A Novel Three-Choice Touchscreen Task to Examine Spatial Attention and Orienting Responses in Rodents

Abstract Mammalian orienting behavior consists of coordinated movements of the eyes, head, pinnae, vibrissae, or body to attend to an external stimulus. The present study aimed to develop a novel operant task using a touch-screen system to measure spatial attention. In this task, rats were trained to nose-poke a light stimulus presented in one of three locations. The stimulus was presented more frequently in the center location to develop spatial attention bias toward the center stimulus. Changes in orienting responses were detected by measuring the animals’ response accuracy and latency to stimuli at the lateral locations, following reversible unilateral chemogenetic inactivation of the superior colliculus (SC). Additionally, spontaneous turning and rotation behavior was measured using an open-field test (OFT). Our results show that right SC inactivation significantly increased the whole body turn angle in the OFT, in line with previous literature that indicated an ipsiversive orientating bias and the presence of contralateral neglect following unilateral SC lesions. In the touch screen orienting task, unilateral SC inactivation significantly increased bias toward the ipsilateral side, as measured by response frequency in various experimental conditions, and a very large left-shift of a respective psychometric function. Our results demonstrate that this novel touchscreen task is able to detect changes in spatial attention and orienting responses because of e.g. experimental manipulations or injury with very high sensitivity, while taking advantage of the touch screen technology that allows for high transferability of the task between labs and for open-source data sharing through https://www.mousebytes.ca.


Introduction 80
Orienting behaviour in mammals consists of highly coordinated movements of the eyes, head, 81 pinnae, vibrissae, and body towards salient sensory stimuli. Sensory information relevant to spatial 82 orienting, such as visual stimulus location, is represented topographically in the superior colliculus 83 (SC) across a wide range of vertebrate species (Gaither and Stein, 1979;May, 2006). The SC is a 84 laminar midbrain structure that is critical for the generation of orienting behaviours, serving the 85 goal of aligning the sensory apparatus of an animal with objects of interest in the surrounding 86 environment. Consistent with the evolutionarily highly conserved nature of this structure, SC 87 lesions result in severe orienting impairments in a range of vertebrate species as diverse as tree 88 shrews (Casagrande and Diamond, 1974), cats (Sprague and Meikle, 1965), and nonhuman 89 primates (Schiller et al., 1987). Along the same line of evidence, electrical microstimulation of the 90 SC has revealed a topographic organization of orienting behaviours. In the rhesus macaque, 91 stimulation of rostral regions of the SC evokes small amplitude contraversive saccades (Robinson, 92 1972), while stimulation of the caudal SC evokes large amplitude contraversive saccades and head 93 movements (Corneil et al., 2002;Freedman et al., 1996;Robinson, 1972). Saccades are rapid eye 94 movements common to primate species that form a considerable part of their orienting behavior 95 by moving their fovea to the visual stimuli of interest. Although rodents do not possess a well-96 defined fovea to produce the same kind of saccades, electrical stimulation of the SC in rodents has 97 evoked contraversive movements of the eyes and coordinated eye, head, pinnae, vibrissae and Sprague and Meikle, 1965; Tehovnik, 1989). The former deficit has been characterized as a failure 104 of motor implementation of orienting, while the latter has been ascribed to changes in spatial 105 9 In test sessions, the central panel was illuminated on 75% of trials (center-only trials). The 261 remaining 25% of trials were divided equally into 5 double stimulus trial types: stimuli were 262 presented in both the leftmost and rightmost panels with the following stimulus onset asynchrony 263 (SOA) conditions: simultaneous left + right stimulus presentation; left stimulus followed by right 264 stimulus with either a 0.5 or 1s SOA; or right stimulus followed by left stimulus with either a 0.5 265 or 1s SOA (see Fig. 2B). Trials were presented in a randomized order. For all double-stimulus 266 trials, a correct response was defined as nose-poking either the left or the right panel after a 267 stimulus was presented, irrespective of which one appeared first, and the animal was rewarded 268 accordingly. An incorrect response in double-stimulus trials referred to when the animal nose-269 poked the center panel (in which no stimulus was presented), or the leftmost or rightmost panels 270 before any stimulus was presented. Touching any other panel between the center and the sides was 271 recorded but neither rewarded nor punished. Trial types were presented in pseudorandomized 272 order. Performance of the full task involved completing 200 trials over 90 min (100 trials in 60 273 min, for an early cohort). The session was set to end once the animal finished all the trials or once 274 the time limit was reached. Each animal was tested twice with CNO and twice with vehicle, in a 275 counterbalanced order, 20 minutes following the injection (one early cohort was tested only once 276 for each condition). Tests were performed at least 3 days apart to ensure injection washout, and 277 rats were maintained on center attention baseline training between the test sessions. Data from 278 both sessions following the same injection was combined and used to calculate the respective 279 animal's performance parameters. The total number of completed trials, accuracy rate, and 280 omission rate were analyzed, as well as the number of responses made to each panel and the 281 respective latencies, separated for each trial type (center only or SOA trials), using the ABET II 282 Touch software (Lafayette Instruments, Lafayette IN, USA). All training and testing ABET II files 283 are provided with the manuscript as Extended Data 1. Touchscreen response rates were calculated 284 as a percentage of total trials where a response was made (excluding trials where the animals made 285 an omission) to avoid biasing the data with trials in which the animal was not facing the panels 286 (e.g. while grooming, eating), as was the case in a small fraction of total trials. In contrast, omission 287 rate was calculated as a percentage of all trials of a particular trial type presented in a session. 288 289 Immunohistochemistry 290 To confirm the expression of the DREADDs in the SC, all animals were perfused with saline 291 followed by 4% paraformaldehyde (PFA) and brains were harvested, post-fixed in PFA for 1 h, 292 and stored in 30% sucrose at 4°C until completely sunken. Brains were sliced into 4 series of 40 293 µm coronal sections using a freezing microtome (Microm HM 560 M) and stored at -20°C in 294 cryoprotectant solution. One series of the sections was used for immunohistochemistry. Free 295 floating tissue sections were thoroughly washed in 0.1M phosphate-buffered saline (PBS) between 296 incubations and all incubations were performed at room temperature with gentle agitation. Sections 297 were blocked with 1% H2O2 in 0.1 M PBS for 10 minutes and pre-absorbed in PBS+ (0.1% bovine 298 serum albumin, 0.4% TritonX100 in PBS) followed by overnight incubation with rabbit anti 299 mCherry (Abcam Cat# ab167453, RRID: AB_2571870) in PBS+. Subsequently, the sections were 300 incubated with biotinylated goat anti-rabbit (Vector Labs Cat# BA-1000, RRID: AB_2313606) for 301 1 hour (1:500 in PBS) and by the avidin horseradish peroxidase for 1 hour (ABC-elite, 1:1000 in 302 PBS; Vector Laboratories, Burlingame, CA, USA). Finally, the peroxidase complex was 303 visualized by exposure for 10 minutes to a chromogen solution containing 0.02% 3,3'-304 diaminobenzidine tetrahydrochloride. At the end of the staining protocol, sections were washed 305 thoroughly with 0.1 M PB, mounted onto plus-charged glass slides with 0.3% gelatin in distilled 306 water and cover-slipped with DPX mounting medium (EMD Millipore, USA). Imaging was 307 performed using a Nikon Eclipse Ni-U upright microscope with a DS-Qi2 high-definition color 308 camera and imaging software NIS Elements Color Camera (Nikon Instruments, Melville, NY, 309 USA). 310 311

Statistical analysis 312
All statistical analyses were carried out using SPSS (IBM). 313 Testing for normality, outliers, homogeneity of variance, and sex: Before conducting statistical 314 analyses, data were scanned for normality using the Shapiro-Wilk test, for outliers using box and 315 whisker plots, and for homogeneity of variance assumptions for relevant analyses. This was 316 conducted for each measure of interest in the OFT and touch screen 3-choice orienting task. For 317 data that did not exhibit a normal distribution according to the Shapiro-Wilk test (p<0.05), we 318 conducted nonparametric Mann-Whitney or Friedman tests, instead of independent t-tests and one-319 way repeated measures ANOVA (see below). For data that did not violate normality but violated 320 the assumption of homogeneity of variance, an adjusted p value for t-tests was used. Data of both 321 sexes were merged throughout the study, as there were no statistically significant differences 322 between sexes, except for overall locomotor activity. The introduction of difference scores for 323 each animal between CNO and vehicle trials normalized the data and eliminated the difference in 324 overall locomotion. 325 326 Open field test: A CNO-vehicle difference score was calculated for total distance travelled, total 327 rotations, the percentage of clockwise rotations, the head turn angle sum and the turn angle sum. 328 Independent t-tests were conducted for all measured except total distance travelled, comparing the 329 DREADD group with the sham group, and p values were chosen accordingly based on whether 330 the assumption of homogeneity of variance was met (only case is for turn angle sum). For the total 331 distance, a Mann-Whitney U test was performed. 332 Touch screen baseline performance: For training data across days on the 3-choice orienting task, 333 pre-test baseline training days just prior to respective CNO or vehicle testing days were analyzed, 334 using only animals that had 2 CNO and 2 vehicle test days (4 pre-test days in total). Accuracy 335 percentage, omission percentage, reward collection latency, and correct response latency were 336 analyzed. After determining no difference in performance across pre-test days between DREADD 337 and sham animals, data of the two groups was merged. 338 Accuracy percentage, omission percentage and correct response latency were analyzed separately 339 for left stimulus trials, center stimulus trials, and right stimulus trials. Preliminary analysis using 340 repeated measures ANOVA or Friedman test revealed no significant difference between pre-test 341 days, after adjusting for multiple comparisons, in percentage accuracy, percentage omissions or 342 correct response latency, for any of the 3 trial types. Therefore, data from all 4 pre-test days were 343 combined and used to calculate overall trial type-specific percentage accuracy, percentage 344 omission and correct response latency. Then, the percentage accuracy, percentage omission and 345 correct response latency were compared between trial types using a one-way repeated measures 346 ANOVA or Friedman test, depending on whether normality and homogeneity of variance criteria 347 were met.  Curves were fitted and PES values were calculated only for animals that had more than 5 trials on 362 all SOAs in which they made a response, which led to the exclusion of 1 Sham animal and 5 363 DREADD animals. CNO effects were so strong on some DREADD animals that they responded 364 almost always to the right, so the PES could not be calculated (Fig. S6). In these cases the smallest 365 SOA value used to generate the fitted curve (SOA = -1) was used as a conservative estimate. A 366 two-way repeated measures ANOVA was performed with group (Sham vs DREADD) as a 367 between subject factor and injection (vehicle vs CNO) as a within subject factor. 368 369

370
This study examined the effects of DREADD-induced transient deactivation of the right SC on 371 orienting behavior in Long Evans rats. The OFT was initially utilized to determine changes in 372 spontaneous turning and rotation behavior upon deactivating the right SC, followed by a novel The purpose of baseline center training sessions was to train the rats to orient towards the center 418 stimulus more than the flanking stimuli by using a higher proportion of center stimulus trials (80% 419 center, 10% left, 10% right). Given that the experimental design involves repeated testing of 420 animals, it was important that animals maintained a stable baseline performance between test 421 sessions, especially on pre-test days prior to their CNO or vehicle test sessions. The baseline 422 training was able to achieve this stable performance, as demonstrated by the lack of significant In order to investigate the orienting bias produced by baseline training in more detail, we analyzed 432 trial-type-specific responses. For this purpose, all trials from pre-test days were combined after 433 confirming that trial-type-specific measures of interest did not significantly differ between the 434 different pre-test days (adjusted for multiple comparisons). Trial-type-specific responses during 435 pre-test days showed a strong center bias, with higher accuracy on center trials compared to left or successfully produced a strong center bias, as also easily observed in a representative video taken 444 during one of the baseline training days on stage 7 (see movie 2). 445

Behavior: 3-choice orienting task testing 446
The testing protocol was comprised of 75% baseline center trials to retain the animals' center bias, 447 and 25% double stimulus side trials with varying SOAs to test the orienting preference to varying 448 degrees following CNO or vehicle injections (see Fig. 2B). Separate analyses were conducted for 449 baseline center-only trials and double-stimulus trials.  (Figures 8C & 8D). 477 There was a significant interaction between group and injection (F(1,4)  In this present study we developed a touchscreen-based experimental approach to measure spatial 484 attention and orienting responses in rodents. Given the well-established role of the SC in orienting 485 behavior of rodents, we validated the novel task through unilateral SC inactivation using 486 DREADDs. Our results show that the touchscreen 3-choice orienting task was highly sensitive to 487 DREADD-induced SC inactivation, providing a variety of measures to assess orienting bias. 488

DREADD expression 490
In most of the animals, DREADDs were expressed throughout a large portion of the right SC, from 491 superficial layers down to the deeper layers (see Fig. 3). In some animals, we also observed a low We first analyzed OFT behaviour to detect general alterations in the animals' spontaneous 505 locomotor behavior after deactivation of the right SC. We found that CNO increased clockwise 506 body turning behaviour, but not head turning behaviour or the percentage of 360-degree clockwise 507 rotations in DREADD animals. The observation of a higher ipsiversive body turn angle is 508 consistent with a deficit in orienting to the contralateral visual field as a consequence of unilateral 509 SC deactivation and the ipsilesional circling behavior that has previously been reported in lesion 510

SC manipulations impact orienting responses 518
Various studies reported that damage to the SC impaired the ability to perform orienting 519 movements towards visual stimuli presented in the hemifield contralateral to the lesioned SC, 520 particularly those presented in the periphery (Schneider, 1975 Asdourian, 1981; Imperato and Di Chiara, 1981). We did not observe any noticeable changes in 532 the rats' posture upon CNO administration in any of our experiments. However, studies reporting 533 bodily asymmetry often used the GABA agonist muscimol for SC inhibition, which might be 534 responsible for postural asymmetries observed, since manipulation of GABAergic mechanisms 535 has previously been reported to cause postural asymmetry and muscular rigidity in rats (Turski et 536 al., 1984). The DREADDS used in this study affected all cell types. Indeed, more recent studies 537 that applied optogenetic techniques to unilaterally deactivate the SC did also not report on any The importance of a centre bias 546 Training on the novel 3-choice orienting task was successful in developing a center-orienting 547 preference, evident from higher accuracy and shorter response latency on center trials compared 548 to left or right trials during the test sessions. Accuracy and reward collection latency were 549 consistent across all pre-test days, indicating that animals returned to approximately the same 550 baseline orienting bias and motivation before each test session. In the varying SOA trials during 551 test sessions, DREADD but not sham rats showed a substantial left shift in their rightward response 552 psychometric function following CNO injection. This indicated a rightward bias which was also 553 shown by a significant decrease in PES in DREADD animals post-CNO compared to vehicle. The 554 rightward bias was rather extreme in some cases, causing animals to respond to the right on all 555 trials in all SOAs, even when they were primed to respond towards the left side in -0.5 and -1 SOA 556 trials, indicating an ipsiversive orienting bias and contraversive neglect after CNO administration. 557 Even in analysis looking at baseline center-only trials, a rightward orienting bias following CNO 558 could be detected during test sessions. Animals were still very accurate on these center-only trials 559 (center response rate of > 90%), but there was a significant decrease in CNO-vehicle score of 560 percentage center responses, whereby animals chose the center choice less frequently and tended 561 towards choosing the right choice more frequently instead, even though only the center choice was 562 illuminated and choosing the non-illuminated right panel was punished. Similar findings were 563 reported by Sinnamon and Garcia (1988) using an operant orienting task with 2 lateral choices and 564 one center choice, however, their results also showed a neglect to center positions after unilateral 565 lesions of the SC, probably due to a lack of enforcing and maintaining a centre orientation bias. 566 Without this bias, it is difficult to interpret the behavioural data, as the animal might simply shift 567 its body posture and attention towards the preferred side. 568 569

Benefit of the touchscreen system 570
The touchscreen-based orienting task is a purely visual method that provided great advantages in 571 terms of precise control of the stimulus duration and location. Compared to tasks that require 572 animals to exert pressure to push a door or a lever, our orienting task benefits from detecting the 573 nose-poke via infrared beam without requiring any pressure, providing a more direct association 574 with orienting behavior (Bussey et al., 2008). Cook et al. (2004) found that rats learned a visual 575 discrimination much faster when they were required to nose-poke the stimuli on a touchscreen as 576 opposed to pushing a lever underneath the stimuli, likely due to differences in the spatial contiguity 577 of stimuli and responses (Cook et al., 2004). Spatial contiguity between stimulus and response is 578 crucial when orienting towards the stimulus is the main goal of the task, and it allows stimulus 579 location to be varied without interfering with the nature of the task or learning speed. The novel 580 task presented here can be easily adapted to studies of orienting response to various directions, 581 since stimuli can be presented anywhere on the touchscreen, or response to stimuli of varying 582 shapes, sizes and salience (brightness). Based on our findings, training rats on this orienting task 583 can be achieved in a relatively short period of time, reaching baseline training of the highest 584 difficulty level (stimulus duration of 1.5s) in less than 3 weeks on average. Similar to all other 585 touchscreen-based tasks, the 3-choice orienting task is highly automated and easy to run and task 586 parameters such as stimulus duration, timeout duration or the amount of reward can easily be 587 modified if needed. Although we did not perform video analysis in our touchscreen experiments, 588 the systems is equipped with video recorders, which can be used in combination with software like 589 ANY-maze to track the location of the animal's head at specific times during the task. Given the 590 broad utilization of touchscreen-based platforms for cognitive testing, this task offers the ability 591 to test orienting behaviour across labs under identical experimental paradigms, benefiting from the 592 numerous advantages these touchscreen systems provide (Mar et al., 2013). Touchscreen-based 593 platforms can also be combined with other techniques such as optogenetics or electrophysiology, 594 to manipulate or record neural activity at precise time points during a testing session (for example, 595 just before a response is made or right after a trial is initiated), and this can be used to more 596 accurately study the role of structures implicated in spatial attention and orienting behaviour 597 20 (Bussey et al., 2008(Bussey et al., , 1994Horner et al., 2013).  animals. Expression is displayed across six slices, from Bregma -5.52 mm as the most rostral, to 784 Bregma -6.60 mm as the most caudal slice. The color gradient on the SC represents the number of 785 animals that had DREADD expression at each subregion of the SC: the darker the gradient at a 786 subregion, the higher the number of animals that had DREADD expression in that subregion. Each 787 schematic slice is accompanied by an immunohistochemistry photo of the same slice from a 788 representative animal. The images of the slices were taken from The Rat Brain Atlas in Stereotaxic 789 Coordinates from Paxinos and Watson (2006). 790 791 Figure 4. Body turning but not head turning behavior is increased following CNO in DREADD 792 but not sham animals. 793 Rats spontaneously explored an enclosed arena 20 minutes after injection of vehicle or CNO. A 794 difference score was calculated for each measure of interest between CNO and vehicle for each 795 animal, and the sham and DREADD groups were compared using independent t-tests or Mann-796 Whitney U test (#). To complete a full rotation, an animal must make consecutive turns that add up (cumulative turn 870 angle) to 360 degrees. Reversal (turning towards the opposite direction) at any point before the 871 cumulative angle reaches 360 degrees resets the cumulative angle to 0 and starts adding up 872 successive turns to the opposite direction (e.g considering panels C1-C4, the cumulative angle 873 counted towards a full rotation is -35 because the animal made a reversal, despite the head turn 874 angle sum being +85). Because of the effects of the reversal, we believe the turn angle sum is a 875 more accurate representation of turning behaviour than complete 360-degree rotations.