A motorized pellet dispenser to deliver high intensity training of the single pellet reaching and grasping task in rats

Accepted Manuscript Title: A motorized pellet dispenser to deliver high intensity training of the single pellet task in rats Authors: Abel Torres-Esp´ın, Juan Forero, Emma Schmidt, Karim Fouad, Keith K. Fenrich PII: DOI: Reference:

S0166-4328(17)30665-4 http://dx.doi.org/10.1016/j.bbr.2017.08.033 BBR 11052

To appear in:

Behavioural Brain Research

Received date: Revised date: Accepted date:

18-4-2017 11-7-2017 19-8-2017

Please cite this article as: Torres-Esp´ın Abel, Forero Juan, Schmidt Emma, Fouad Karim, Fenrich Keith K.A motorized pellet dispenser to deliver high intensity training of the single pellet task in rats.Behavioural Brain Research http://dx.doi.org/10.1016/j.bbr.2017.08.033 This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

A motorized pellet dispenser to deliver high intensity training of the single pellet task in rats

*Abel Torres-Espín a,b, *Juan Forero a,b, Emma Schmidt a, Karim Fouad a,b, Keith K. Fenrich a,b

a

Neuroscience and Mental Health Institute, b Department of Physical Therapy, Faculty

of Rehabilitation Medicine, 3-88 Corbett Hall, University of Alberta, Edmonton, AB T6E 2G4, Canada. *These authors contributed equally to this study

Corresponding author: Keith Fenrich Faculty of Rehabilitation Medicine 3-88 Corbett Hall University of Alberta Edmonton, AB T6E 2G4 CANADA Email: [email protected] Phone: 780 492 0938 Fax: 780 492 4429

Highlights: 

A low-cost 3D printed motorized pellet dispenser is described



Dispensers allow high-intensity training of the single pellet grasping task



Dispensers reduce workload and variability in results between experiments

1

ABSTRACT The single pellet reaching and grasping (SPG) task is widely used to study forelimb motor performance in rodents and to provide rehabilitation after neurological disorders. Nonetheless, the time necessary to train animals precludes its use in settings where high-intensity training is required. In the current study, we developed a novel highintensity training protocol for the SPG task based on a motorized pellet dispenser and a dual-window enclosure. We tested the protocol in naive adult rats and found 1) an increase in the intensity of training without increasing the task time and without affecting the overall performance of the animals, 2) a reduction in the variability within and between experiments in comparison to manual SPG training, and 3) a reduction in the time required to conduct experiments. In summary, we developed and tested a novel protocol for SPG training that provides higher-intensity training while reducing the variability of results observed with other protocols.

Abbreviations: SPG: single pellet reaching and grasping APP: automated pellet presenter CV: coefficient of variance SR: success Rate PPR: pellet presentation rate 1wM: 1 window Manual 1wM+: 1 window Manual extended time 2wM: 2 windows Manual 2wD: 2 windows Dispenser

Key words: Single pellet grasping, high-intensity, training, motor task, forelimb

2

1. INTRODUCTION Rehabilitative training is probably the most common approach used in the clinic to promote functional recovery after central nervous system (CNS) injury and/or disease. Nonetheless, there are still several unanswered questions about how rehabilitation translates to recovery, particularly in terms of the mechanisms involved in the recovery and how to optimize its effects. For instance, the amount of rehabilitative training and its intensity, defined as the number of repetitions of a task by time unit, have been associated to locomotor recovery after stroke [1,2] or spinal cord injury (SCI) [3,4] in humans. However, in the clinical setting the link between training intensity and recovery is not well stablish and its mechanisms are poorly understood. Surprisingly, although more research has been done in this topic in pre-clinical studies, the role of training intensity in recovery is still under-investigated. For example, in upper limb recovery and skilled reaching, the research in experimental stroke suggest that a minimal amount of training repetitions and intensity are necessary for recovery[5,6]. Nonetheless, direct evidence of how intensity affects performance in skilled reaching is scarce [7,8]. One approach to investigate how the amount of training affects forelimb recovery is to increase the amount of time spent training each day [8]. However, this approach increases the number of task repetitions per day but not necessarily its intensity. Moreover, in other neurological conditions such as experimental cervical SCI the importance of intensity has to be yet established. This knowledge gap makes it very challenging to translate findings, including promising rehabilitative treatments, from the pre-clinical to the clinical setting. Furthermore, since rehabilitative training is almost universally administered in the clinical setting to treat CNS injuries and/or disease, it should thus be part of any pre-clinical studies that aim to develop treatments for these conditions. A major limitation of rehabilitative training in experimental animal models is the large amount of time required to train the number of animals needed to obtain reliable and statistically meaningful results. This limitation is perhaps one of the main reasons rehabilitative training is not included in more pre-clinical studies, and thus new strategies for studying rehabilitative training are urgently needed. In pre-clinical animal models, a variety of skilled reaching tasks are widely used to study forelimb fine motor performance in rodents [9,10] and to provide task-specific rehabilitation after SCI [11–13], traumatic brain injury (TBI) [14], stroke [15], and neurodegenerative disease models [16]. Among these tasks are the Montoya staircase 3

test [17], the food tray task [18], the isometric pull task [19], the forelimb reaching task [20] and the single pellet reaching and grasping (SPG) task [10,21]. Among them, it is the SPG task (also named the skilled reaching task or single pellet reaching task [22,23]) that is largely used in studies involving rodents. Rats can be easily trained to perform the SPG task using movement patterns that mirror those observed in humans [24]. Moreover, several new measures have been developed for the SPG task to understand and study the different aspects of motor improvement after injury to the nervous system [25]. Even though the SPG task can be considered a highly useful approach for studying rat forelimb movements and fine motor control, and for understanding the mechanisms involved in recovery after injury to the CNS, it is probably the most tedious and time consuming of the forelimb reaching tasks. Despite these drawbacks, different approaches are commonly used to reduce the time and effort needed for SPG training and testing. For instance, the number of animals tends to be low (less than 6 per experimental group) [26] or performed in batches [13], hence increasing chances for the study to become underpowered. Also, the amount of training can be reduced to accommodate for time limitations, but this approach hinders the use of the SPG task as a rehabilitative training tool. Other approaches have been made to provide alternatives to increase the throughput of the SPG task. For instance, the task has been automated to reduce the time needed from the researcher when performing the experiment and to increase reproducibility [27,28]. These automated systems have been proven to be useful to study and analyze motor performance in the SPG task. However, since the animal still needs around 60 minutes to reach for a high amount of pellets (an average of 50), these systems still lack the ability to deliver highly intensive training. Recently, we developed a robotic system called the Automated Pellet Presentation (APP) system [29,30] to automate training and testing of rats in the SPG task before and after SCI. The APP system is attached to the home-cages of the rats, which allows them to train ad libitum 24 hours a day, 7 days a week. Rats trained with APP devices made 4-10 more attempts per day compared to rats trained using the standard manual SPG task. However, because APP rats could train throughout the day and night, attempts were often distributed over many hours, thus reducing the intensity of training despite the high attempt rate.

4

In this paper, we present a new protocol for the SPG task based on a motorized pellet dispenser that facilitates high-intensity training for 10 minutes per day. We found that high-intensity training in the SPG task is achievable in rats without increasing the training time. Moreover, the results obtained by training with the motor dispenser are more consistent between researchers. 2. MATERIAL AND METHODS 2.1 Animals Adult female Lewis rats (Charles River, 200-250g) were group-housed (4-5 animals per cage) and received water and food ad libitum. The animals were housed in a control environment of temperature, humidity and a light/dark cycle of 12/12h. During the training period, animals were food restricted (see below). The study was approved by the University of Alberta Animal Care and Use Committee, and complies with the Canadian Council for Animal Care an ARRIVE guidelines. 2.2 Experimental design This study aimed at developing and testing the efficacy of new pellet dispenser connected to a modified training enclosure for high-intensity training in the SPG task. Data derived from other experiments performed in our lab were used to compare the traditional training protocol (i.e., manual training using single-window enclosures), to a variation of the traditional training protocol (i.e., manual training using two-window enclosures), and the newly proposed protocol using a motorized pellet dispenser (i.e., pellet dispenser-based training using dual-window enclosures). Pellet dispenser (Fig. 1 and 2): Each dispenser consists of a pellet holder connected to the tip of an arm that can move up and down inside a container holding the pellets (pellet hopper). The pellet holder was made of a short plastic tube (6.4 mm outer diameter, 4.6 mm inner diameter and 6.4 mm in height). Pellets were presented at a distance of 20 mm from pellet centre to the inner wall of the enclosure. Dispensers were attached to the enclosure in one of two different lateral positions, each 5 mm to the left or right of the midline of the slit to accommodate the animal’s paw preference. The arm was moved by a micro servo motor (Hitec HS-82MG, Hitec, Poway, California, USA) attached to the back of the hopper and connected to the arm by a micro pushrod system (DU-BRO, #665 and #605, Wauconda, Illinois, USA). The motor was controlled by an

5

Atmel ATTiny84 microcontroller incorporated in a Servo Trigger board (SparkFun Electronics, WIG-13118, Niwot, Colorado, USA). The Servo Trigger board allowed fine controlling for the up and down position of the arm as well as the speed at which the arm moved between the two positions (for more details visit SparkFun website, http://www.sparkfun.com). Each motorized dispenser was connected to a custom-made control box (Fig. 2A) that provided power and a push button to manually activate the dispenser. Each control box could be connected to 3 independent dispensers and was powered using an AC/DC wall adapter (output of 9 V and 2.5 A) connected to a linear voltage regulator (output of +5 V and a maximum current rating of 1.5 A). Upon activation of the dispenser (i.e., by pushing the button), starting from the up position the arm moves down into the container to the down position and then immediately returns to the up position. In nearly all cases, during the upward motion a pellet becomes lodged in the pellet holder and is successfully presented to the rat. Presentation of a pellet takes less than 5 seconds from the moment the dispenser is activated with the push of the button until the moment the pellet is in place. The pellet hopper can hold approximately 1000 pellets. All pieces of the pellet dispenser and the control box were 3D printed using the MakerBot Replicator 2 3D printer (MakerBot, New York, USA). The 3D files for the dispenser can be obtained in https://drive.google.com/drive/folders/0B4tLAWCy1g1KNmU2V1JlcEQyMVE?usp=s haring. SPG enclosure: The SPG training enclosure consisted of an acrylic (0.7 cm thick) corridor (40 cm long, 14 cm wide, 45 cm tall) with one (single-window) or two (dualwindow, one per end) windows consisting each of a 1 cm wide and 10 cm tall vertical slit (Fig. 2A). The floor of the enclosure is made of a metal wire grid to prevent the animal eating pellets dropped inside the enclosure during training. An external shelf to hold the pellets or a pellet dispenser (depending on the training protocol, see below) was placed 3 cm away from the gridded floor (Fig. 2A). Pellets were presented at a distance of 20 mm from pellet centre to the inner wall of the enclosure. The specifications for the pellet placement are the same as the ones described before, although the task may have other names such as skilled reaching or single pellet reaching [22,23]. SPG training: In the first stage of training, rats were familiarized with the task. Briefly, rats were individually placed inside the training enclosure until they showed no signs of

6

exploration (1 or 2 days, 10 min per session), then pellets were presented on the shelf close to the slit. In the second stage of training, the pellets were presented progressively further from the slit until the final distance was reached. The second stage of training lasted between 5 to 7 days, 10 min per session. During this time, the preferred paw was determined by counting the number of times animals tried to reach with each paw. The third and final stage of SPG training started after the animals became familiarized with the task and their preferred paw was determined. The different training protocols used were: 1) 10–minute–sessions of manual training using single–window enclosures, 1 window Manual (1wM); 2) 15–minute–sessions of manual training using singlewindow enclosures, 1 window Manual extended time (1wM+); 3) 10-minute sessions of manual training using dual-window enclosures, 2 window Manual (2wM); and 4) 10minute-sessions of motorized training using dispensers attached to dual-window enclosures, 2 window Dispenser (2wD). Manual training sessions with single-window enclosures followed the training protocol described previously [12]. Briefly, animals were placed at the back of the box and a pellet was presented on the shelf. Once the rat had completed a grasp attempt, the trainer placed a sugar pellet on the enclosure floor at the back of the enclosure to encourage the rat to move to the back of the enclosure. Once the rat had shuttled to the back of the enclosure, a new pellet was placed on the shelf and the process was repeated for the entire session. Training using dual-window enclosures (manual or dispenser) followed the same protocol described above for single-window enclosures, but instead of encouraging the animal to move to the other side of the enclosure by placing a pellet on the floor, pellets were placed on the shelf on the other side (pellet holder when using dispensers). All animals were trained for 3–5 sessions per week for 4–6 weeks (see SPG data set description). To increase motivation in the task, banana flavoured sugar pellets (45 mg, TestDiet, 5TUT sucrose tab, St. Louis USA) were used for all training protocols. Additionally, food in the home-cages was limited the night before each training session to a measured amount of rat chow (10 g/rat), sufficient to allow a constant increase in the animal’s body weight over time. 2.3 SPG data recording and processing

7

Performance of animals trained on the 1wM protocol was recorded by the researcher on paper spreadsheets in real–time. Because of the higher intensity, the performance of animals trained on the 2wD protocol was recorded on video and analysed off–line. Video recording: two conventional cameras (JVC, GZ-MG50U and Panasonic, DMCFZ200) were placed on each side of the dual–window enclosures. Video acquisition was set at 30 fps with a resolution of 640 by 480 pixels. The cameras where situated outside of the enclosures and facing the slit at an angle that would allow an unobstructed view of the animals, windows, and shelves or dispensers. SPG analysis: the SPG task was marked as follows: a trial started when a pellet was presented and lasted until either the animal moved the pellet from the shelf/dispenser or it moved to the opposite end of the box. If no attempts were made, then the trial was considered incomplete and not counted. For each successful trial, attempts were defined as clear forward motions of the forepaw beyond the window slit and towards the pellet. Each attempt was classified as: 1) Reaching hit, when the animal reached for and touched the pellet; 2) Grasping hit, when the animal reached and grasped the pellet; or 3) Success (retrieve hit), when the animal reached, grasped and successfully retrieved the pellet through the slit and ate it. Each performance metric (i.e., reaching hit, grasping hit, and success) is presented as a percentage of the number of attempts (or number of presented pellets if indicated). 2.4 SPG data sets description All data sets for the single-window protocols included in this paper were obtained from experiments conducted in our laboratory during a period of 7 years. A total of 8 experiments and 200 animals (in 4 protocols of training) were included. The experiments were performed at separate times following the same protocol described above. For the 1wM protocol, 82 animals from 3 different experiments were included (Exp1: n = 38, 5 weeks of training; Exp2: n = 20, 6 weeks of training; Exp3: n = 24, 6 weeks of training). These experiments were performed by two different researchers in three different experiments. For the 1wM+ protocol, 30 animals from a single experiment were included (6 weeks of training). For the 2wM protocol, 28 animals from a single experiment were included (4 weeks of training). For the 2wD protocol, 60 animals from 3 different experiments were included (Exp1: n = 6, 6 weeks of training; Exp2: n = 30, 6 weeks of training; Exp3: n = 24, 5 weeks of training). For the data to be consistent across experiments, only data from the first 5 weeks of training from each 8

data set were included in the analysis. It is important to notice that data from the first week of training on the 2wM protocol was not recorded hence it is missing from the analysis. 2.5 Statistical analysis All the statistical methods were implemented in R [31] (version 3.2.4)using R studio [32] (version 1.0.136). All performance measures are represented as the mean per week. The coefficient of variance (CV) of each performance measure for each week was calculated as the mean divided by the standard deviation of that performance for that week and represented as a percentage. For statistical inference, all the results were analysed by fitting a linear-mixed model using the R package lme4 [33] considering the animal as the random effect. Time (weeks of training) and grouping variable (experiment) were considered as the fixed effect. ANOVA table and p-values of the fitted linear-mixed model were generated using the ANOVA function in R and multiple pairwise comparisons were performed using the R package lsmeans [34] adjusting the p-value by the Tukey method. All the p-values for the comparison shown in figures 3 to 6 are provided in the supplementary tables. An alpha value of 5% was considered as criteria for significance. 3. RESULTS 3.1 SPG training using the pellet dispenser We compared different training protocols to evaluate the efficacy of using the pellet dispenser for high-intensity SPG training (Fig. 3). The number of pellets presented each training session can be increased either by increasing the duration of each training session (i.e., increasing amount of training), by using dual-window enclosure instead of single-window enclosure (i.e., increasing the intensity of the training), or by automating the presentation of the pellet (i.e., presenting the pellet faster and more efficiently). The different training protocols used in this project were aimed at 1) comparing singleversus dual-window enclosure to increase training intensity, and 2) testing the effect of replacing manual training with the use of a motorized pellet dispenser. It is important to note that all the data used from the single-window approach was gathered from various experiments performed in our laboratory (see Methods) including both published [35] and unpublished results.

9

3.2 SPG training using dual-window dispensers increases training intensity The number of pellets presented each week consistently increased throughout training for all training protocols (main time effect: p < 0.001, Fig. 3B), with some protocols resulting in more pellets being presented than others (interaction effect: time x protocol, p < 0.001). Specifically, we found that increasing the intensity or amount of training, either by having longer training sessions (i.e., training for 15 min instead of 10 min, 1wM / 1wM+) or by training the animals in a dual-window enclosure (2wM) rather than a single-window enclosure (1wM), results in higher weekly average number of presented pellets. Moreover, the number of presented pellets was significantly higher in animals trained with the 2wD protocol compared to all the other protocols (after 5 weeks of training, 2wD compared to: 1wM, p < 0.001; 1wM+, p < 0.001; 2wM, p < 0.001). To quantify the intensity of training we calculated the pellet presentation rate (PPR), which we defined as the time required to present 20 pellets, a number commonly used when analysing fine motor function using the SPG task [25,26,36]. Statistical analysis of the average weekly PPR showed a consistent increase in training intensity (i.e., decrease in PPR) as training progressed in all the protocols (Fig. 3C; main time effect: p < 0.001), which was linked to the protocol being tested (interaction effect: time x protocol, p < 0.001). Except for the first week of training, post-hoc comparisons showed no significant differences in the PPR for animals trained on single-window enclosures (p-values in table S1). On the contrary, animals trained using dual-window enclosures showed a significant increase in training intensity over time compared to animals trained using single-window enclosures. Over the 5 weeks of training, animals trained with single-window enclosures increased their training intensity by 65% compared to an increase of 90% for animals trained on dual-window enclosures. Finally, to provide a global measure of intensity that combines the amount and rate of training, we divided the number of pellets presented by the number of minutes per training session. We found that animals trained with the 2wD protocol were presented about four times as many pellets per minute than animals trained using single-window enclosures (mean ± SEM for week 5: 2wD, 7.42 ± 0.59; 1wM, 1.82 ± 0.09; 1wM+, 1.96 ± 0.12; 2wM, training was stopped before week 5, p < 0.001 for each comparison). Taken together, these analyses show that animals trained with dispensers received more training at a higher intensity than the manually trained animals. 10

Success rate (see Methods), is a measure frequently used to evaluate performance in the SPG task (Fig. 3D). Statistical analysis showed that training had an effect on success rates and that these effects were dependent on the training protocol (interaction effect: time x protocol, p < 0.001). For instance, animals trained using single-window enclosures showed different trends depending on the duration of the training. Animals trained for 15 min per session (1wM+) showed a rapid increase in success rate followed by a slow reduction (starting after week 2). In comparison, animals trained for 10 minutes per session (1mW) showed a small reduction in success rate after 2 weeks in comparison to after one week of training and then a smooth increase in performance thereafter. Interestingly, animals trained in dual-window enclosures showed a consistent increase in the success rate as training progressed. Pairwise comparisons showed that animals trained in dual-window enclosures using the dispenser had higher success rates than animals trained in single-window enclosures, being significant for the 1wM+ group (p-values in table S1). Even though data from animals trained on the 2wM protocol was unavailable for weeks 1 and 5, a similar trend was observed for animals trained using the dispenser during weeks 2, 3 and 4. Although animals trained with the 2wM protocol showed a consistent increase in success rate over time, they appear to plateau in weeks 3 and 4, and it is uncertain whether they would have reached higher success rate if trained for a longer time. We used the coefficient of variance for the success rate (CVSR) as a measure of intersession constancy for the learning task over time (Fig. 3E). Consistent for all protocols, CVSR values started high at week 1, but dropped as training continued. However, the measured change depends on the training protocol (interaction effect: time x protocol, p < 0.001). Values measured for the CVSR decrease consistently with training in the 1wM, 2wM and 2wD protocols but not the 1wM+ protocol, for which a sharp increase in value was measured at week 3. A consistent reduction in the CVSR would be related to an improvement in the stability of the SPG task performance (measured in this case by the success rate), possibly linked to the material learning component of the task. At the end of training, CVSR values were lowest for the 2wD protocol compared to others, but without being statistically significant (p-values in table S1). 3.3 The pellet dispenser reduces the variability between and within experiments We studied the within-experiment variability for the 1wM and 2wD protocols using data collected from different experiments performed at different times by different 11

researchers (3 sets of data for each protocol). The different measures used to quantify within-experiment variability (for details, see Methods) are presented as the average per week for the 1wM protocol in Fig. 4, and for the 2wD protocol in Fig. 5. The results of training animals using the 1wM protocol are strongly influenced by the different conditions under which the experiments were run (e.g., the experimenter/trainer, the time when experiment was run). Statistical analysis showed a significant interaction effect between training time and experimental conditions, and an experimental main effect on all the measured parameters (number of presented pellets, number of attempts per trial, percentage of reaching hits, percentage of grasping hits and success rate; each case p < 0.003). Post-hoc comparisons confirmed significant single effect between experiments at different time points (Fig. 4). Thus, the repeatability of experiments based on the SPG task using the 1wM protocol is highly sensitive to the experimental conditions. Interestingly, results obtained from animals trained using the 2wD protocol were less sensitive to the experimental conditions than results from the 1wM protocol (c.f., Figs. 4 and 5). With the 2wD training protocol, main effect differences were shown only in the number of presented pellets (p = 0.028) and reaching hits (p = 0.033). However, there was a significant interaction effect of the experimental conditions (time x experiment) on the measures for the number of presented pellets, percentage of reaching hits, the percentage of grasping hits and the success rate but not for the number of attempts per pellet (p-values in table S2). This means that the experimental conditions can affect performance when training with the 2wD protocol. Nonetheless, post-hoc comparisons showed that the number of presented pellets was not significantly different between experiments for weeks 1 through 4 but only for week 5 (p-values in table S2). Post-hoc comparisons also showed that reaching hits were not significantly different between experiments at any time point. Thereby, the use of the 2wD protocol reduced the effect of the experimental conditions on the repeatability of experiments. In particular, the effect of the experimenter was greatly reduced when using the 2wD protocol compared to the 1wM protocol. Out of the three data sets collected for the 1wM data presented above, two were conducted by a single experimenter (Exp1 and Exp2) and yet significant differences were present in the results. Conversely, all three experiments using the 2wD protocol were conducted by different experimenters, yet the differences

12

between experiments were not significant for most measures, indicating that the use of the dispensers reduced the effect of the experimenter during training. We calculated the CV for number of presented pellets, number of attempts per pellet, reaching hits, grasping hits and success rate to analyse the inter-session consistency of the results depending on the training protocol over time (Fig. 6). The data from each set of experiments was grouped separately for the 1wM and the 2wD protocols. Statistical analysis revealed that the variability in all five measures decreased significantly with training independent of the protocol used (main effect of time: p < 0.001 for all five measures; significant interaction between time x protocol for the CV of number of presented pellets and grasping hits, p < 0.001, but not for the CV of attempts/pellet, p = 0.167, for the CV of reaching hits, p = 0.187, or for the CV of success rate, p = 0.386). In addition, a significantly lower CV for attempts/pellets, reaching hits, grasping hits and success rates, but not for number of presented pellets, was found for animals trained on the 2wD protocol compared to the 1wM protocol (main effect of protocol: CV of number of presented pellets, p = 0.108; CV of attempts/pellet, p < 0.001; CV of reaching hits, p = 0.003; CV of grasping hits, p < 0.001; CV of success rate, p < 0.001). Those main effects were translated to a single effect at different time point by the posthoc comparison (Fig. 6). Taken together, training using the dispenser resulted in a significant reduction in the variability of performance outcome measures. 3.4 The total time needed to perform an experiment is reduced when using the pellet dispenser The PPR using the pellet dispensers on dual-window enclosures is much faster than using manual presentation on single-window enclosures. Measures of the time necessary to conduct experiments using the 1wM, 1wM+ and 2wD protocols is presented in Table 1. Training using the dispenser allows the researcher to run 3 animals at a time, compared to 2 animals using manual pellet presentation (in our experience, manual training of more than 2 animals can be challenging and reduces the consistency of presenting pellets). Thus, training 30 animals requires 10 rounds on the 2wD protocol compared to 15 with the 1wM and 1wM+ protocols. In other words, training 30 animals on the 2wD protocol requires a total training time of 100 min (1 hr 40 min) compared to 150 min (2 hr 30 min) and 225 min (3 hr 45 min) when using the 1wM and 1wM+ protocols respectively. On the other hand, SPG task is commonly used to analyse motor function. For this, animals are presented with a standard number of pellets (e.g., n = 20, 13

Alaverdashvili and Whishaw 2013; García-Alías et al. 2015; Ishikawa et al. 2015) to have enough presented pellets to evaluate reaching and grasping movements with certainty. Using single-window protocols, Lewis rats usually require 4 to 5 weeks of training to achieve 20 pellets in a single 10 to 20-minute session. In contrast, that goal can be reached with less than 3 weeks of training using the 2wD protocol (see Fig. 3). Moreover, the time needed for those 20 pellets can be less than 4 minutes per session with 5 weeks of training using the 2wD protocol. Thus, testing 30 animals in the 20 pellets test requires around 3 h 15 min using the 1wM 10' protocol or 4 h 30 min using the 1wM 15' protocol. Using the 2wD protocol, training 30 animals takes around 2 h 10 min. Such a reduction in training and testing time will significantly reduce the time needed to execute the experiment. That reduction may result in less variability in the task performance by shortening the time window for training in each session. 4. DISCUSSION Pre-clinical research in animal models commonly uses the SPG task for evaluating fine motor function and, more recently, to evaluate the effect of rehabilitative training on recovery after injury to the CNS such as SCI, stroke and TBI [11–15]. However, the standard manual SPG task can be tedious and time consuming, limiting its usability in rehabilitative training experiments with a small number of animals. Moreover, the standard protocol does not allow for high intensity training and therefore new protocols of training are necessary. Here we present a novel design of a high-intensity SPG training protocol using a motorized pellet dispenser on dual-window SPG enclosures. We demonstrated that Lewis rats can receive high-intensity training using the pellet dispenser on dual-window enclosures and achieve performance outcomes that are comparable to those obtained from manual training but show less variability and require less time from the experimenters. 4.1 High intensity training using the pellet dispenser The standard manual SPG protocol allows for a limited training intensity. We showed how rats trained in a manual one-window enclosure reach on average 18 and 30 pellets in 10 and 15 minute sessions respectively, which results in no more than 2 pellets per minute. These values may be close to the ceiling effect associated with this protocol. Even when the one-window task is , a condition under which one would expect a higher efficiency in presenting pellets, the reported number of pellets presented per minute of 14

training was still low [27,28] (authors reported on average 50 pellets presented over a 60 minutes session; an estimated rate of 0.83 pellets per minute). Thus, the reduction in human intervention during the SPG task when using those automatic systems does not result in a high intensity training protocol. In addition, completely automatic systems are complex and sophisticated, hindering their implementation in rehabilitation experiments with high numbers of animals. The number of animals that can be trained per session can be increased by attaching an automatic pellet dispenser to the rats’ home cage. We have previously determined the feasibility of training rats using an automated, ad libitum, APP system. The APP system was found to be useful for training un-injured rats on a SPG task [30]and to promote recovery of forelimb function after acute SCI [29]However, the APP system uses a variation of the traditional training paradigm by allowing animals to train when they want (i.e., ad libitum) instead of following a set schedule (i.e., training within a set time-window). As a result, training can be distributed throughout a 24-hour period, thus potentially reducing the intensity of training in comparison to the new two-window dispenser protocol. The pellet dispenser described here is an intermediate approach that still allows for a short (10 min) daily high-intensity training session under a paradigm more comparable to that of the standard SPG task, and allows the researcher to perform and control the training during each session. In the present study, we demonstrated that combining dual-window SPG enclosures with the use of the pellet dispenser increases the number of pellets presented to the animals in each session. An approach to increasing the number of presented pellets is to extend the duration of the training session (1wM+ animals). Although extending the duration of each session for 1wM+ animals did not result in increased training intensity (reflected by the PPR), it did result in a slight increase in the overall number of pellets presented per session, resulting in an increased success rate. Therefore, increasing the amount of training without increasing the intensity produced a high SPG performance, supporting the idea that the quantity of training has an influence on task learning. Although animals did reach peak performance sooner with longer sessions (15 min instead of 10 min), these rapid gains were followed by a consistent decrease in performance. This result is likely related to a reduction in motivation to perform the task rather than fatigue (such as that observed in patients, [37] ) since this effect was not found in 2wD animals. Animals trained using the novel 2wD protocol were presented

15

with 3 to 4 times the number of pellets in only 10 minutes. Although animals trained with this novel protocol showed lower success rates in the first weeks of training, these values increased steadily and, after 5 weeks of training, exceeded the values observed with the traditional protocol. Thus, changing the protocol influenced the rate at which the animals learned the SPG task, but had no effect on their performance by the end of the training period. Acquiring a reaching task by rats is characterized by a high learning rate (i.e. over time improvement in the performance) at the beginning of the training process with a posterior decay [38]. This pattern of motor learning is reflected by a progressive reduction in the inter-session variability of the task performance over time. In our study, large differences between consecutive days of training in the performance, showed by a high CV, at the beginning of training were followed by a reduction of those differences. In comparison with the animals trained using the 1wM protocol, animals trained using the 2wD protocol showed lower CV in the three measures (reaching hit, grasping hit and success rate) of the SPG performance. That may reflect a faster inter-session learning rate by animals trained at high intensities than those trained at low intensities [38]. 4.2 Reducing experimental time and variability Although the capacity to deliver high-intensity training is probably the most relevant feature of using the pellet dispenser on the SPG task, we found other important advantages. For instance, rats that are well-trained using the 1wD protocol need between 10 and 20 min to reach the 20 attempted pellets mark. In contrast, animals trained with the pellet dispenser on dual-window enclosures required less than 4 min to be presented with the same number of pellets. In addition, the simplicity of presenting pellets by pressing a button, instead of placing pellets manually, allows a single researcher to train (or test) a group of animals more quickly. Overall, the novel dispenser protocol allows an increase in the efficiency of training and, because the number of animals trained simultaneously can be increased, there is a reduction in the time the researcher is exposed to the animals during training sessions. It has been shown that animals’ activity is affected by the time of the day at which behavioural testing is done [39]. Therefore, animals trained at the beginning of the session might not perform similarly to those trained at the end (e.g., 2 or 3 hours later). The shortening of the training time window by using 2wD protocol could then have an influence in reducing 16

the variability by allowing animals to be trained in a similar time of the day. Moreover, we observed that experiments conducted by different researchers using the novel 2wD protocol produced more consistent results than experiments conducted using the 1wM training even when done by a single researcher. Thus, the use of the dispenser might reduce data variability by combining a very consistent way of presenting pellets to the animals with a reduced time window for training in each session, which results in an increase in the reproducibility between experiments. Traditionally, data from the SPG task is collected in real-time annotation. However, simultaneously tracking the performance of 3 animals on dual-window enclosures is too complex for a single researcher using the dispenser or dual-window enclosures. To overcome this hurdle, we used video recording to collect data. Although this could be seen as a drawback, since it is necessary to analyse the videos after training, video recording offers various advantages. For instance, having video recordings of the experiment allows the researcher to better differentiate performance details (e.g., repetitive attempts, grasping technique) that occur too rapidly to detect in real-time. Also, because of the higher number of presented pellets, larger samples of data would be analysed thus the computed metrics (over 3-4 times more presented pellets) would better account for performance variability within session for each animal. 4.3 Implications for rehabilitation after CNS injuries Task-specific training of reaching has been demonstrated to induce functional recovery after acute or subacute CNS injuries [11,12,14,15]. However, the recovery achieved by skilled reaching training in the chronic scenario is limited or non-existent without a combined treatment [13,40]. One plausible reason for the poor efficacy of the existent training protocols in chronic CNS injuries to produce recovery is that they are poorly suited to deliver high-intensity training. Early after CNS injuries there is a period of increased plasticity, which facilitates the rewiring of spared neural circuitry [41,42]. Through rehabilitative training, rewiring can be directed towards meaningful functional recovery [42,43]. However, the window of opportunity for rehabilitative-induced recovery is time-dependent. In some cases, training too early could exacerbate the damage while delaying the onset of treatment for too long can reduce its benefits [41,44]. Thus, the standard reaching protocols may provide enough training intensity to induce recovery during that window of high plasticity, but may not be enough to drive or induce neural changes when the system is less plastic. This concept may be behind 17

the importance of the training intensity to induce recovery in humans, where the rehabilitation is usually started late (i.e. weeks) after CNS injury [45,46]. In that case, we provide here a tool to deliver high-intensity SPG training to experimental animals. This method will allow researchers to study the importance of intensity to induce recovery after CNS injuries. 4.4 Conclusion In conclusion, we presented a novel device and training protocol that allows animals to train with high-intensity in the SPG task, that does not affect performance, reduces the variability between and within experiments, and reduces the time researchers expend training animals. 5. ACKNOWLEDGMENTS This study was supported by operating grants from the Wings for Life foundation (RES0025323) and the Canadian Institute for Health Research (CIHR, RES0011338). KKF was supported by fellowships from CIHR and Alberta Innovates: Health Solutions. We thank Romana Vavrek and Pamela Raposo for technical assistance. 6. REFERENCES [1]

M. Pohl, J. Mehrholz, C. Ritschel, S. Rückriem, Speed-Dependent Treadmill Training in Ambulatory Hemiparetic Stroke Patients, Stroke. 33 (2002) 553–558. doi:10.1161/hs0202.102365.

[2]

K.J. Sullivan, B.J. Knowlton, B.H. Dobkin, Step training with body weight support: Effect of treadmill speed and practice paradigms on poststroke locomotor recovery,

Arch.

Phys.

Med.

Rehabil.

83

(2002)

683–691.

doi:10.1053/apmr.2002.32488. [3]

K.A. Leech, C.R. Kinnaird, C.L. Holleran, J. Kahn, T.G. Hornby, Effects of Locomotor Exercise Intensity on Gait Performance in Individuals With Incomplete Spinal Cord Injury., Phys. Ther. (2016) 1–37. doi:10.2522/ptj.20150646.

[4]

M. Wessels, C. Lucas, I. Eriks-Hoogland, S. De Groot, Body weight-supported gait training for restoration of walking in people with an incomplete spinal cord injury: A systematic review, Assist. Technol. Res. Ser. 26 (2010) 297–299. doi:10.3233/978-1-60750-080-3-297.

[5]

J.A. Kleim, T.A. Jones, Principles of Experience-Dependent Neural Plasticity: 18

Implications for Rehabilitation After Brain Damage, J. Speech Lang. Hear. Res. 51 (2008) S225. doi:10.1044/1092-4388(2008/018). [6]

R. Teasell, J. Bitensky, K. Salter, N.A. Bayona, The Role of Timing and Intensity of Rehabilitation Therapies, Top. Stroke Rehabil. 12 (2005) 46–57. doi:10.1310/ETDP-6DR4-D617-VMVF.

[7]

C.L. MacLellan, M.B. Keough, S. Granter-Button, G.A. Chernenko, S. Butt, D. Corbett, A Critical Threshold of Rehabilitation Involving Brain-Derived Neurotrophic Factor Is Required for Poststroke Recovery, Neurorehabil. Neural Repair. 25 (2011) 740–748. doi:10.1177/1545968311407517.

[8]

J.A. Bell, M.L. Wolke, R.C. Ortez, T.A. Jones, A.L. Kerr, Training Intensity Affects Motor Rehabilitation Efficacy Following Unilateral Ischemic Insult of the Sensorimotor Cortex in C57BL/6 Mice, Neurorehabil. Neural Repair. 29 (2015) 590–598. doi:10.1177/1545968314553031.

[9]

E. Azim, J. Jiang, B. Alstermark, T.M. Jessell, Skilled reaching relies on a V2a propriospinal

internal

copy

circuit,

Nature.

508

(2014)

357–363.

doi:10.1038/nature13021. [10] G.A.S. Metz, I.Q. Whishaw, Skilled reaching an action pattern: Stability in rat (Rattus norvegicus) grasping movements as a function of changing food pellet size, Behav. Brain Res. 116 (2000) 111–122. doi:10.1016/S0166-4328(00)00245-X. [11] A. Krajacic, M. Ghosh, R. Puentes, D.D. Pearse, K. Fouad, Advantages of delaying the onset of rehabilitative reaching training in rats with incomplete spinal cord injury.,

Eur.

J.

Neurosci.

29

(2009)

641–51.

doi:10.1111/j.1460-

9568.2008.06600.x. [12] J. Girgis, D. Merrett, S. Kirkland, G. a S. Metz, V. Verge, K. Fouad, Reaching training in rats with spinal cord injury promotes plasticity and task specific recovery., Brain. 130 (2007) 2993–3003. doi:10.1093/brain/awm245. [13] D. Wang, R.M. Ichiyama, R. Zhao, M.R. Andrews, J.W. Fawcett, Chondroitinase combined with rehabilitation promotes recovery of forelimb function in rats with chronic

spinal

cord

injury.,

J.

Neurosci.

31

(2011)

9332–44.

doi:10.1523/JNEUROSCI.0983-11.2011. [14] A. DeAnna L., F. Lindsay, S. Lance, A. Pevtsov, K. McDonough, J. Stamschror, 19

T.A. Jones, D.A. Kozlowski, Combining Multiple Types of Motor Rehabilitation Enhances Skilled Forelimb Use Following Experimental Traumatic Brain Injury in

Rats,

Neurorehabil.

Neural

Repair.

29

(2015)

989–1000.

doi:10.1177/1545968315576577. [15] A. Schmidt, J. Wellmann, M. Schilling, J.-K. Strecker, C. Sommer, W.-R. Schäbitz, K. Diederich, J. Minnerup, Meta-analysis of the efficacy of different training strategies in animal models of ischemic stroke., Stroke. 45 (2014) 239–47. doi:10.1161/STROKEAHA.113.002048. [16] P. Vergara-Aragon, C.L.R. Gonzalez, I.Q. Whishaw, A Novel Skilled-Reaching Impairment in Paw Supination on the “Good” Side of the Hemi-Parkinson Rat Improved with Rehabilitation, J. Neurosci. 23 (2003) 579–586. [17] C.P. Montoya, L.J. Campbell-Hope, K.D. Pemberton, S.B. Dunnett, The “staircase test”: a measure of independent forelimb reaching and grasping abilities in rats, J. Neurosci. Methods. 36 (1991) 219–228. doi:10.1016/0165-0270(91)90048-5. [18] I.Q. Whishaw, W.T. O’connor, S.B. Dunnett, W.T.O. Connor, S.B. Dunnett, The contributions of motor cortex, nigrostriatal dopamine and caudate-putamen to skilled

forelimb

use

in

the

rat,

Brain.

109

(1986)

805–843.

doi:10.1093/brain/109.5.805. [19] S.A. Hays, N. Khodaparast, A.M. Sloan, D.R. Hulsey, M. Pantoja, A.D. Ruiz, M.P. Kilgard, R.L. Rennaker, The isometric pull task: A novel automated method for quantifying forelimb force generation in rats, J. Neurosci. Methods. 212 (2013) 329–337. doi:10.1016/j.jneumeth.2012.11.007. [20] G.W. Schrimsher, P.J. Reier, Forelimb motor performance following cervical spinal cord contusion injury in the rat, Exp. Neurol. 117 (1992) 287–298. http://www.ncbi.nlm.nih.gov/pubmed/1397165. [21] I.Q. Whishaw, S.M. Pellis, The structure of skilled forelimb reaching in the rat: A proximally driven movement with a single distal rotatory component, Behav. Brain Res. 41 (1990) 49–59. doi:10.1016/0166-4328(90)90053-H. [22] M. Alaverdashvili, A. Foroud, D.H. Lim, I.Q. Whishaw, “Learned baduse” limits recovery of skilled reaching for food after forelimb motor cortex stroke in rats: A new analysis of the effect of gestures on success, Behav. Brain Res. 188 (2008)

20

281–290. doi:10.1016/j.bbr.2007.11.007. [23] J.E. McKenna, I.Q. Whishaw, Complete compensation in skilled reaching success with associated impairments in limb synergies, after dorsal column lesion in the rat., J. Neurosci. 19 (1999) 1885–1894. [24] L.A.R. Sacrey, M. Alaverdashvili, I.Q. Whishaw, Similar hand shaping in reaching-for-food (skilled reaching) in rats and humans provides evidence of homology in release, collection, and manipulation movements, Behav. Brain Res. 204 (2009) 153–161. doi:10.1016/j.bbr.2009.05.035. [25] M. Alaverdashvili, I.Q. Whishaw, A behavioral method for identifying recovery and compensation: Hand use in a preclinical stroke model using the single pellet reaching

task,

Neurosci.

Biobehav.

Rev.

37

(2013)

950–967.

doi:10.1016/j.neubiorev.2013.03.026. [26] Y. Ishikawa, S. Imagama, T. Ohgomori, N. Ishiguro, K. Kadomatsu, A combination of keratan sulfate digestion and rehabilitation promotes anatomical plasticity after rat spinal cord injury, Neurosci. Lett. 593 (2015) 13–18. doi:10.1016/j.neulet.2015.03.015. [27] D.J. Ellens, M. Gaidica, A. Toader, S. Peng, S. Shue, T. John, A. Bova, D.K. Leventhal, An automated rat single pellet reaching system with high-speed video capture,

J.

Neurosci.

Methods.

271

(2016)

119–127.

doi:10.1016/j.jneumeth.2016.07.009. [28] C.C. Wong, D.S. Ramanathan, T. Gulati, S.J. Won, K. Ganguly, An automated behavioral box to assess forelimb function in rats, J. Neurosci. Methods. 246 (2015) 30–37. doi:10.1016/j.jneumeth.2015.03.008. [29] K.K. Fenrich, May Zacincte, Torres-Espín Abel, Forero Juan, Bennetta J. David, Fouad Karim, Single pellet grasping following cervical spinal cord injury in adult rat using an automated full-time training robot, Behav. Brain Res. 22 (2015) 134– 139. doi:10.1177/0963721412473755.Surging. [30] K. Fenrich, Z. May, C. Hurd, C.E. Boychuk, J. Kowalczewski, D.J. Bennett, I.Q. Whishaw, K. Fouad, Improved single pellet grasping using automated ad libitum full-time

training

robot.,

Behav.

Brain

Res.

281

(2015)

137–48.

doi:10.1016/j.bbr.2014.11.048.

21

[31] R Development Core Team, R: A Language and Environment for Statistical Computing, R Found. Stat. Comput. Vienna Austria. 0 (2016) {ISBN} 3-90005107-0. doi:10.1038/sj.hdy.6800737. [32] Rs.

Team,

RStudio:

Integrated

Development

for

R,

(2015).

http://www.rstudio.com/. [33] D. Bates, M. Mächler, B.M. Bolker, S.C. Walker, Fitting linear mixed-effects models using lme4, J. Stat. Softw. 67 (2015) 1–48. doi:10.18637/jss.v067.i01. [34] R. V Lenth, Least-Squares Means: The R Package lsmeans, J. Stat. Softw. 69 (2016) 1–33. doi:10.18637/jss.v069.i01. [35] D. Wei, C. Hurd, D. Galleguillos, J. Singh, K.K. Fenrich, C.A. Webber, S. Sipione, K. Fouad, Inhibiting cortical protein kinase A in spinal cord injured rats enhances efficacy of rehabilitative training, Exp. Neurol. 283 (2016) 365–374. doi:10.1016/j.expneurol.2016.07.001. [36] G. García-Alías, K. Truong, P.K. Shah, R.R. Roy, V.R. Edgerton, Plasticity of subcortical pathways promote recovery of skilled hand function in rats after corticospinal and rubrospinal tract injuries, Exp. Neurol. 266 (2015) 112–119. doi:10.1016/j.expneurol.2015.01.009. [37] M.J. Castro, D.F. Apple Jr., R.S. Staron, G.E. Campos, G.A. Dudley, Influence of complete spinal cord injury on skeletal muscle within 6 mo of injury, J. Appl. Physiol.

86

(1999)

350–358.

http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&do pt=Citation&list_uids=9887150. [38] M.M. Buitrago, T. Ringer, J.B. Schulz, J. Dichgans, A.R. Luft, Characterization of motor skill and instrumental learning time scales in a skilled reaching task in rat, Behav. Brain Res. 155 (2004) 249–256. doi:10.1016/j.bbr.2004.04.025. [39] P. Saibaba, G.D. Sales, G. Stodulski, J. Hau, Behaviour of rats in their home cages: daytime variations and effects of routine husbandry procedures analysed by time sampling

techniques,

Lab.

Anim.

30

(1996)

13–21.

doi:10.1258/002367796780744875. [40] A.S. Wahl, W. Omlor, J.C. Rubio, J.L. Chen, H. Zheng, A. Schröter, M. Gullo, O. Weinmann, K. Kobayashi, F. Helmchen, B. Ommer, M.E. Schwab, Asynchronous 22

therapy restores motor control by rewiring of the rat corticospinal tract after stroke., Science. 344 (2014) 1250–5. doi:10.1126/science.1253050. [41] A.-S. Wahl, M.E. Schwab, Finding an optimal rehabilitation paradigm after stroke: enhancing fiber growth and training of the brain at the right moment, Front. Hum. Neurosci. 8 (2014) 1–13. doi:10.3389/fnhum.2014.00381. [42] K. Fouad, A. Krajacic, W. Tetzlaff, Spinal cord injury and plasticity: opportunities and

challenges.,

Brain

Res.

Bull.

84

(2011)

337–42.

doi:10.1016/j.brainresbull.2010.04.017. [43] M. El Amki, P. Baumgartner, O. Bracko, A.R. Luft, S. Wegener, Task-Specific Motor Rehabilitation Therapy After Stroke Improves Performance in a Different Motor

Task:

Translational

Evidence,

Transl.

Stroke

Res.

(2017).

doi:10.1007/s12975-016-0519-x. [44] J. Biernaskie, G. Chernenko, D. Corbett, Efficacy of rehabilitative experience declines with time after focal ischemic brain injury., J. Neurosci. 24 (2004) 1245– 54. doi:10.1523/JNEUROSCI.3834-03.2004. [45] C. Granger, S. Markello, J. Graham, A. Deutsch, K. Ottenbacher, The uniform data system for medical rehabilitation: report of patients with stroke discharged from comprehensive medical programs in 2000-2007, Am. J. Phys. Med. Rehabil. 93 (2014) 231–244. doi:10.1097/PHM.0b013e3182a92c58. [46] K. Putman, L. De Wit, W. Schupp, H. Beyens, E. Dejaeger, W. De Weerdt, H. Feys, W. Jenni, F. Louckx, M. Leys, Inpatient stroke rehabilitation: A comparative study of admission criteria to stroke rehabilitation units in four European centres, J. Rehabil. Med. 39 (2007) 21–26. doi:10.2340/16501977-0006.

23

Figures

Fig. 1. Details of the motorized pellet dispenser. (A and B) infographics showing two different views of the 3D model of the pellet dispenser. (C) an exploded view of the 3D model to better visualize each 3D printed component. Left and right refers to the piece location considering the closest edge to the final position as the front and the motor site as the back of the dispenser. (D) a front view of the SparkFun servo trigger board used to control the motor (www.sparkfun.com, image under CC BY-NC-SA 3.0 licence https://creativecommons.org/licenses/by-nc-sa/3.0). 3 on-board potentiometers are used to adjust the start/end position (potentiometers A and B) and the time transition between both positions (potentiometer T). (E) a close view of the motor and the motor holder on the back of the dispenser. (F) detailed views of the assembled motor and the connection to the arm using a pushrod system. (G) details of the plastic piece used as pellet holder in the tip of the arm. Dimensions in mm.

24

Fig. 2. Description of SPG training enclosure and pellet dispenser. (A) lateral and front perspective views, and close-up view of a 3D model of the dispenser and a dualwindow training enclosure with the most relevant length measures in cm. (B) Picture of an assembled dispenser showing the back and the placement of the electronic components (i.e., motor and motor controller with the connectors). (C) Picture showing a dual-window training enclosure with two dispensers, one for each window, and the control boxes with the buttons to trigger the motors. In a regular training session, 3 enclosures, each with 2 dispensers, are placed together and controlled by the two control boxes (notice that each box controls the 3 dispensers on one side), allowing the training 3 animals at the time. (D-G) details of the sequence of the dispenser arm collecting one pellet and lifting it up to the top of the lid. Around 1000 pellets can be placed in the hopper (a small number of pellets was placed for illustrative purposes).

25

Fig. 3. Training protocol comparison showing differences in the number of pellets per session. (A) infographic showing the three different setups used to compare the 4 training protocols. (B) graph showing the average number of pellets presented per session at each week of training for the 4 training protocols. An increase of 3 to 4 times in intensity of training (nº of presented pellets) was observed comparing 2wD training with 1wM and 1wM+ training protocols. This can be also appreciated in the weekly mean of the pellets presentation rate (PPR), which was calculated as the time (in minutes) to reach 20 presented pellets per session (C) The ordinate axis is presented as the log10 of the PPR to facilitate visualization. The dotted line is used as a reference indicating 10 min of training time. The weekly means of the success rate calculated by number of presented pellets (D) and the percentage of coefficient of variation for the success rate (CVSR) for

26

each week of training (E) showed some differences in the performance and performance variation, respectively, between training protocols. 1wM: 10 minute sessions of manual training in single-window enclosures, 1wM+: 15 minute sessions of manual training in single-window enclosures, 2wM: 10 minute sessions of manual training in dual-window enclosures, 2wD: 10 minute sessions of motorized dispenser training in dual-window enclosures. In each graph the significant pair-comparisons (p < 0.05) are represented by letters corresponding to the pairs indicated in the group differences legend. The mean ± SEM are represented.

Fig. 4. A comparison of 3 different experiments using the 10 min single-window manual training protocol reveals inter-experimental variability. This figure shows the weekly mean of the number of presented pellets (A), the ratio of attempts per presented pellet (B), the percentage of reaching hits (C), the percentage of grasping hits (D), and the success rate calculated by number of attempts (E) for the 3 analysed experiments conducted using the 10min single-window manual training protocol (1wM). Significant differences were observed between the 3 analysed experiments in all the 27

measures (A-E) indicating the susceptibility of the 1wM training protocol to experimental variations. In each graph the significant pair-comparisons (p < 0.05) are represented by letters corresponding to the pairs indicated in the group differences legend. The mean ± SEM are represented.

Fig. 5. A comparison of 3 different experiments using the dual-window dispenser training protocol reveals the stability of the method. This figure shows the weekly mean of the number of presented pellets (A), the ratio of attempts per presented pellet (B), the percentage of reaching hits (C), the percentage of grasping hits (D) and the success rate calculated by number of attempts (E) for the 3 analysed experiments conducted using the dual-window dispenser training protocol (2wD). Different to what we observed using 1wM protocol (see fig. 4), all the measures (A-E) obtained by training using the 2wD protocol were similar between experiments and fewer statistical differences were found. This shows that results are more stable and less susceptible to experimental changes when using the 2wD protocol compared to the 1wM protocol. In 28

each graph the significant pair-comparisons (p < 0.05) are represented by letters corresponding to the pairs indicated in the group differences legend. The mean ± SEM are represented.

Fig. 6. 2wD training protocol shows higher inter-training session stability than the 1wM training protocol. This figure shows the comparison of the weekly coefficient of variation (CV) for the number of presented pellets (A), the number of attempts per pellet (B), the percentage of reaching hits (C), the percentage of grasping hits (D) and success rate (E) between the 10min single-window manual (1wM) and dual-window dispenser (2wD) protocols. In both training protocols, the CV is decreased over time, indicating a reduction in the variability of the SPG performance with the accumulation of training. Regarding number of presented pellets, 2wD protocol showed a lower CV one week after training onset than protocol 1wM. For the rest of measures (B-E), a lower weekly CVs were observed by training using the 2wD protocol in comparison to 1wM protocol. This indicates that there is less variation in SPG performance between the different training sessions of each week using 2wD than 1wD protocols. *p < 0.05. **p < 0.01, ***p < 0.001. The mean ± SEM are represented.

29

Table 1:

Training Protocol

nº animals that

Time per

can be trained

session

simultaneously

(min)

Training

nº rounds

time for

per

30 animals

30

(min)*

animals**

Time between rounds (3 min per round)

Total time training per day (min)

Mean time to

Total time

obtain 20

to test

trials per

PPR for 30

animal

animals

(min)

(min)

1wM 10'

10

2

150

15

45

195

17.5

307.5

1wM 15'

15

2

225

15

45

270

12.92

238.8

2wD

10

3

100

10

30

130

3.75

67.5

*Calculated as:

(𝑇𝑟𝑎𝑖𝑛𝑖𝑛𝑔 𝑡𝑖𝑚𝑒 𝑝𝑒𝑟 𝑠𝑒𝑠𝑠𝑖𝑜𝑛)∗ (𝑛° 𝑜𝑓 𝑎𝑛𝑖𝑚𝑎𝑙𝑠) 𝑛° 𝑜𝑓 𝑎𝑛𝑖𝑚𝑎𝑙𝑠 𝑡ℎ𝑎𝑡 𝑐𝑎𝑛 𝑏𝑒 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑠𝑖𝑚𝑢𝑙𝑡𝑎𝑛𝑒𝑜𝑢𝑠𝑙𝑦

**Calculated as:

𝑛° 𝑜𝑓 𝑎𝑛𝑖𝑚𝑎𝑙𝑠 𝑛° 𝑜𝑓 𝑎𝑛𝑖𝑚𝑎𝑙𝑠 𝑡ℎ𝑎𝑡 𝑐𝑎𝑛 𝑏𝑒 𝑡𝑟𝑎𝑖𝑛𝑒𝑑 𝑠𝑖𝑚𝑢𝑙𝑡𝑎𝑛𝑒𝑜𝑢𝑠𝑙𝑦

30