Adult, male C57Bl/6 mice (Charles River Laboratories, MA) were group-housed under a standard 12 h light/dark cycle in a temperature- and humidity-controlled vivarium. Training was initiated when mice were 9–10 weeks old. Mice were allowed access to water for 2 h each day and maintained approximately 85–90% of their ad libitum weight during behavioral training. Food was available ad libitum throughout the experiment. All animal protocols were approved by the Institutional Animal Care and Use Committee at Icahn School of Medicine at Mount Sinai. Data of the 5CSRTT have been analyzed previously (Norman et al., 2021b), but only the mice which underwent both 5CSRTT and 2CSRTT with optogenetics (6 out of 8 mice) were re-analyzed in this study.

Following procedures previously described in Norman et al. (2021b), mice were anesthetized with 2% isoflurane and head-fixed in a mouse stereotaxic apparatus (Narishige, East Meadow, NY). AAV2-CamKII-eNpHR3.0-eYFP (UNC Viral Vector Core, Chapel Hill, NC) was injected bilaterally into the ACA. Bilateral ACA injection sites relative to Bregma area are: AP + 0.7 mm, ML ± 0.2 mm, DV −0.7 mm; AP + 0.2 mm, ML ± 0.2 mm, DV −0.7 mm; AP −0.3 mm, ML ± 0.2 mm, DV −0.7 mm. Bilateral VIS injection sites relative to lambda are: AP + 0.0 mm, ML ± 3.0 mm, DV −0.4 mm; AP + 0.1 mm, ML ± 2.85 mm, DV −0.4 mm; AP + 0.1 mm, ML ± 3.15 mm, DV −0.4 mm. Each infusion (500 nl) was made at 150 nl/min using a microinjector set (Nanoject III) and glass pulled syringe. The glass pipettes (1.14 mm outer diameter and 0.53 mm inner diameter, 3-000-203-G/X, Drummond Scientific, PA) were pulled on a P-97 Flaming/Brown type micropipette puller (Sutter Instrument, CA). The tip of the pulled glass pipettes was approximately 50 μm. The syringe was left in place for 1 min following the injection to reduce backflow of virus. Bilateral LEDs (Amuza, San Diego, CA) of 500 μm diameter that delivered 470 nm light were implanted at the VIS. A glass cannula associated with LED was implanted at the region of interest, and was glued onto the brain and light from the LED traveled about 1 mm through glass cannula into the brain. Model info of the LED and light intensity info is also included. Behavioral testing occurred at least 3 weeks after viral injection to allow for maximal viral expression. The location of the cannula was mapped onto blank coronal slice templates taken from Paxinos and Franklin mouse brain atlas as previously described (Norman et al., 2021b). One mouse was excluded due to the off-target cannula implantation.

Following procedures previously described in Norman et al. (2021b), optogenetic experiments were conducted using Teleopto system (Amuza, San Diego, CA). Transistor-transistor logic signals were sent to a signal generator (Rigol Technologies, Beijing, China) that drove the light at specific time points during the trials. During separate test sessions, bilateral blue LED cannula with 1 mm length of optic fiber and 1.3 mm for the center to center distance of optic fibers (TeleLCD-B-1000-500-1.3, Amuza, San Diego, CA) were stimulated during either −5:−2 s of ITI, −3:0 s of ITI or during the length of the stimulus presentation. One timing of stimulus duration was used per test day in a counterbalanced manner. During behavioral 5CSRTT and 2CSRTT, continuous 470 nm light was delivered bilaterally via LED optic fiber implanted at the VIS pseudo randomly at specific time points throughout a trial during 50% of trials. In our study, we chose to use 470 nm LED to suppress eNpHR3.0-expressing ACA neurons, because previous studies found that red-shifted light frequencies resulted in off-target effects in control mice that reduced 5CSRTT performance (White et al., 2018). During optogenetics testing, stimulus duration was pseudo randomly shuffled between 2.0 and 1.0 s. The power at the fiber-optic tip measured using an optical power meter (Thorlabs) was approximately 10 mW.

Statistical analyses were performed using Prism (Graphpad, San Diego, CA) and in R v. 3.5.3. Analyses were conducted using general linear mixed models, ANOVA, or t-tests as indicated. Bar graphs are represented as the mean and error bars represent standard error of the mean (SEM).

We first sought to determine whether attentional behavior outcomes depend on the extent of task demand by manipulating a number of locations where visual stimulus is presented for mice to scan and touch to successfully obtain food reward. Specifically, we compared the performance between mice performing the 5CSRTT (Carli et al., 1983; Bari et al., 2008) and 2CSRTT with varying numbers of choice locations. Mice were first trained to perform the 5CSRTT (Figure 1A and Supplementary Figure 1A). The task required mice to sustain and divide their attention across 5 response windows in their lower visual field in anticipation of a brief white square stimulus appearing pseudorandomly at 1 of the 5 locations in automated standardized Bussey-Saksida touch-screen operant chambers (Bussey et al., 2012; Mar et al., 2013). The number of trials with correct touches to the flashed screen, compared to omitted and incorrect touches is a primary operational definition of attention. Correct response rate provides a broad and general index of attention performance and reflects what is captured by both accuracy and omission, two readouts that have been used to assess some aspects of attention in previous studies using 5CSRTT (Carli et al., 1983). It should be noted that omission probability is sensitive to both lapse in attention and motivation and every operant response depends to some degree on both. However, another readout of the 5CSRTT, the latency to collect reward following a correct response, serves as an indicator of motivation to assess if omission change reflects attention or motivation level. In addition, other 5CSRTT behavioral measures such as premature, and preservative response assess relatively independent indices of cognition (Dalley et al., 2004; Lustig et al., 2013; Koike et al., 2016; Supplementary Figure 1B). Mice had to reach baseline criteria of 4 out of 5 sessions with ≥ 50 trials, ≥ 80% accuracy, ≤ 20% omission at 2 s stimulus duration before 5CSRTT testing in which the stimulus duration length was shuffled between 2.0, 1.5, 1.0, and 0.8 s in a pseudorandomized order. Following 5CSRTT testing, mice began 2CSRTT baselining with 2 s stimulus duration in which the number of response windows was reduced to 2. To pass 2CSRTT baselining stage, mice had to reach same baseline criteria as the 5CSRTT before undergoing 2CSRTT testing. We found that task performance of mice was significantly higher during the 2CSRTT compared to the 5CSRTT as indicated by increased correct trials [Figure 1B: two-way repeated measures analysis of variance (RM ANOVA) task type P = 0.0059]. The improved task performance during the 2CSRTT was driven by improved accuracy (Figure 1C: two-way RM ANOVA, task type P = 0.0118) with a lesser degree of change in omissions (Figure 1D: two way RM ANOVA, task type P = 0.1146). Furthermore, CRL was slower during the 5CSRTT (Figure 1E: t-test P = 0.0002), while reward collection latency(s) (Figure 1F: t-test P = 0.7720), premature responses (%) (Figure 1G: t-test P = 0.3218), and perseverative responses (%) (Figure 1H: t-test P = 0.2377) were not statistically different between both the 5CSRTT and 2CSRTT. These data suggests that each task engaged similar levels of motivation, impulsivity, and compulsivity, respectively.

Of note, there was no difference in 2CSRTT performance between mice with or without prior 5CSRTT testing (Figure 2). Specifically, neither correct% (Figure 2B: two-way RM ANOVA, task type P = 0.9379), accuracy% (Figure 2C: two-way RM ANOVA, task type P = 0.9489) nor omission% (Figure 2D: two-way RM ANOVA, task type P = 0.9383) was different during 2CSRTT between mice with or without prior 5CSRTT testing. No difference in CRL (Figure 2E: t-test P = 0.3119), reward collection latency (Figure 2F: t-test P = 0.9363), premature responses (Figure 2G: t-test P = 0.2913), and perseverative responses (Figure 2H: P = 0.1111) was observed between the two groups. These results suggest that the order of training has limited influence to attentional behavior. Collectively, these data support that lower task demand of 2CSRTT compared to 5CSRTT is reflected to better task performance in 2CSRTT.

While our previous studies have shown that ACA_VIS projection neurons are required for attentional behavior during 5CSRTT in mice (Norman et al., 2021a,b), it is unknown whether the contribution of this subpopulation of frontal cortex projecting neurons are task demand dependent. To this end, we selectively suppressed ACA_VIS axon terminal activity in visual cortex using the inhibitory opsin halorhodopsin eNpHR3.0 at key time points during both the 5CSRTT and 2CSRTT in which the stimulus duration length was shuffled between 2 and 1 s in a pseudorandomized order to determine its effect on task performance (Figures 3A,B). Data of the 5CSRTT have been analyzed previously (Norman et al., 2021b), but only the mice which underwent both 5CSRTT and 2CSRTT (6 out of 8 mice) were re-analyzed in this study. Viral spread and implant location were validated and reported in a previous study that used the mice to this study (Norman et al., 2021b). This study also showed that patch-clamp recording from ACAvis neurons in slices effectively suppressed spike activities without rebound excitation (Norman et al., 2021b).

Correct response rate provides a broad and general index of attention performance and reflects what is captured by both accuracy and omission, two readouts that have been used to assess some aspects of attention in previous studies using 5CSRTT (Carli et al., 1983). We found that suppressing ACA_VIS projections during the 3 s immediately before stimulus presentation (−3:0 s) impaired performance during the 5CSRTT, reducing correct trials [Figure 3F general linear mixed model (GLMM) effect of light during 5C task P = 0.0114] by increasing omission (Figure 3G effect of light during 5C task P = 0.0001), however, no effect was observed during the 2CSRTT (Figures 3F,G GLMM, effect of light during 2C task: correct P = 0.9784, omission P = 0.8267). We did not observe effect of light illumination on accuracy (Figure 3H GLMM, effect of light P = 0.6049). These findings indicate that it is mainly the omission increase that contributed to the correct rate reduction by the optogenetic suppression of ACAvis projections during 5CSRTT.

Performance of either task was not affected by adjusting the LED illumination time to earlier in the ITI (−5: −2 s) (Figures 3C–E: GLMM, effect of light: correct trials% P = 0.1010, omissions% P = 0.2873, accuracy% P = 0.1796) or during stimulus presentation (0:1 s) (Figures 3I–K: GLMM, correct trials%: effect of light P = 0.4721, omission: effect of light during 5C task P = 0.6911, during 2C task P = 0.9702, accuracy%: effect of light P = 0.2600). Of note, our recent study showed that LED illumination over VIS to control mice expressing eYFP in ACA induced no behavioral effects during 5CSRTT (Norman et al., 2021b), suggesting that behavioral changes depend on halorhodopsin expression and are unlikely to be due to other factors such as light induced heat change. It should be though noted that eYFP control experiments were not performed during 2CSRTT. ACA_VIS terminal inhibition during any time point examined had no effect on reward collection latency (Supplementary Figure 2A: t-test, ITI −5:−2 s P = 0.3846, ITI −3:0 s P = 0.7773, stimulus period P = 0.9072), CRL (Supplementary Figure 2B: t-test, ITI −5:−2 s P = 0.7012, ITI −3:0 s P = 0.2568, during stimulus period P = 0.7620), premature responses (Supplementary Figure 2C: t-test, ITI −5:−2 s P = 0.1002), or perseverative responses (Supplementary Figure 2D: t-test, ITI −5:−2 s P = 0.5593, ITI −3:0 s P = 0.5554, stimulus period P = 0.2756) during the 2CSRTT.

Taken together, these data suggests that ACA_VIS projections are causally important for attentional task performance in the period just before stimulus presentation during 5CSRTT, but become dispensable once the task switch to 2CSRTT, suggesting that ACA_VIS projections are necessary only when task demand is high.