Contrast-dependent orientation discrimination in the mouse

As an important animal model to study the relationship between behaviour and neural activity, the mouse is able to perform a variety of visual tasks, such as orientation discrimination and contrast detection. However, it is not clear how stimulus contrast influences the performance of orientation discrimination in mice. In this study, we used two task designs, two-alternative forced choice (2AFC) and go/no-go, to examine the performance of mice to discriminate two orthogonal orientations at different contrasts. We found that the performance tended to increase with contrast, and the performance at high contrast was better when the stimulus set contained a single contrast than multiple contrasts. Physiological experiments in V1 showed that neural discriminability of two orthogonal orientations increased with contrast. Furthermore, orientation discriminability of V1 neurons at high contrast was higher in the single than in the multiple contrast condition, largely due to smaller response variance in the single contrast condition. Thus, the performance of mice to discriminate orientations at high contrast is adapted to the contrast range in the stimuli, partly attributed to the contrast-range dependent capacity of V1 neurons to discriminate orientations.

For the anaesthetized experiments, the mice were sedated with an intraperioneal injection of chlorprothixene (4 mg/kg) and anaesthetized with urethane (1.2 g/kg). Dexamethasone (2 mg/kg) and atropine (0.3 mg/kg) were injected subcutaneously to reduce edema and secretions. We monitored the toe-pinch reflex of the animal and supplemented additional urethane (0.2 -0.3 g/kg) as needed. For the awake experiments, the headplate of the mouse was fixed to a holder attached to the stereotaxic apparatus and the mouse's body was restricted in a circular plastic tube. Following anaesthesia with isoflurane (~ 1%), a craniotomy (~ 1 mm diameter) was made above V1, and the exposed cortex was protected with a silicone elastomer (Kwik-Cast, World Precision Instruments). The mouse was let to recover from the anaesthesia for at least 3 hours, after which the animal was head-fixed to a holder via the headplate. The dura was removed before the recordings. The recordings lasted 3 -5 hours in anaesthetized mice and 2 -3 hours in awake mice. After the recordings, the mice were euthanized by an overdose of pentobarbital (0.5 g/kg) followed by cervical dislocation.

Behaviour
In the multiple contrast condition, each block included four trials of the same contrast. For the 2AFC task, each block included two trials in which the target was on the left side and two trials in which the target was on the right side, and the sequence of the four trials within each block as well as the contrasts in different blocks were randomized. For the go/no-go task, the four trials in each block included two targets and two non-targets of the same contrast, and the contrasts in different blocks as well as the sequence of stimuli within each block were randomized.
For the multiple contrast condition in the 2AFC task, the contrasts ranged from 15% to 100%. For the go/no-go task, the multiple contrasts ranged from 15% to 100% or from 15% to 80%, and all animals were trained on both versions of the multiple contrast condition.

Electrophysiology
The neural signals were amplified at 30 kHz and filtered with a Cerebus 64-channel system (Blackrock microsystem). We band-pass filtered the signals at 0.25 to 7.5 kHz and set a threshold at 4 s.d.
of the background noise to detect spike waveforms. Spikes were sorted offline with the Offline Sorter (Plexon Inc.) using cluster analysis of principal component amplitudes. Spike clusters were considered to be single units if their interspike interval was > 1 ms and P < 0.05 for multivariate ANOVA tests on clusters. To determine whether a single neuron was recorded by more than one site in the electrode, we computed correlation coefficients (binned at 1 ms) between all pair-wise combinations of units in the same recording. Those pairs with a correlation coefficient > 0.1 were considered to contain duplicate units, and the unit with the lower firing rate in the pair was discarded 1 .

Analysis of neuronal responses
In Fig. 5 and Fig. 7, 188 neurons from anaesthetized mice and 49 neurons from awake mice were included for the responses measured with contrasts ranging from 15% to 100%, and 148 neurons from anaesthetized mice were included for the responses measured with contrasts ranging from 15% to 80%. In contrast, respectively, and 17 neurons from awake mice were included for the responses at 100% contrast.
For each contrast, we compiled a receiver operating characteristic (ROC) curve for the pair of response distributions using a range of 30 criterion values spanning from the minimal to the maximal response 2 . The probability of correct discrimination was calculated as the area under the ROC curve (ROC area). After performing this analysis for all contrast levels, we generated a neurometric function for each neuron. When we compared the ROC area at high contrast (100% or 80%) between the single and the multiple contrast conditions, we only used those cells in which both conditions were tested in the same cell.  (a) Hit rate at 100% contrast was above 90% for both the single and the multiple contrast conditions. The single contrast condition was tested both before (left bar) and after (right bar) the multiple contrast condition. (b) FA rate at 100% contrast was significantly lower in the single than in the multiple contrast condition. (c) Response bias at 100% contrast was significantly lower in the single than in the multiple contrast condition. (d-f), Hit rate, FA rate, and response bias for 80% contrast in the single and the multiple contrast conditions, similar as those described in (a-c). The single contrast condition was tested after the multiple contrast condition. Error bars, s.d., n = 14 mice. *, P < 0.05, **, P < 0.01, ***, P < 0.001, Wilcoxon signed rank test. Figure S3. Neurometric functions exhibit a variety of shapes. (a) Neurometric functions for 6 example neurons measured using the contrast range of 15% to 100%. (b) Neurometric functions for 6 example neurons measured using the contrast range of 15% to 80%. Figure S4. Comparison of the maximum ROC area between cells whose peak ROC areas were at the highest contrast and those cells whose peak ROC areas were at lower contrasts. (a) The responses were measured with contrasts ranging from 15% to 100%. P = 0.34, Wilcoxon rank sum test, n = 135 for the monotonic cells (cells whose peak ROC areas were at the highest contrast) and n = 102 for the non-monotonic cells (cells whose peak ROC areas were not at the highest contrast). (b) The responses were measured with contrasts ranging from 15% to 80%. P = 0.007, Wilcoxon rank sum test, n = 93 for the monotonic cells and n = 55 for the non-monotonic cells. Error bars, s.e.m., **, P < 0.01.   Figure S7. Behavioural performance of the four trials with the same contrast within each block in the multiple contrast condition and that of the four trials within each block in the single contrast condition for the go/no-go task. For all conditions, each block of trials included two targets and two non-targets of the same contrast and the sequence of stimuli within each block were randomized. For the multiple contrast condition, the contrasts in different blocks were also randomized. (a) Multiple contrast condition, contrasts ranging from 15% to 100%. For 100% contrast, the discriminability increased over the four trials (Spearman's rank correlation coefficient r = 1, P = 0.04, n = 14). For the lower contrasts, the discriminability among the four trials were not significantly different (P = 0.63, 0.99, 0.3, and 0.81, respectively, n = 14, ANOVA). (b) Similar as described in (a) except that the contrasts ranged from 15% to 80%. For 80% contrast, the discriminability increased over the four trials (Spearman's rank correlation coefficient r = 1, P = 0.04, n = 14). For the lower contrasts, the discriminability among the four trials were not significantly different (P = 0.58, 0.6, 0.47, and 0.75, respectively, n = 14, ANOVA). (c) Single contrast condition. Left: contrast = 100%, tested before the multiple contrast condition. Middle: contrast = 100%, tested after the multiple contrast condition. Right: contrast = 80%, tested after the multiple contrast condition. For both single contrast conditions, the discriminability among the four trials were not significantly different (P > 0.8, ANOVA). Error bars, s.d., ***, P < 0.001, Wilcoxon signed rank test.