A new type of mouse gaze shift is led by directed saccades

Animals investigate their environments by directing their gaze towards salient stimuli. In the prevailing view, mouse gaze shifts are led by head rotations that trigger compensatory, brainstem-mediated eye movements, including saccades to reset the eyes. These “recentering” saccades are attributed to head movement-related vestibular and optokinetic cues. However, microstimulating mouse superior colliculus (SC) elicits directed head and eye movements that resemble SC-dependent sensory-guided gaze shifts made by other species, raising the possibility mice generate additional types of gaze shifts. We investigated this possibility by tracking eye and attempted head movements in a head-fixed preparation that eliminates head movement-related sensory cues. We found tactile stimuli evoke gaze shifts involving directed saccades that precede attempted head rotations. Optogenetic perturbations revealed SC drives touch-evoked gaze shifts. Thus, mice make sensory-guided, SC-dependent gaze shifts led by directed saccades. Our findings uncover diversity in mouse gaze shifts and provide a foundation for studying head-eye coupling.

movement-related sensory cues. We found tactile stimuli evoke gaze shifts involving directed 23 saccades that precede attempted head rotations. Optogenetic perturbations revealed SC drives 24 touch-evoked gaze shifts. Thus, mice make sensory-guided, SC-dependent gaze shifts led by 25 directed saccades. Our findings uncover diversity in mouse gaze shifts and provide a foundation 26 for studying head-eye coupling. 27 28

Introduction: 29
Natural environments are complex and dynamic, and animals frequently redirect their gaze to 30 scrutinize salient sensory stimuli. Gaze shifts employ head and eye movement coupling 31 strategies that depend on context and can vary between species (Goldring et al., 1996;Land, 1 2019;Land and Nilsson, 2012;Populin, 2006;Populin and Rajala, 2011;Populin et al., 2004a;2 Ruhland et al., 2013;Tollin et al., 2005). Mice are an increasingly important model organism in 3 vision research, yet the strategies they use to shift their gaze remain incompletely understood. 4 Revealing these strategies is essential to understanding mouse visual ethology and the 5 underlying neural mechanisms. 6 7 The prevailing view holds that species such as mice whose retinae lack high-acuity 8 specializations (afoveates) generate gaze shifts driven by head movements followed by 9 "recentering" saccades (Land and Nilsson, 2012). Indeed, recent studies tracking head and eye 10 movements in freely moving mice found that spontaneous and visually evoked mouse gaze 11 shifts matched this description (Meyer et al., 2018(Meyer et al., , 2020Michaiel et al., 2020;Payne and 12 Raymond, 2017). Specifically, during a gaze shift, slow eye movements stabilize the retinal 13 image by countering the rotation of the head and are punctuated by fast saccadic eye 14 movements to recenter the eyes in the orbits as they approach the end of their range of motion. 15 These recentering saccades-also known as "compensatory" saccades or the quick phase of 16 nystagmus-are centripetal, occur in the direction of the head movement, and are thought to be 17 driven by vestibular or optokinetic signals acting on circuits in brainstem (Curthoys, 2002;Hepp 18 et al., 1993;Kitama et al., 1995;Meyer et al., 2020;Michaiel et al., 2020;Payne and Raymond, 19 2017). These recent observations have buttressed the view that gaze shifts in mice and other 20 afoveates are led by head movements, with eye movements made only to compensate for the 21 effects of head movements. In contrast, primates and other foveate species are capable of an 22 additional form of gaze shift led by directed saccades, with or without directed head movements, 23 to redirect their gaze towards salient stimuli (Bizzi et al., 1972;Freedman, 2008;Lee, 1999;24 Zangemeister and Stark, 1982). Directed saccades differ from recentering saccades in that they 25 have endpoints specified by the location of the stimulus (and therefore are often centrifugally 26 3 directed), typically occur simultaneously with or slightly before (20-40 ms) head movements 1 during gaze shifts, and are driven by midbrain circuits, particularly the superior colliculus (SC). 2 To date, there is no behavioral evidence that mice or any afoveate species generate directed 3 saccades or gaze shifts led by eye movements. 4 5 However, three observations are inconsistent with the model that all saccades in mice are 6 exclusively recentering and made to compensate for head movements. First, mouse saccades 7 are not only a product of vestibular or optokinetic cues, because head-fixed mice, in which these 8 signals do not occur, generate saccades, albeit less frequently. Second, neuroanatomical and 9 functional studies suggest that the circuits that underlie directed saccades are conserved in 10 mice (May, 2006;Sparks, 1986Sparks, , 2002. Specifically, microstimulation of the mouse superior 11 colliculus (SC) showed that it contains a topographic map of saccade and head movement 12 direction and amplitude (Masullo et al., 2019;Wang et al., 2015) roughly aligned with maps of 13 visual, auditory, and somatosensory space (Drager andHubel, 1975, 1976). These SC sensory 14 and motor maps resemble those believed to underlie primates' and cats' ability to make gaze 15 shifts led by directed saccades towards stimuli of these modalities (Sparks, 1986(Sparks, , 2002. Third, 16 saccade-like eye movements occurring in the absence of head movements have occasionally 17 been observed in freely moving mice, albeit infrequently and usually in close proximity to head 18 movements (Meyer et al., 2018(Meyer et al., , 2020Michaiel et al., 2020). 19 20 We therefore hypothesized that mice innately generate gaze shifts that incorporate directed 21 saccades. We predicted that this ability was obscured in previous studies for several reasons. 22 First, in freely moving mice it is difficult to uncouple the contributions of reafferent vestibular and 23 optokinetic inputs from those of exafferent (extrinsic) sensory inputs to saccade generation. 24 Second, previous analyses in mice were mostly confined to spontaneous or visually guided 25 gaze shifts, and there is evidence in humans, non-human primates, and cats that gaze shifts in 26 response to different sensory modalities can involve distinct head-eye coupling strategies 1 (Goldring et al., 1996;Populin, 2006;Populin and Rajala, 2011;Populin et al., 2004a;Ruhland 2 et al., 2013;Tollin et al., 2005). Third, in freely moving mice it is difficult to present stimuli in 3 specific craniotopic locations. We therefore reasoned that by using a head-fixed preparation 4 both to eliminate vestibular and optokinetic cues and to present stimuli of different modalities at 5 precise craniotopic locations, we could systematically determine whether mice are capable of 6 gaze shifts involving saccades whose endpoints depend on stimulus location and show different 7 coupling to head movements. We found that tactile stimuli evoke saccades whose endpoints 8 depend on stimulus location, that these saccades precede attempted head rotations, and that 9 these touch-evoked gaze shifts are SC-dependent. Together, these results resolve an apparent 10 discrepancy between mouse neuroanatomy and behavior, demonstrating that mice are capable 11 of generating gaze shifts led by directed saccades. 12 13

Stimulus-evoked gaze shifts in head-restrained mice 15
To test the hypothesis that mice possess an innate ability to make sensory-evoked gaze shifts 16 that incorporate directed saccades coupled to head movements, we head-fixed naïve, wild-type 17 adult animals and used infrared cameras to track both pupils and a strain gauge (also known as 18 a load cell) to measure attempted head rotations. Previous studies in head-fixed mice observed 19 occasional undirected saccades in response to changes in the visual environment (Samonds et 20 al., 2018) and visually guided saccades only after weeks of training and at long (~1 s) latencies 21 (Itokazu et al., 2018). We therefore tested a panel of stimuli of different modalities to determine 22 whether they could evoke saccades. We began by testing the following stimuli from a constant 23 azimuthal location: 1) a multisensory airpuff that provides tactile input to the ears and generates 24 a loud, broadband sound; 2) an auditory stimulus consisting of the same airpuff moved away 25 from the animal so as not to provide tactile input; 3) a tactile stimulus consisting of a bar that 26 Tactile stimuli evoke directed saccades whose endpoints depend on stimulus location 13 To determine whether sensory-evoked saccades are directed, we asked whether saccade 14 endpoints are dependent on stimulus location. We began by examining the endpoints of 15 saccades evoked by left and right ear airpuffs ( Fig. 2A). We found that left ear airpuffs evoked 16 saccades with endpoints far left of center (with center defined as the mean eye position), 17 whereas right ear airpuffs evoked saccades with endpoints far right of center (left: -5.4 ± 4.5°, 18 right: 5.4 ± 3.4°, mean ± s.d., p < 10 -10 Welch's t-test, n = 2155 trials). To understand how this 19 endpoint segregation arises, we examined the trajectories of individual saccades (Fig. 2E). We 20 found that left ear airpuffs elicited nearly exclusively leftward saccades (94.2 ± 3.1%, mean ± 21 s.d., n = 5 mice), whereas right ear airpuffs elicited nearly exclusively rightward saccades (96.0 22 ± 2.1%)-often from the same eye positions. By definition, from any eye position, one of these 23 directions must lead away from center and is thus centrifugal rather than centripetal. In addition, 24 puff-evoked saccades that began towards the center often reached endpoints at eccentricities of 25 5 to 10 degrees. To further test whether saccade endpoints are specified by stimulus location, 26 we repositioned the airpuff nozzles to stimulate the whiskers and repeated the experiments. We 27 reasoned that saccade endpoints should become less eccentric as stimulus eccentricity 28 8 decreases. Indeed, airpuffs applied to the whiskers evoked saccades with endpoints central to 1 those evoked by ear airpuff stimulation, such that the ordering of saccade endpoints mirrored 2 that of stimulus locations (Mean endpoints: left ear, -5.4 ± 4.5°; left whiskers, -0.4 ± 3.8°; right 3 whiskers, 0.8 ± 3.9°; right ear, 5.4 ± 3.4°; mean ± s.d., p < 0.05 for all pairwise comparisons, 4 paired two-tailed Student's t-test ) ( Fig. 2A,B; Supplementary Fig. 3). In this cohort, the 5 separation between whisker-evoked saccade endpoints, although significant, was small, but in 6 other cohorts we have observed larger separation (as well as higher auditory-evoked saccade 7 probabilities) ( Supplementary Fig. 4). Taken together, these data suggest that touch-evoked 8 saccades are directed towards particular eye positions that are specified by stimulus location. We next examined the endpoints of saccades evoked by tactile and auditory stimuli. Similar to 10 the multisensory airpuff stimuli, tactile stimuli delivered to the left and right ears evoked 11 saccades whose endpoints were significantly different (-3.9 ± 5.3° (left) vs. 1.5 ± 5.0° (right); 12 mean ± s.d., p < 10 -10 , Welch's t-test, n = 452 trials) and whose directions were largely opposite 13 from nearly all eye positions (left stimuli evoked 77.5 ± 23.7% leftward saccades, right stimuli 14 evoked 78.3 ± 11.6% rightward saccades, n = 5 mice). This result suggests that tactile stimuli 15 are sufficient to induce gaze shifts that involve directed saccades. We next examined the 16 endpoints and trajectories of saccades evoked by left and right auditory stimuli (Fig. 2D, H). 17 Strikingly, the saccade endpoint locations did not differ significantly for left and right auditory 18 stimuli and were located centrally (0.1 ± 5.1° (left) vs. -0.1 ± 5.1° (right); mean ± s.d., p = 0.72, 19 Welch's t-test, n = 298 trials). Because we had fewer trials with sound-evoked gaze shifts 20 overall, to confirm that this lack of statistically significant endpoint separation was not a result of 21 lower statistical power, we repeated our analyses on matched numbers of sound-and touch-22 evoked saccades sampled at random, once again observing that left and right ear airpuff-, 23 whisker airpuff-, and ear tactile-evoked saccade endpoints were significantly different (ear 24 airpuff, p < 10 -10 ; whisker airpuff, p = 0.0039; ear tactile, p < 10 -10 ; auditory airpuff, p = 0.50; 25 Welch's t-test, see figure for complete statistics). The central endpoints of sound-evoked 26 saccades arose because, in contrast to touch-evoked saccades, saccades evoked by both right 27

Stimulus-evoked gaze shifts involve a different type of head-eye coupling 1
Next, we examined the relationship between attempted head rotations and saccades during 2 gaze shifts. We first analyzed the relative timing of the head and eye components of 3 spontaneous gaze shifts. Previous studies in freely moving and head-fixed mice found that on 4 average, spontaneous and visually evoked saccades in freely moving mice are preceded by 5 head rotations, and that spontaneous saccades in head-fixed mice are similarly preceded by 6 attempted head rotations (Meyer et al., 2018(Meyer et al., , 2020Michaiel et al., 2020;Payne and Raymond, 7 2017). Consistent with these observations, we found that on average spontaneous saccades 8 were preceded by attempted head rotations (Fig. 3B, E). Interestingly, attempted head rotations 9 during spontaneous saccades appeared biphasic, with a slower phase beginning before 10 saccades (median: 90 ms before saccade onset) followed by a fast phase beginning slightly 11 after saccade onset. This biphasic response closely resembles the eye-head coupling pattern 12 reported in freely moving mice (Meyer et al., 2020). , permutation test).

14
We next asked whether the relative timing of eye and attempted head movements was similar 15 for stimulus-evoked gaze shifts. Surprisingly, ear airpuff-evoked saccades were not preceded by 16 slow attempted head rotations but were nevertheless accompanied by fast attempted head 17 rotations that began slightly after saccade onset (Fig. 3B). This pattern was mirrored in the 18 average eye and attempted head movement traces for whisker airpuff-, auditory airpuff-, and 19 ear tactile-evoked saccades (Supplementary Fig. 6A-D). To better understand how these 20 patterns arose, we examined head and eye movement timing at the single-trial level. For 21 spontaneous gaze shifts, head movement onset fell along a continuum, with the majority of 22 attempted head movements beginning before saccade onset (70%), with a median latency of 90 23 ms before saccade onset (Fig. 3E, G). In contrast, for stimulus-evoked gaze shifts, the vast 24 majority of attempted head movements began after saccade onset (74%), with a median latency 25 of 30 ms after saccade onset (Fig. 3F,G). 26   27   15 We next examined the amplitudes of head and eye movements during spontaneous and 1 stimulus-evoked gaze shifts. During spontaneous gaze shifts, attempted head rotations were in 2 the same direction as saccades and scaled with saccade amplitude (Fig. 3A, B, H). These data 3 are consistent with eye-head coupling patterns previously observed in both head-fixed and 4 freely moving conditions (Meyer et al., 2020). Similarly, stimulus-evoked gaze shifts involved 5 attempted head rotations that were made in the same direction as saccades and scaled with 6 saccade amplitude (Fig.3A, B, H). Interestingly, stimulus-evoked saccades of a given amplitude 7 were coupled to an average of 37% smaller attempted head movements than were 8 spontaneous saccades, due largely to the lack of slow pre-saccadic head movements observed 9 during spontaneous gaze shifts (linear regression slopes 0.135 vs. 0.214, p < 10 -5 , permutation 10 test). To confirm that the differences in coupling we observed were not an artifact of differences 11 in saccade size and starting position between saccade types, we performed an additional 12 analysis using subsets of gaze shifts matched for saccade amplitude and initial eye position and 13 The superior colliculus mediates airpuff-evoked gaze shifts 1 We next sought to identify the neural circuitry underlying airpuff-evoked gaze shifts. As 2 discussed previously, in other species, stimulus-evoked gaze shifts involving directed head and 3 eye movements are driven by SC (Freedman, 2008;Freedman et al., 1996;Guitton, 1992;4 Guitton et al., 1980;Paré et al., 1994). In contrast, it is widely believed that the recentering 5 saccades observed in mice are driven by brainstem circuitry in response to head rotation 6 (Curthoys, 2002;Hepp et al., 1993;Kitama et al., 1995;Meyer et al., 2020;Michaiel et al., 2020;7 Payne and Raymond, 2017). To determine whether SC is required to generate touch-evoked 8 gaze shifts in mice, we pursued an optogenetic strategy to perturb SC activity in the period 9 surrounding airpuff onset. For inhibition experiments, we stereotaxically injected adeno-10 associated virus (AAV) encoding the light-gated chloride pump eNpHR3.0 under the control of a 11 pan-neuronal promoter and implanted a fiber optic in right SC (Gradinaru et al., 2010). 12 Consistent with data in foveate species (Hikosaka and Wurtz, 1985;Robinson, 1972;Schiller 13 and Stryker, 1972), optically reducing right SC activity shifted airpuff-evoked saccade endpoints 14 to the right (i.e., ipsilaterally) for both left (-3.7 ± 4.3° (control) vs. -2.3 ± 4.7° (LED on), p < 15 0.001, Welch's t-test) and right ear airpuffs (4.5 ± 4.1° (control) vs. 5.4 ± 4.7° (LED on), p = 16 0.011, Welch's t-test) ( Fig. 4A-C). To control for potential mismatches in starting eye position 17 between LED-off and LED-on trials, we performed additional analyses using matched trials and 18 found that the endpoint and amplitude differences persisted ( Supplementary Fig. 10). For 19 stimulation experiments, we stereotaxically injected AAV encoding the light-gated ion channel 20 ChR2 under the control of a pan-neuronal promoter and implanted a fiber optic in right SC 21 (Gradinaru et al., 2010). Because strong SC stimulation can evoke saccades, we used weak 22 stimulation (50-120 μW) in order to bias SC activity. This manipulation caused the reciprocal 23 effect of right SC inhibition, biasing endpoints leftwards (i.e., contraversively) for both left (-5.8 ± 24 4.6° (control) vs. -8.0 ± 6.1° (LED on), p = 0.0016, Welch's t-test) and right (5.6 ± 3.5° (control) 25 vs. 1.9 ± 6.1° (LED on), p < 10 -5 , Welch's t-test) airpuffs (Fig. 4E-G , permutation test).

23
To understand how SC manipulations affect attempted head movements and head-eye 24 coupling, we examined the distribution of head movements as a function of saccade amplitude. 25 As expected given the role of SC in generating both head and eye movements, attempted head 26 movements were shifted to the right (i.e., ipsiversively) by SC inhibition (Fig. 4D) and to the left 27 (i.e., contraversively) by SC excitation (Fig. 4H). To examine the effects of SC manipulations on 28 head-eye coupling, we identified trials with identical saccade trajectories in LED-on and LED-off 29 conditions and examined the corresponding head movement amplitudes. Interestingly, SC 30 manipulations had no effect on the relationship between saccade and head movement 1 amplitudes, suggesting that SC manipulations do not change head-eye coupling during gaze 2 shifts. Taken together, these bidirectional manipulations indicate that SC serves a conserved 3 necessary and sufficient role in generating ear airpuff-evoked gaze shifts in which it specifies 4 overall gaze shift amplitude. 5 6 Discussion: 7 Here we investigated whether mouse gaze shifts are more diverse than had previously been 8 appreciated. In the prevailing view, mouse gaze shifts are led by head rotations that trigger 9 compensatory eye movements, including saccades that function to reset the eyes (Land, 2019;10 Land and Nilsson, 2012;Liversedge et al., 2011;Meyer et al., 2020;Michaiel et al., 2020;11 Payne and Raymond, 2017). These "recentering" saccades are attributed to head movement-12 related vestibular and optokinetic cues (Curthoys, 2002;Meyer et al., 2020;Michaiel et al., 13 2020;Payne and Raymond, 2017). Working in a head-fixed context to eliminate vestibular and 14 optokinetic cues and to present stimuli of different modalities at precise craniotopic locations, we 15 found that mice are capable of an additional type of gaze shift. As discussed below, we 16 identified numerous features that distinguish this new type of gaze shift from the previously 17 studied type, including the endpoints of the saccades, their timing and amplitude relative to 18 attempted head movements, and the brain regions that drive them. 19 20 Touch-evoked saccades are directed 21 The first indication that touch-evoked gaze shifts differ from those previously observed in mice 22 was an endpoint analysis revealing that touch-evoked saccades are directed rather than 23 recentering. This conclusion is based on three lines of evidence. First, endpoints of saccades 24 evoked by left and right ear airpuffs are near the left and right edges, respectively, of the range 25 of eye positions observed and overlap minimally, despite trial-to-trial variability. In contrast, 26 endpoints for saccades evoked by left and right auditory stimuli are indistinguishable and well 1 described by the recentering model. Second, left and right ear airpuffs evoke saccades traveling 2 in opposite directions from most eye positions; by definition, one of these directions must lead 3 away from center and is thus centrifugal rather than centripetal. In contrast, saccades evoked by 4 both left and right auditory stimuli travel centripetally from all initial eye positions. Third, from 5 many eye positions, touch-evoked saccades that begin towards the center pass through to 6 reach endpoints at eccentricities between 5 to 10 degrees and cannot accurately be termed 7 centripetal. For these reasons, we conclude that touch-evoked saccades are directed and do 8 not serve to recenter the eyes. 9 10 Our findings contrast with and complement previous studies contending that rodents, like other 11 afoveates, use saccades to reset their eyes to more central locations (Meyer et al., 2018(Meyer et al., , 202012 Michaiel et al., 2020;Wallace et al., 2013). One recent analysis suggested that gaze shifts 13 made during visually guided prey capture involve resetting centripetal saccades that "catch up" 14 with the head (Michaiel et al., 2020). Another found that saccades away from the nose recenter 15 the eye, whereas saccades toward the nose move the eye slightly beyond center (Meyer et al., 16 2020). Although we observed this as well, it does not contribute to our results because we 17 averaged the positions of the left and right eyes, eliminating this asymmetry. Earlier studies in 18 head-fixed mice observed occasional, undirected saccades in response to changes in the visual 19 environment (Samonds et al., 2018) and found that mice could be trained to produce visually 20 guided saccades only after weeks of training and at extremely long (~1 s) latencies (Itokazu et 21 al., 2018). To our knowledge, ours is the first study demonstrating innate gaze shifts involving 22 directed saccades in mice (or any species lacking a fovea). 23 24 25 26

Touch-evoked gaze shifts are not led by head movements 1
The prevailing view holds that head movements initiate and determine the amplitude of mouse 2 gaze shifts, with eye movements compensatory by-products. In support of this model, one study 3 found that spontaneous saccades in head-fixed mice are preceded by attempted head rotations. 4 A careful comparison with gaze shifts occurring during visually guided object tracking and social 5 interactions in freely moving mice led the authors to suggest that head-eye coupling is not 6 disrupted during head-fixation, and that gaze shifts in both contexts are head-initiated (Meyer et 7 al., 2020). Another study tracked the eyes and head during visually guided cricket hunting and 8 found that gaze shifts are driven by the head, with the eyes following to stabilize and recenter 9 gaze (Michaiel et al., 2020). Together, these findings have bolstered the prevailing view that 10 afoveates such as mice generate gaze shifts driven by head movements, with eye movements 11 compensatory by-products (Land, 2019;Land and Nilsson, 2012;Liversedge et al., 2011). 12 However, while our findings for spontaneous saccades are consistent with those in the 13 literature, we have shown that touch-evoked gaze shifts are initiated by saccades, a finding that 14 contrasts with and complements earlier studies. 15 16 This finding that touch-evoked gaze shifts are initiated by saccades suggests that mouse 17 saccades are not always a simple by-product of head movements. Additional support for this 18 idea came from an analysis of head and eye movements as a function of eye position. If gaze 19 shifts were determined solely by the location of the stimulus relative to the head and saccades 20 were a compensatory by-product of this calculation, eye position should have no effect on head 21 movement amplitude. However, we found that the amplitudes and directions of both saccades 22 and attempted head movements evoked by saccades vary with initial eye position. The 23 influence of eye position on touch-evoked eye and head movements further indicates that 24 saccades are not compensatory by-products of head movements. Instead, touch-evoked head 25 movements and directed saccades are specified simultaneously as parts of a coordinated 26 movement whose component movements take into account both stimulus location and initial 1 eye position. 2 3 The relative timing of head and eye components during touch-evoked gaze shifts in mice 4 resembles that observed during gaze shifts in cats and primates. For example, head-fixed cats 5 and primates generate gaze shifts using directed saccades and then maintain their eyes in the 6 new orbital position, similar to what we have observed in mice (Freedman, 2008;Guitton et al., 7 1980). In addition, saccades in head-fixed cats and primates are often accompanied by 8 attempted head rotations, similar to those we observe during touch-evoked gaze shifts in head-9 fixed mice (Bizzi et al., 1971;Guitton et al., 1984;Paré et al., 1994). In primates and cats able to 10 move their heads, gaze shifts are usually led by directed saccades (with some exceptions), 11 likely because the eyes have lower rotational inertia and can move faster (Pelisson and 12 Guillaume, 2009;Ruhland et al., 2013). These saccades tend to be followed by a head 13 movement in the same direction that creates vestibular signals that drive slow, centripetal 14 counterrotation of the eyes to maintain fixation (Bizzi et al., 1972;Freedman, 2008;Freedman 15 and Sparks, 1997;Guitton et al., 1984). In this way, the animal can rapidly shift its gaze with a 16 directed saccade yet subsequently reset the eyes to a more central position as the head moves. 17 It is tempting to speculate that a similar coordinated sequence of head and eye movements 18 occurs during freely moving touch-evoked gaze shifts, enabling mice to rapidly shift gaze with 19 their eyes while eventually resetting the eyes in a more central orbital position. 20 21

Head-eye amplitude coupling differs during spontaneous and evoked saccades 22
An additional feature that distinguishes spontaneous and touch-evoked gaze shifts is the 23 relative contributions of head and eye movements. We found that spontaneous saccades of a 24 given amplitude are coupled to larger head movements than are touch-evoked saccades. This 25 difference arises largely from the absence of a pre-saccadic attempted head movement during 26 touch-evoked gaze shifts. This differential pairing of head and eye movements is reminiscent of 1 reports in primates and cats that the relative contributions of head and eye movements vary for 2 gaze shifts evoked by different sensory modalities (Goldring et al., 1996;Populin, 2006;Populin 3 and Rajala, 2011;Populin et al., 2004a;Ruhland et al., 2013;Tollin et al., 2005). However, in 4 those species, vision typically elicits gaze shifts dominated by saccades while hearing typically 5 evokes gaze shifts entailing larger contributions from head movements. In contrast, we 6 observed that sound-and touch-evoked gaze shifts involve larger contributions from saccades 7 than do spontaneous gaze shifts, whereas visual stimuli did not evoke gaze shifts at all. This 8 indicates that although there is general conservation of the involvement of SC in sensory-driven 9 gaze shifts, modality-specific features are not conserved, which may reflect differences in 10 sensory processing across species. For example, virtually every cell in primate intermediate and 11 deep SC that is responsive to tactile stimuli is also responsive to visual stimuli (Groh and 12 Sparks, 1996). In contrast, a recent study of Pitx2 + neurons in the intermediate and deep mouse 13 SC, which project to brainstem oculomotor centers and when optogenetically stimulated evoke 14 orienting movements of the eyes and head reminiscent of the gaze shifts we describe (Masullo 15 et al., 2019), reported that these neurons responded robustly to whisker airpuffs but did not 16 respond to visual stimuli (Xie et al., 2021). We also observed limited variability across mice in 17 the relative contributions of head and eye movements to gaze shifts, which contrasts with the 18 observation that different human subjects are head "movers" and "non-movers" during gaze 19 shifts ( Supplementary Fig. 9) Thus, our results reveal that mice are capable of using multiple 20 strategies to shift their gaze but with key differences from other species. 21

22
The role of superior colliculus in sensory-evoked mouse gaze shifts 23 In other species, SC drives sensory-evoked gaze shifts, and microstimulation and optogenetic 24 stimulation of mouse SC has been shown to elicit gaze shifts (Masullo et al., 2019;Wang et al., 25 2015). However, to our knowledge, no study had identified a causal involvement of SC in mouse 26 gaze shifts. We performed bidirectional optogenetic manipulations that revealed that touch-1 evoked gaze shifts depend on SC, identifying a conserved, necessary and sufficient role for SC 2 in directed gaze shifts. In addition, we found that SC manipulations did not alter head-eye 3 amplitude coupling. This observation suggests that SC specifies the overall gaze shift amplitude 4 rather than the individual eye or head movement components, consistent with observations in 5 other species (Freedman et al., 1996;Paré et al., 1994). 6 7 Ethological significance 8 Prior to the present study, it was believed that species with high-acuity retinal specializations 9 acquired the ability to make directed saccades to scrutinize salient environmental stimuli, 10 because animals lacking such retinal specializations were thought incapable of gaze shifts led 11 by directed saccades (Land, 2019;Land and Nilsson, 2012;Liversedge et al., 2011;Walls, 12 1962). Our discovery that sensory-guided directed saccades are present in mice-albeit without 13 the precise targeting of stimulus location seen in foveate species-raises the question of what 14 fovea-independent functions these movements serve. Although mice have lateral eyes and a 15 large field of view, saccades that direct gaze towards a stimulus, as seems to occur with both 16 ear and whisker tactile stimuli, may facilitate keeping salient stimuli within the field of view. As 17 natural stimuli are often multimodal, directing non-visual stimuli towards the center of view 18 maximizes the likelihood of detecting the visual component of the stimulus. Alternatively, despite 19 mouse retinae lacking discrete, anatomically defined specializations such as foveae or areas 20 centralis, there are subtler nonuniformities in the distribution and density of photoreceptors and 21 retinal ganglion cell subtypes, and magnification factor, receptive field sizes, and response 22 tuning vary across the visual field in higher visual centers; it may be desirable to center a salient 23 tactile stimulus on a particular retinal region to enable scrutiny of the visual component of the 24 stimulus (Ahmadlou and Heimel, 2015;Baden et al., 2013;van Beest et al., 2021;Bleckert et 25 al., 2014;Drager and Hubel, 1976;Feinberg and Meister, 2015;Li et al., 2020;de Malmazet et 26 25 al., 2018). Although touch-evoked saccades alone may be too small to center the stimulus 1 location on any particular region of the retina, they may do so in concert with directed head 2 movements. 3 4 Why tactile stimuli evoked directed saccades in our preparation whereas auditory and visual 5 stimuli do not is unclear. One possibility is the aforementioned speed of saccades relative to 6 head movements may be especially beneficial because tactile stimuli typically derive from 7 proximal objects and as a result may demand rapid responses. Alternatively, because the 8 spatial acuity of the tactile system is higher than that of the auditory system in mice, this may 9 enable more precise localization of the tactile stimuli (Allen and Ison, 2010;Diamond et al., 10 2008). Auditory stimuli, in contrast, may alert animals to the presence of a salient stimulus in 11 their environments whose location is less precisely ascertained, and as a result drive gaze shifts 12 whose goal is to reset the eyes to a central position that maximizes their chances of sensing 13 and responding appropriately. Finally, our set of stimuli was not exhaustive, and it is possible 14 that as yet unidentified visual or auditory stimuli could elicit gaze shifts with directed saccades. 15 Alternatively, the recent report that mouse SC Pitx2 + neurons respond to tactile but not visual 16 stimuli may indicate that visual orienting is driven by a distinct neural pathway or that head-fixing 17 gates visual but not tactile responses of SC neurons. 18

Future Directions 20
In this study we used a head-fixed preparation to eliminate the confound of head movement-21 related sensory cues and to present stimuli from defined locations. However, in the future, it will 22 be interesting to compare touch-evoked gaze shifts in head-fixed and freely moving animals. 23 For example, as noted previously, whereas head-fixed primates and cats generate gaze shifts 24 using directed saccades and then maintain their eyes in the new orbital position, similar to what 25 we have observed, in freely moving primates and cats, gaze shifts are led by directed saccades 26 but typically followed by head movements during which the eyes counterrotate centripetally in 1 order to maintain gaze in the new direction. It will be interesting to know whether similar 2 differences distinguish saccade-led touch-evoked gaze shifts in head-fixed and freely moving 3 mice. By expanding on methods recently described by other groups, it may be possible to 4 investigate these and other questions (Meyer et al., 2018(Meyer et al., , 2020Michaiel et al., 2020;Payne 5 and Raymond, 2017). 6 7 Furthermore, a practical implication of our identification of mouse SC-dependent gaze shifts is 8 that this behavioral paradigm could be applied to the study of several outstanding questions. 9 First, there are many unresolved problems regarding the circuitry and ensemble dynamics 10 underlying target selection (Basso and May, 2017) and saccade generation (Gandhi and 11 Katnani, 2011), and the mouse provides a genetically tractable platform with which to 12 investigate these and other topics. Second, gaze shifts are aberrant in a host of conditions, such 13 as Parkinson's and autism spectrum disorder (Liversedge et al., 2011). This paradigm could be 14 a powerful tool for the study of mouse models of a variety of neuropsychiatric conditions. Third, 15 directing saccades towards particular orbital positions during these gaze shifts requires an 16 ability to account for the initial positions of the eyes relative to the target, a phenomenon also 17 known as remapping from sensory to motor reference frames. Neural correlates of this process 18 have been observed in primates (Groh and Sparks, 1996;Jay and Sparks, 1984)  We have found that mice make an unexpectedly broad range of gaze shifts, including a new 2 type that is elicited by sensory stimuli, led by directed saccades, and incorporates smaller head 3 movements. Prior studies in species whose retinae lack high-acuity specializations had never 4 observed this type of gaze shift, but our study used a broader range of stimuli than previously 5 tested in a preparation that allowed spatially precise delivery. Detailed perturbation experiments 6 determined that the circuit mechanisms of sensory-evoked gaze shifts are conserved from mice 7 to primates, suggesting that this behavior may have arisen in a common, afoveate ancestral 8 species long ago. More broadly, our findings suggest that analyzing eye movements of other 9 afoveate species thought not to make directed saccades-such as rabbits, toads, and 10 goldfish-in response to a diverse range of multimodal stimuli may uncover similar abilities to 11 make diverse types of gaze shifts.  14 15