The functional contribution of the SC in specific behavior has been investigated in a variety of experiments, including electrophysiological recording, inactivation and lesion approaches (Binns 1999; Huberman and Niell 2011) but little has been done at a more integrated level in animal models with congenital defects.
We first asked whether the modified collicular retinotopy affects visual acuity using the visual cliff test, which measures visual depth perception in rodents. Mice from all three experimental groups spent significantly more time on the opaque side compared to the cliff side (side: F1,18 = 10.15, p = 0.005; Fig. 2a) and stepped earlier onto the opaque side than onto the cliff side (side: F1,18 = 16.61, p < 0.001; Fig. 2b) indicating normal visual perception. There was no significant difference between genotypes for the latency to step down and the time spent on either the checkered side or the cliff side (no genotype effect or genotype × side interaction). We next tested visual acuity by stimulating and measuring the optokinetic reflex (OKR). This reflex mediates compensatory head motions elicited by moving full-field visual stimuli, to maintain a constant image on the retina. Mice from all three genotypes showed similar threshold values for the minimum contrast that triggers an OKR at spatial frequencies ranging from 0.064 to 0.272 cycles/degree (Fig. 2c). Together, these results indicated normal visual acuity in EphA3KI/KI and EphA3KI/+ mice.
General locomotor activity, sensory motor coordination and circadian rhythm
We next tested locomotor activity using horizontal cage activity and wheel running. Mice of each experimental group showed a similar decrease in locomotor activity over the course of a 3-h session corresponding to habituation to the new cage (15-min block: F11,198 = 55.17, p < 0.0001; Fig. 3a) and no significant effect of the genotype was observed in total wheel running activity, all three genotypes showing normal rhythmic activity (Fig. 4b, Online resource 2). The key role of the SC in the integration of sensorimotor modalities led us to test sensorimotor coordination. All three genotypes underwent the beam walking test and showed similar latencies to leave the start segment (genotype x trial: F6,54 = 0.65, p > 0.10, not shown) and to reach the platform, which decreased significantly during subsequent trials (trial: F3,54 = 16.48, p < 0.0001; Fig. 3b). Sensorimotor coordination and latency to leave the start segment were similar among genotypes. Moreover, we tested whether the running activity of knock-in mice follows light-entrained and endogenous circadian patterns. All three genotypes showed similar running activity in 12 h light–dark and dark–dark cycles with similar endogenous period (WT: 23.57 ± 0.26 h, EphA3KI/+: 23.76 ± 0.35 h and EphA3KI/KI: 23.69 ± 0.26 h; Fig. 4). Together these results indicate normal locomotor activity, sensory motor processing and circadian activity in EphA3KI/KI and EphA3KI/+ animals.
Visuo-spatial orientation and memory
We then tested vision and motor skills using the Morris water maze visible platform test, where mice must locate a cue at close range, and swim toward it. After 2 days of habituation, mice were tested for their performance in reaching a visible platform. Swim speed and distance were measured in four trials. Swim speed remained stable and similar for all groups. Swimming distance was similarly reduced among all groups over the four consecutive trials (trial: F3,54 = 16.07, p < 0.0001). No significant difference was observed among genotypes or genotype × trial interactions (Fig. 5a). Next we used a variant of the Morris water maze test where the platform is hidden to evaluate visuo-spatial learning and memory. Here, mice must find the hidden platform based on distant visual cues outside the pool. Over the course of the four training days, mice of all three genotypes showed similar swim speeds and learned the position of the hidden platform equally well (day: F3,54 = 20.67, p < 0.0001; Fig. 5b). No difference was observed between genotypes, suggesting that EphA3KI/+ and EphA3KI/KI animals are able to learn a task requiring visuo-spatial orientation abilities. In a probe test performed 24 h later, all mice showed a clear bias toward the target quadrant where they spent significantly more time than the 15-s chance level (WT: t
6 = 6.68, p = 0.0005, EphA3KI/+: t
6 = 4.62, p = 0.004; EphA3KI/KI: t
6 = 6.01, p = 0.001; Fig. 5c). Taken together, these results indicated normal visuo-spatial orientation, preserved motivation to reach a visible and hidden platform and intact spatial learning and memory in EphA3KI/KI and EphA3KI/+ mice.
Anxiety, response inhibition
As the behavioral output in several tasks (e.g., visual cliff, Go/No-Go and Morris water maze) can be modulated by levels of anxiety, they were determined in the Isl2-EphA3 knock-in mice using the light/dark box test (Crawley 2007). This conflict test evaluates anxiety based on the tendency of a mouse to explore a novel environment against the aversive effect of a brightly lit open field (the light box). We measured both the time spent in the light box (aversive environment) and the number of attempts to enter this box (defined as an incomplete body entrance). Animals from the three genotypes spent a similar amount of time in the aversive environment (the light box) indicating comparable levels of anxiety (Fig. 6a). In support of that, habituation times in a novel activity cage and latency to leave the start segment in the beam walking test, presented above, did not differ between the three genotypes further suggesting that the Isl2-EphA3KI animals exhibit normal levels of anxiety. Surprisingly, EphA3KI/KI and EphA3KI/+ mice made significantly fewer attempts to enter the light box (incomplete body entrances) compared to their WT littermates (attempts: F2,21 = 4.24, p < 0.05, NK post hoc: p < 0.05; Fig. 6b). In other words, EphA3KI/KI and EphA3KI/+ mice were less hesitant and entered the light box more readily suggesting that they fail to refrain from exploring an aversive environment. In addition, EphA3KI/KI and EphA3KI/+ mice showed a decreased latency for complete body entrance into the light box compared to WT littermates (latency: F2,21 = 3.24, p = 0.06; Fig. 6c). This provides further evidence that they did not hesitate to enter an aversive environment. However, EphA3KI/KI mice showed no increase in time spent in the light box and no impairment in the visual cliff test, optokinetic reflex and both versions of the water maze in which performance depends on intact visual abilities (Yassine et al. 2013). Alternatively, reduced hesitation to enter the light box could be related to a diminished response inhibition, a key feature of impulsivity (Chamberlain and Sahakian 2007).
To confirm defects in response inhibition of knock-in mice, we performed a Go/No-Go task. Go/No-Go paradigms are based on a cue discrimination conditioning and are commonly used to assess attention and response inhibition, but also learning and memory functions in humans and mice (Meziane et al. 1993; Aron and Poldrack 2005; Gubner et al. 2010; Loos et al. 2010). This test required food restriction, during which the mice were kept at 85 % of their weight to ensure motivation for food reward. Mice of all three genotypes showed similar weight loss and motivation for food during food restriction (not shown) (Meziane et al. 1993). In our version of the task, mice were conditioned to run successively down two runaways differing in colors, one color runaway being always baited with food (Go trail) and the other never baited (No-Go trial). Both EphA3KI/+ and WT littermates progressively learned to discriminate between the reinforced (Go trials) and non-reinforced (No-Go trials) runways as indicated by a significant decrease in running time on Go trials and stable running times on No-Go trials (Go trials: F2,34 = 18.9, p < 0.0001; Fig. 7a, b) as usually observed in this task (Meziane et al. 1993). This suggested normal learning, motivation and response inhibition in EphA3KI/+ and WT mice. Running duration of EphA3KI/KI animals decreased similarly than WT and EphA3KI/+ littermates on Go trials. Surprisingly, and in contrast to WT and EphA3KI/+, EphA3KI/KI running times also significantly decreased on No-Go trials (No-Go trails: F4,34 = 4.03, p < 0.01, NK p < 0.05; Fig. 7a, b) indicating their failure to refrain themselves from running in the non-reinforced runway on No-Go trials. Preserved performances of the EphA3KI/KI animals on Go trials suggested intact motivation for food and efficient learning. A discrimination learning deficit in these mice is unlikely since amnesic treatments are known to affect essentially Go running times (Meziane et al. 1993, 1998). In addition, their performance in the visible and hidden versions of the Morris water maze as well as in the visual cliff test and optokinetic reflex suggests that their visual acuity and visuo-spatial memory are comparable to those of WT and EphA3KI/+ littermates. Taken together, these results further support the hypothesis of a defective response inhibition in the EphA3KI/KI animals.
In principle, this defective behavior could be caused by impaired attention or increased distraction (Barkley 2004). To test this possibility, we repeated the reinforced Go task, but added visual (flashing light) and auditory (tone) distractors. Mice of all genotypes showed significantly increased running times by reducing their speed in trials with tones (70 dB tone: F1,18 = 5.48, p < 0.05; 90 dB tone: F1,18 = 9.18, p < 0.01; Fig. 7c) and flash lights (F1,18 = 92.06, p < 0.0001; Fig. 7c) compared to non-distracted trials. Notably, all EphA3KI/KI mice increased their running times when exposed to a flashing light, (one mouse stopped to look toward the origin of the stimulus) although the difference between EphA3KI/KI and WT littermates did not reach statistical significance (Flash latency: F2,18 = 1.17, p = 0.33; Fig. 7c). These data indicate that a flashing light and loud tones are effective distractors during the Go task.
Analysis of regional monoamine levels
The observed defective response inhibition in EphA3KI/KI mice, corresponding to an ADHD phenotypic feature, could be induced by abnormal catecholamine levels (van der Kooij and Glennon 2007; Sontag et al. 2010). To test this possibility, we determined levels of monoamine neurotransmitters in distinct areas of the mouse brain, namely the superficial layers of the superior colliculus (SC), the prefrontal cortex, the striatum, the parietal cortex and the cerebellum, all involved in attentional processes and motor control (Himelstein et al. 2000; Aron and Poldrack 2005; Biederman and Faraone 2005; Overton 2008). Levels of dopamine, adrenaline and serotonin were not significantly different between genotypes in the five structures studied (Fig. 8; Online resource 2). In contrast, the levels of noradrenaline were significantly increased in the superficial layers of the SC of EphA3KI/KI compared to their EphA3KI/+ and WT littermates (KW test p < 0.05; Figs. 8a, 9). The increase in noradrenaline in the superficial layers of the SC prompted us to examine the expression of receptors, transporters and enzymes that are involved in monoaminergic metabolism and associated with attention-deficit diseases (Himelstein et al. 2000; Biederman and Faraone 2005). All three genotypes showed similar expression of transporters, metabolic enzymes and downstream receptors of dopamine, noradrenaline, adrenaline and serotonin in the superficial layers of the SC and in other brain regions (Online resource 2).