicular neurons, inhibitory circuits, nigro-collicular inputs), implement predetermined control of specific whiskers to map multi-whisker summation of GoW/LoW, and determine how habituation and predictive signals interact to drive behavior.

• Lack of predetermined control over which whiskers gained/lost contact during external motion limited detailed mapping of multi-whisker summation rules. - The precise circuit mechanisms driving rapid adaptation, predictive attenuation, and slow habituation in SC remain unresolved (e.g., roles of cerebellum, cortex, brainstem, and local inhibition). - While locomotion and whisking were carefully controlled in dedicated paradigms, naturalistic free-moving contexts were not tested. - Generalization to other somatosensory contexts or species remains to be established.