Locus coeruleus norepinephrine contributes to visual spatial attention by selectively enhancing perceptual sensitivity

Summary

Selectively focusing on a behaviorally relevant stimulus while ignoring irrelevant stimuli improves perception. Enhanced neuronal response gain is thought to support attention-related improvements in detection and discrimination. However, understanding of the neuronal pathways regulating perceptual sensitivity remains limited. Here, we report that responses of norepinephrine (NE) neurons in the locus coeruleus (LC) of non-human primates to behaviorally relevant sensory stimuli promote visual discrimination in a spatially selective way. LC-NE neurons spike in response to a visual stimulus appearing in the contralateral hemifield only when that stimulus is attended. This spiking is associated with enhanced behavioral sensitivity, is independent of motor control, and is absent on error trials. Furthermore, optogenetically activating LC-NE neurons selectively improves monkeys’ contralateral stimulus detection without affecting motor criteria, supporting NE’s causal role in granular cognitive control of selective attention at a cellular level, beyond its known diffuse and non-selective functions.

Graphical Abstract

eTOC blurb

Ghosh and Maunsell show that LC norepinephrine neurons are selectively modulated by visual spatial attention towards contralateral stimulus. By activating LC optogenetically during sensory selection, they demonstrate that it drives perceptual sensitivity, improving attentional performance. This LC contribution is independent of motor processing and general arousal.

INTRODUCTION

Experiments that examined selective attention-related changes in the visual system demonstrate an amplification or gain of visual neuron’s spike response associated with improvements in perceptual detection and discrimination at the attention location or feature.^1–4 Importantly, enhanced perceptual sensitivity to the attended stimulus is one of the crucial cognitive factors that accounts for the response gain in the visual cortex and the improved behavior associated with selective attention.⁵ However, we have limited understanding about the cellular substrate that mediates a causal role on selective sensory processing and attentional performance in primates, owing to previous limitations on temporally-precise control of neuronal activity of regulatory circuitry in non-human primates doing tasks that allow control of attention and measurement of sensory performance free of confounds from decision and motor actions.

Neuromodulators have long been thought to play crucial roles in cognitive functions, including selective attention.⁶ Neuromodulation by norepinephrine (NE) has a powerful effect on general arousal and wakefulness.⁷ The locus coeruleus (LC) is a primary source of NE in the brain and is known to contribute to diffuse, non-specific arousal.^8–11 However, a correlation between LC activation and behavioral accuracy in non-human primates,¹² coupled with NE-induced enhancement of the signal-to-noise of cerebral cortex sensory representations,^8,13,14 suggest that LC-NE might improve perceptual performance by selectively facilitating relevant sensory inputs, thereby contributing to selective attention. It remains unknown whether LC-NE neuromodulation causally contribute to perceptual sensitivity in a selective way that is distinct from arousal, and attentional performance in primates.

We therefore trained monkeys to perform a novel visual spatial attention task that allowed us to measure spatially selective perceptual sensitivity independently of decision and motor selection (or action). We measured neuronal responses from LC-NE neurons, and optogenetically enhanced their responses to visual stimuli to measure their causal contributions to spatially selective perceptual detection performance. We found that LC neurons display robust spatially selective visual attention-related modulation that is specific to perceptual sensitivity. By optogenetically activating these LC-NE neurons, we identified that the behaviorally relevant sensory stimulus-locked neuronal activity of LC-NE neurons drives perceptual sensitivity to visual stimuli in a spatially selective way.

RESULTS

Selective spike modulation of LC neurons associated with visual spatial attention

We trained two rhesus monkeys to selectively attend one of two visual stimulus locations while performing a visual orientation change detection task (STAR Methods, Figure 1A).¹⁵ Once the animal fixated, two Gabor sample stimuli appeared for 200 ms, one in each hemifield. After a brief delay (200–300 ms), a Gabor test stimulus appeared in one of the locations. The monkey had to report whether the test stimulus had a different orientation than the sample that had appeared in that location by making a saccade to one of two colored saccade targets after the fixation spot went off. The locations of the saccade targets were interchanged randomly across trials. In this way, signals related to motor planning¹⁶ were dissociated from the attention-related perceptual sensitivity we wanted to examine. Spatially selective attention was controlled in blocks of trials by different reward sizes and by varying the test stimulus probability for the two locations (STAR Methods). Manipulating the expectation of a behaviorally relevant stimulus over a block of trials allowed the monkeys to develop a stable state of attention (Figure S4A).

Figure 1. Modulation of LC neurons with changes in visual spatial attention.

(A) Visual orientation change detection task. Monkeys maintained fixation (400–800 ms) while attending to one of two sample Gabor presented (200 ms) in opposite hemifields. After a short delay (200–300 ms), a test stimulus (200 ms) appeared at one of the stimulus locations. Monkeys reported whether the orientation changed between the sample and test by making a saccade to the appropriate saccade target (which had different colors and shapes). In the non-match trial, the test orientation changed from the sample; in the match trial, the test orientation remained unchanged. Positions of the match and non-match saccade targets were randomized on each trial. Monkeys’ spatial attention was controlled between the stimulus locations in alternate blocks of trials by varying both the reward size for correct detections and by using different test stimulus probabilities at the two locations. (B-C) Monkeys’ performance averaged across sessions as a function of orientation changes (monkey S, 29 sessions; monkey P, 22). Behavioral sensitivity (d’) is much better on trials when the test stimulus appeared at the attended location (valid trials; filled circles) compared to the unattended location (invalid trials; open circles). Error bars, 95% confidence intervals. (D-E) Example and population PSTHs of spike rates of phasic LC neurons for all correct trials when monkeys attended either contralateral or ipsilateral to the recorded LC (phasic, monkey S, n = 89, monkey P, n = 74; non-phasic, monkey S, n = 88, monkey P, n = 53). Error bars, ± 1 SEM. (F-G) Same as in (D-E) but for non-phasic LC neurons.

Monkeys’ detection performance improved at the attended location relative to the unattended location as measured by increased behavioral sensitivity (d’) (Figures 1B–C). While the monkeys did the attention task, we recorded simultaneously from populations of LC neurons (monkey S, n = 229; monkey P, n = 185). NE neurons in the LC were identified using stereotaxic coordinates and characteristic neurophysiological responses to salient sound stimuli.^17,18 Recorded neurons were classified as phasic responsive to visual stimulus (phasic, monkey S, n = 89/229; monkey P, n = 74/185), pre-stimulus fixation responsive (non-phasic, monkey S, n = 88/229; monkey P, n = 53/185) or saccade related (saccade, monkey S, n = 104/229; monkey P, n = 124/185) (Figure S1A). Responses of phasic LC neurons to the sample stimuli were stronger when attention was directed to the stimulus contralateral to the recorded LC (Figures 1D–E; monkey S, neuronal modulation (d’_neuron) over sample period = 0.84 ± 0.09, p < 10⁻¹⁵; monkey P, d’_neuron = 0.65 ± 0.07, p < 10⁻⁹; signed-rank test). This modulation of LC activity by attentional shifts was associated with increased behavioral d’, and was independent of changes in perceptual decision criteria and motor criteria (Figure S2).

Spike rates of non-phasic LC neurons also increased when the monkeys shifted their attention to the location contralateral to the recorded LC (Figures 1F–G; monkey S, d’_neuron over the fixation period = 0.78 ± 0.05, p < 10⁻¹⁵; monkey P, d’_neuron = 0.67 ± 0.06, p < 10⁻¹³; signed-rank test). Responses of saccade-related neurons were unchanged throughout the trial, except for a delayed response following the saccade, and were not affected by the animals’ location of attention (Figure S1B).

Previous studies have found a shift in central fixation eye position towards the cued attended location when monkeys’ spatial attention was controlled over long blocks of trials.^19,20 We also observed that the monkeys’ averaged fixation eye positions marginally (but significantly) shifted towards the attention location (mean ± 1 SEM shift in fixation eye position across sessions, monkey S, 0.15° ± 0.03°, p < 10⁻³; monkey P, 0.025° ± 0.006°, p < 10⁻²; signed-rank test; STAR Methods). Therefore, we investigated whether these shifts in fixation eye position had any confounding effects on the observed neuronal modulation of LC neurons. We selected subsets of trials from the two attention conditions to match the distribution of fixation eye positions (STAR Methods). The attentional modulation of LC neurons did not differ significantly for the fixation-matched trials compared to all trials (phasic LC neurons, monkey S, p = 0.75, monkey P, p = 0.49; non-phasic LC neurons, monkey S, p = 0.53, monkey P, p = 0.80, signed-rank test; Figures S1D–E). Thus, attention related modulations in LC neurons were not associated with small offsets in gaze between attention states.

The latency to visual response (half-max spiking response) of phasic responsive LC neurons had a broad distribution (median, 39 ms, 1^st and 3^rd quartiles, 18 ms and 56 ms; Figure S1C; STAR Methods). On average, spike rates of phasic LC neurons gradually increased before the onset of visual stimuli, with a strong response occurring at the onset of visual stimuli when spatial attention was directed towards the contralateral location (Figure 1E). These short latency or early responses of phasic LC neurons might be associated with priming or expectation of the contralateral attended stimulus, but are not related to shifts in fixation eye position (Figure S1D).

Furthermore, the spike rate modulation of the phasic LC neurons was not temporally locked to fixation initiation (p > 0.6, signed-rank test; aligned with reference to fixation onset). The onset of fixation did not elicit a reliable spike response in these LC neurons (time to half max relative to the fixation, median, 445 ms, 1^st and 3^rd quartiles, 432 ms and 458 ms; latency of 32/163 LC phasic neurons were reliably measured; STAR Methods). In addition, the spontaneous spike rate of LC neurons during the intertrial intervals (during 500 ms before the trial start) did not differ between the contra- and ipsi- lateral attended trial blocks (p > 0.4, signed-rank test).

Together, these results suggest that the modulation of LC neurons is associated with a spatially selective (hemifield) regulation of the processing of behaviorally relevant sensory stimuli, rather than a state of general covert orienting or arousal.

Spike modulation of LC neurons depends on contralateral attention and correct perceptual detection

Previous studies have suggested that the spiking of LC neurons closely varies with behavioral performance during visual target detection tasks.¹² Thus, we decomposed population firing rates into demixed principal components (dPC)²¹ to quantify how the activity of LC neurons relates to perceptual detection (correct versus error) and the focus of selective attention (contra versus ipsi hemifields relative to the recorded LC) in single trials (STAR Methods; Figures 2A and 2C; Figure S3). Spike trains of LC neurons carried significant information about the location of selective attention (left, Figures 2B and 2D) and detection of stimulus orientation change (middle, Figures 2B and 2D) as measured by decoding accuracy in cross-validated (leave-one-out) single trials (STAR Methods).

Figure 2. Single trial prediction of selective attention and sensory detection by LC population code.

(A) Population PSTHs of spike rates of phasic LC neurons for correct and error trials while attention was directed either contra- or ipsi-lateral to the recorded LC (n = 163, two monkeys). Error bars, ± 1 SEM. (B) Time course of single-trial decoding accuracy (cross-validation leave-one out trials) for different cognitive variables: selective attention, correct detection and their interaction based on demixed principal components of population PSTHs of phasic LC neurons as in (A) (STAR Methods). Error bars, 95% confidence intervals from shuffled trials (C-D) Same as in (A-B), but for non-phasic LC neurons (n = 141, two monkeys).

Furthermore, some dPCs represented significant information about the interaction between the spatial location of attention and correct detection (Figure S3; right, Figures 2B and 2D). Together, these components of LC neuronal activity isolated trial-by-trial relative representations of cognitive and behavioral factors associated with enhanced perceptual sensitivity at the attended location that directly relate to correctly detecting a stimulus change (Figure S4). Neurons in other neuromodulatory systems, including dopamine and acetylcholine, are also known to exhibit mixed selectivity to multiple task-relevant factors.⁶

Optogenetic activation of LC-NE neurons enhances spatially selective perceptual detection

To directly examine the causal role of stimulus-locked attentional modulation of LC-NE neurons’ spiking, we selectively expressed excitatory channelrhodopsin (ChR2) in NE neurons in the LC (Figure 3). The same monkeys were unilaterally injected with two adeno-associated viruses (STAR Methods) in an approach that has been found to give robust cell type-specific opsin expression in monkeys.²² The first virus, AAV9-DBH-Cre-P2A-mCherry-SV40pA, delivered Cre-recombinase under the control of the dopamine b-hydroxylase (DBH) promoter. The second virus, AAV5-EF1a-DIO-hChR2(H134R)-EYFP, delivered a Cre-recombinase-dependent hChR2 construct (STAR Methods). Subsequently, NE neurons in the LC were identified by characteristic neurophysiological responses to salient sound stimuli^17,18 and excitation by optical stimulation using brief blue light (Figures 3A–B). Reliable co-expression of ChR2 and DBH was identified in putative LC-NE neurons (Figure 3C).

Figure 3. Optogenetic regulation of neuronal spiking of LC-NE neurons in monkeys.

(A) Example LC-NE neuron. Left, Optogenetic excitation by 100 ms blue laser pulses (top, single trial spike trains; bottom, spike rate PSTH). Right, response to brief white-noise sound. (B) Same as in (A), for the population of same LC-NE neurons in Figure 1. Optogenetic stimulation was tested for 10 and 20 mW laser intensities. Error bars, ± 1 SEM. (C) Immuno-histological confirmation of ChR2 expression in LC-NE neurons. A representative immuno-stained coronal brain section of monkey S showing colocalization of NE neurons and ChR2 in the LC (STAR Methods). Panels from left to right, DAPI nuclear staining, dopamine-beta-hydroxylase (DBH) immunostaining of LC-NE neurons, enhanced yellow fluorescent protein (eYFP) in DBH:Cre:hChR2 (AAV5-EF1a-DIO-hChRh2(H134R)-EYFP) neurons in the LC, and merged image of DAPI, DBH and eYFP. Yellow cells represent colocalization of eYFP and DBH. Bottom panels represent high magnification images of the box area in the top panel.

On a random half of the trials, we optogenetically activated LC-NE neurons while the monkeys performed the attention task (Figure 4A). These stimulation sessions were a subset of the total experimental sessions used for the neurophysiological recordings in Figures 1 and 2. A 400 ms train of eight 10 ms optical pulses (20 Hz) was delivered to the LC starting 200 ms before the sample stimuli. Previous studies suggest that the neuronal activity of the LC-NE system is closely linked with the pupillary area,²³ which provides an index of arousal and cognitive engagement. Unilateral optogenetic activation of LC-NE neurons increased the pupil area relative to unstimulated trials, documenting the efficacy of LC-NE stimulation in altering the physiological state (Figures 4B–C, monkey S, p < 0.01; monkey P, p < 0.01; paired t-test).

Figure 4. Optogenetic excitation of LC-NE neurons selectively improves performance contralateral to the stimulated LC.

(A) Optogenetic stimulation during the sample stimulus in spatially selective visual attention task (Figure 1A). LC was stimulated unilaterally on a randomly interleaved 50% of validly cued (instruction trials) trials by laser pulses for 400 ms starting 200 ms before sample stimulus onset (10 ms pulse width, 40 mW). (B) Pupil area in an example session aligned to sample onset. Optogenetic stimulation increased the pupil area relative to unstimulated trials. Error bars, ± 1 SEM. (C) Session averaged mean pupil area for unstimulated and LC-stimulated trials for monkeys S and P. Error bars, ± 1 SEM. (D-E) Session averaged behavioral sensitivities (d’) of monkey S (F; n = 11 sessions) and monkey P (G; n = 11 sessions) as a function of orientation changes. Trials are grouped according to whether monkeys’ spatially selective attention was directed at the location either contra- (left) or ipsi-lateral (right) to the recorded LC. Trials are also separated according to optogenetically stimulated (blue) and unstimulated (yellow) trials. When the animal’s attention was directed to the stimulus contralateral to the stimulated LC, d’s were improved relative to unstimulated trials. Error bars, 95% confidence intervals. (F-G) Session averaged behavioral decision criteria for all valid trials, grouped according to the attended location and stimulation for the same dataset shown in Figures 4D–E. Error bars, 95% confidence intervals.

When the animal’s attention was directed to the stimulus in the visual hemifield contralateral to the stimulated LC, only a specific component of visual attention, perceptual sensitivity as measured by behavioral d’ in that hemifield, was greatly enhanced (associated with fewer false alarms and more hit rates, blue curve, left, Figures 4D–E). In contrast, LC activation caused a weak impairment of the animal’s discrimination of ipsilateral orientation changes (blue curve, right, Figures 4D–E). In contrast, LC activation did not affect the animals’ perceptual decision criteria – another component of visual attention performance (Figures 4F–G). Both the neuronal encoding of selective attention and the effects of LC-NE activation to perceptual performance persisted over trials within the experimental trial blocks (Figure S4). Thus, LC activity improves behavioral performance in a specific (perceptual sensitivity) and spatially selective way when animals perform a visual attention task. This spatially selective (hemisphere-specific) effect of LC stimulation was not driven by enhanced effective retinal illumination due to a larger pupil area (Figures 4B–C). LC stimulation uniformly affected the pupil area irrespective of the location of monkeys’ spatial attention (monkey S, p = 0.94; monkey P, p = 0.86; paired t-test).

A previous report described spike rate modulations of monkey visual cortical neurons that were coupled with the onsets of microsaccades towards the attended location.²⁴ We examined whether microsaccades have a relationship to the selective contribution of LC neurons to spatial attention. Microsaccades were detected based on instantaneous eye velocities (Figure S5; STAR Methods).²⁵ Shortly after the visual stimulus appeared the rate of microsaccade dropped transiently. However, the rate of microsaccade was unaffected by optogenetic stimulation of LC (monkey S, F_{(1, 10)} = 0.03, p = 0.86; monkey P, F_{(1, 10)} = 1.5, p = 0.25; repeated measures ANOVA, Figure S5). Moreover, we found no significant bias in the proportion of microsaccades towards the attended sample stimulus (within <15° solid angle) immediately after the onset of stimulation (monkey S, F_{(1, 10)} = 0.23, p = 0.64; monkey P, F_{(1, 10)} = 1.01, p = 0.34; repeated measures ANOVA; Figure S5). Opto stimulation of LC-NE did not affect the microsaccades directed towards the stimuli (monkey S, F_{(1, 10)} = 0.08, p = 0.78; monkey P, F_{(1, 10)} = 0.19, p = 0.67; repeated measures ANOVA; Figure S5).

We also tested whether optogenetic stimulation of LC-NE neurons affected the monkeys’ gaze angle, which could indirectly enhance contralateral performance. Spatially selective attention led to a shift in the monkeys’ averaged fixation eye positions towards the attention location (monkey S, F_{(1, 10)} = 16.46, p < 0.01; monkey P, F_{(1, 10)} = 15.57, p < 0.01; repeated measures two factor ANOVA; STAR Methods). However, optogenetic stimulation had no effect on fixation eye positions (monkey S, F_{(1, 10)} = 2.8, p = 0.12; monkey P, F_{(1, 10)} = 2.39, p = 0.15; repeated measures two factor ANOVA). These results suggest that the specific contribution of LC-NE activation to spatially selective visual attention is not directly coupled with the rates or directions of microsaccades, and the shift in central fixation eye position.

Spatially selective contribution of LC-NE to visual detection is independent of attentional selectivity and the level of general attentional effort

Previous studies have reported bilateral or non-specific behavioral effects of unilateral optogenetic LC activation or inhibition in rodents.^10,26 Unlike those studies, the monkeys in our task directed their visual attention to one of the two locations in opposite hemifields. It is possible that LC-NE activation enhances performance at an attended location but cannot direct spatially selective attention. Thus, we tested whether the specific effect of LC activation on performance in our task depended on the selectivity of spatial attention. In separate experimental sessions, we instructed monkeys to distribute their visual attention equally between two stimulus locations in opposite hemifields using the same visual detection task as described in the previous section (Figure 1A). In interleaved blocks of trials, we controlled the animals’ spatially non-selective attentional effort (or attentional intensity) between low and high values (at both locations) by varying the task difficulty while keeping the reward size and test stimulus probability the same between the two locations¹⁵ (STAR Methods; Figure 5A). On a random half of the valid trials, where the test stimulus appeared at the attended location, we unilaterally optogenetically stimulated LC-NE neurons. Artificially boosting hemifield-specific LC spiking by optogenetic stimulation improved monkeys’ contralateral behavioral d’ relative to the ipsilateral detection, similar to the way spatially selective attention improved performance at the cued (attended) location in Figures 4D–E (Figure 5B; Table S1). The monkey’s decision criteria did not change with LC activation during the attentional effort manipulation (Figure 5C; Table S1). Together, these results suggest that the selective contribution of enhanced LC spiking (relative to the opposite hemifield) to perceptual performance is not limited to enhancing a previously elevated perceptual sensitivity (i.e., high d’, Figures 4D–E), but it can selectively drive spatial attention (Figure 5B).

Figure 5. Spatially selective effects of LC-NE activation on behavioral performance are independent of attentional selectivity.

(A) Control of non-selective attentional effort between low and high values in alternate blocks of trials by varying task difficulty (STAR Methods). Left, The same visual orientation change detection task was used as described in Figure 1A, except there were two possible test orientation changes (ΔΘ), a contextual (67–60% probability) and a probe (33–40% probability) ΔΘ. The reward size and test stimulus probability between the two stimulus locations in opposite hemifields were the same and unchanged within a session. Right, Contextual ΔΘ was easy (90°) for low effort and difficult (16°−18°) for high effort blocks. The probe ΔΘ remained constant for both effort blocks. On randomly interleaved 50% of the probe trials, LC was optogenetically stimulated during the samples (400 ms, 20 Hz, 10 ms pulse width, 40 mW, starting 200 ms before sample stimulus onset). (B-C) Session averaged behavioral d’s (circles, (B)) and criteria (triangles, (C)) for contra- and ipsi- lateral probe test stimuli when animals (left, monkey S, n = 9; right, monkey P, n = 9) shifted their non-selective attentional effort between low (yellow open circles) and high (yellow filled circles) values. LC stimulation selectively improved behavioral d’ (blue circles) to the contralateral stimulus when the animal’s attention was similar in the two hemifields (yellow circles). Criteria (blue triangles) were unaffected by LC stimulation. Error bars, 95% confidence intervals.

DISCUSSION

We have shown that NE neuromodulation selectively contributes to the perceptual sensitivity of behaviorally relevant stimuli, a crucial neuronal signature of selective attention. By precisely controlling perceptual sensitivity independently of perceptual decision and motor criteria, we isolated a distinct contribution of LC-NE neurons to attention selectively at a contralateral location. Although primarily a non-sensory structure, the modulation of LC spike rates observed in relation to visual attention closely correlates with the increased neuronal gain seen in various cortical and subcortical visual areas associated with increased perceptual sensitivity in the neuron’s response field.^5,15,27 To the best of our knowledge, the consequences of sensory-evoked phasic NE modulation on task-specific sensory processing and behavior have not been previously demonstrated.

These results significantly advance the current understanding of LC activity on arousal, selective attention and task-specific contribution in several ways. First, previously reported increases in LC phasic activity to a reward-predicting visual response target and associated behavioral improvements in non-human primates^12,16,17 could have been related to either improved sensory processing of the target (perceptual sensitivity) or enhanced motor processing. A covert orienting toward the reward-predicting visual target (orienting toward the relevant feature or object) does not link LC phasic activity to any selective sensory processing. Here, we demonstrated that the phasic responses of LC-NE neurons are associated with perceptual sensitivity (a specific component of attention), spatially selective to the contralateral location, and not related to decision or motor action. Second, earlier studies targeting the NE system using pharmacology,^28,29 electrical stimulation³⁰ or optogenetics lacked precise temporal specificity, cell-type specificity and selective isolation of sensory versus response associations^10,31,32 during perceptual tasks. Our results show that temporally restricted phasic activation of NE neurons during sensory selection is sufficient to selectively improve perceptual sensitivity of performance to contralateral visual stimulus – a causal contribution of LC-NE neuronal activity to spatially selective attention (hemisphere selective). By manipulating attentional effort directly, we rule out the alternate interpretation that LC activity boosts behavioral sensitivity in a spatially non-selective manner. Third, similar to spatial expectations (‘where’), prior knowledge of the temporal occurrence or temporal expectations (‘when’) enhances the modulation of visual cortical neurons that correlate with the temporal probability of a relevant stimulus.^33–35 Our findings show that non-phasic LC neurons exhibit ramping spiking following fixation, while phasic LC neurons have short response latencies to attended contralateral stimulus. The neuronal encoding of spatial attention and perceptual detection, along with the causal effects of LC activation on performance, improve with each successive trial as the spatial expectation of the relevant stimulus increases once attention is shifted to the contralateral stimulus location. Together, these suggest that LC-NE modulation contributes to both the temporal and spatial allocation of attentional resources, preparing afferent cortical areas for the appropriate state of attention.

Collectively, the results identify a distinct causal contribution of NE activity in mediating task-relevant selective sensory processing distinct from attention orienting and nonselective arousal.

Sensory-specific LC contribution to attention in relation to non-selective attention

The LC-NE mediated contribution to attentional performance in our task may seem closely related to other forms of non-selective attention, such as ‘arousal’ and ‘effort’, as reported in previous studies.¹⁰ When considering an operational definition of arousal and effort based on objective measures of perceptual performance during attention demanding tasks,³⁶ the LC’s contribution to attention differs from arousal and effort on several accounts. We refer to arousal as a low dimensional cognitive state of heightened responsivity that influences sensory, cognitive, and motor signals across modalities in a non-selective and non-specific manner. In our study, we define attentional ‘effort’ as the non-selective intensity of attention, which can be measured by overall behavioral d’s at two locations.^15,36 Selective improvement of contralateral perceptual sensitivity (i.e., sensory gain) without any changes in decision or motor criteria (a cognitive or motor signal) by brief activation of LC suggests greater top-down attention to a contralateral sensory signal rather than cognitive signals. Moreover, spike rates of LC neurons during the non-task period (intertrial intervals) remained the same regardless of whether the monkey’s attention was directed towards the contra- or ipsi-lateral visual hemifield. This is a contrary to what would have been predicted by hemisphere-specific arousal. Together, these findings suggest that LC neurons were not sensitive to overall or hemisphere-specific arousal and effort in our task. The LC-mediated hemisphere-specific effect on attentional performance has greater specificity compared to arousal.

In agreement with other studies,^16–18 our results suggest that LC neurons are responsive to different task-relevant stimuli and behavioral responses (Figure S1B). Increased phasic LC spiking associated with a rewarded response target¹⁷ or effortful motor action¹⁶ has been shown to be associated with attentional orientation toward a relevant response. The orienting of attention-related activity in a subset of the LC neurons was independent of changes in sensory sensitivity (d’).

Spatial selectivity of LC contribution to attention

One important question is the mechanisms that support NE-mediated spatially selective sensory selection in visual areas targeted by the LC. Previous psychophysical studies in humans and non-human primates suggest that the spatial resolution of selective attention is appreciably poorer than the size of the visual receptive fields.^37–39 Nevertheless, it is clear that spatial attention can be differentially directed to sites separated by a few degrees within one hemifield,² and shifting attention between two stimuli within a receptive field.^40,41 It will be important in future studies to explore the spatial resolution of LC-NE contributions to visual attention.

Our result on unilateral attentional effects of LC optogenetic stimulation in monkeys is surprising, given the widespread LC projections throughout the cerebral cortex, and moderate but non-specific (bilateral) effects of unilateral LC stimulation on the contralateral LC and prefrontal cortex in rodents.⁴² However, LC-NE neuronal spike modulation with selective attention and the unilateral effect of optogenetic stimulation are consistent with the LC’s dominant ipsilateral connectivity with the forebrain and visual cortex,^43,44 and neurophysiological evidence that phasic LC discharge rapidly increases the signal-to-noise of sensory representations and mimics salient sensory stimulus-mediated responses in the cerebral cortex.^8,14,45

Spatially selective attention is known to strongly modulate neuronal responses in many brain areas in the visual system⁴⁶ that receive dense LC-NE projections, including the dorsolateral prefrontal cortex, frontal eye field, area V4, and superior colliculus. Our results, combined with empirical evidence on stimulus-specific selective effects of localized NE in the visual cortex,¹⁴ support the previously proposed “glutamate amplifies noradrenergic effects” (GANE) model⁴⁷ of spatially selective attentional spike modulation in the visual system. In this model, high glutamate release in response to the attended stimulus creates a localized NE hotspot via positive feedback between glutamate and NE release. Such an increased NE concentration could further amplify the salient stimulus-evoked response. In contrast, NE depletion would result in a suppressive effect for representations of unattended stimuli. The sensory-specific distributions of LC-NE projections in multisensory areas remain poorly characterized. Thus, anatomical mapping of sensory-specific selective LC activation would be valuable for better understanding of distinct neuromodulatory controls on perception and cognition.

Attentional modulation of LC neurons in reference to visual structures

Notably, attention-related modulations in LC approach an all-or-none effect, far greater than typical attention-related modulations seen in visual cerebral cortex.^5,15 A direct estimate of the amount of NE release in primate visual cortex associated with an attended sensory stimulus would be valuable to better understand the complexities of local computations.^48,49 Synchronous activation of a population of LC-NE neurons in attention-demanding tasks might collectively elicit many spikes.¹² The precise levels of NE changes in different cortical areas owing to changes in a few spikes per LC neuron (8.1 ± 0.4 spike/s, mean and ± 1 SEM) in perceptual tasks such as ours need to be investigated.

Individuals with attention disorders suffer from a diverse set of behavioral deficits, including difficulties in maintaining or flexibly switching selective attention, and exhibit abnormal regulation of norepinephrine (NE) neuromodulation in the nervous system.⁵⁰ Brain-wide diffuse projections of NE neurons are known to play a role in non-specific arousal. However, the extent of precise, stimulus-specific sensory contributions of NE neuromodulation to attentional performance remains unknown. This study provides important experimental evidence for a distinct NE neuromodulatory role in primates that regulates attention-demanding perceptual performance in a spatially selective way. The results suggest how neuromodulators, including catecholamine and acetylcholine, might intersect with well-defined neuronal signatures of attention that are central to human perception and behavior. The findings have potential clinical implications for designing targeted therapeutic interventions according to the specific deficits in sensory, decision or motor performance observed in disorders of attention, including attention-deficit/hyperactivity disorder and various neurodegenerative diseases.

STAR* METHODS

RESOURCE AVAILABILITY

Lead contact

All requests for additional information and reagents should be directed to and will be fulfilled by the lead contact, Supriya Ghosh (sghosh5@uchicago.edu).

Materials availability

Plasmids used to generate viruses in this study are available from Addgene. This study did not generate new unique reagents.

Data and code availability

All data are available in the main text or the supplementary information (external Data S1-S2). Behavioral task was controlled using custom-written software (https://github.com/MaunsellLab/Lablib-Public-05-July-2016).

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

Monkeys

Two adult male rhesus monkeys (Macaca mulatta, 9 and 13 kg) were each implanted with a titanium head post using aseptic surgical techniques before behavioral training began. After the completion of training (3 to 5 months), we surgically implanted a stainless-steel recording chamber targeting the LC on one side (right, monkey S; left, monkey P) to access the LC, guided by an MRI obtained before the initial surgery. The cylinders were centered on the skull at 3.5 mm A, 10.0 mm L and tilted in the coronal plane to advance toward the midline (monkey P, 9°; monkey S, 11°). The same two monkeys were used in previous studies that described different findings on neuronal responses in area V4.^15,36 All experimental procedures were approved by the Institutional Animal Care and Use Committee of the University of Chicago and followed the U.S. National Institutes of Health guidelines.

METHOD DETAILS

Behavioral task

Monkeys sat in a primate chair facing a calibrated CRT display (1024 × 768 pixels, 100 Hz frame rate) at 57 cm viewing distance inside a darkened room. Binocular eye position and pupil area were recorded at 500 Hz using an infrared camera (EyeLink 1000, SR Research). Trials started once the animal fixated within 1.5° of a central white spot (0.1° square) presented on a mid-level gray background (Figure 1A). The animal had to maintain fixation until its saccade response at the end of the trial. After a randomly selected fixation period of 400 to 800 ms, two achromatic Gabor sample stimuli appeared for 200 ms, one in each visual hemifield. After a random variable delay of 200 to 300 ms, a Gabor test stimulus appeared for 200 ms at one of the two stimulus locations, selected randomly with equal probability. The spatial phases of the sample and test Gabor stimuli were kept fixed (0°, odd symmetric) throughout all the sessions. Shortly after the test stimulus turned off (100 to 150 ms), two saccade targets of different colors and shapes (non-match target, green square; match target, magenta circle; 0.30°−0.55°) appeared in opposite directions along an imaginary line orthogonal to the axis of the samples. A go-signal (fixation spot turning off) occurred 150 to 200 ms after the saccade target appeared and indicated that the animal should make a saccade to the appropriate saccade target depending on the change in the test Gabor orientation relative to the sample. The orientation of the test stimulus changed from the sample stimulus on random half of trials (non-match trial). In the remaining half of the trials, the orientation of the test stimulus remained unchanged (match trial). The locations of the saccade targets were switched randomly across trials. This eliminated motor criteria from the attention-related behavioral d’ and perceptual decision. The orientation of the sample Gabor stimulus at each location was randomized across trials from 0° to 175° (5° interval). Orientations of left and right sample stimuli were independent and never identical. Gabor stimuli were varied each day and remained unchanged throughout each session (monkey P, left Gabor, azimuth, −7.6° to −6.5° [mean = −7.0°, interquartile = −7.6°, −6.5°], elevation, 0°; sigma, 0.65° to 0.80° [mean = 0.72°, interquartile = 0.66°, 0.80°], spatial frequency, 0.45 to 0.61 [mean = 0.53, interquartile = 0.47, 0.56] cycles per degree, right Gabor, azimuth, 6.7° to 7.6° [mean = 7.1°, interquartile = 6.7°, 7.6°], elevation, 0°, sigma, 0.62° to 0.78° [mean = 0.68°, interquartile = 0.64°, 0.70°], spatial frequency, 0.50 to 0.67 [mean = 0.63, interquartile = 0.62, 0.67] cycles per degree; monkey S, left Gabor, azimuth, −6.6° to −6.0° [mean = −6.4°, interquartile = −6.5°, −6.4°], elevation, 0°, sigma, 0.80° to 1.0° [mean = 0.92°, interquartile = 0.90°, 1.0°], spatial frequency, 0.30 to 0.50 [mean = 0.40, interquartile = 0.34, 0.45] cycles per degree, right Gabor, azimuth, 6.0° to 7.0° [mean = 6.1°, interquartile = 6.5°, 6.7°], elevation, 0°, sigma, 0.75° to 1.0° [mean = 0.86°, interquartile = 0.85°, 0.90°], spatial frequency, 0.36 to 0.55 [mean = 0.45, interquartile = 0.40, 0.50] cycles per degree). Behavioral task was controlled using custom-written software (https://github.com/MaunsellLab/Lablib-Public-05-July-2016).

Behavioral control of spatially selective attention

Spatially selective attention (Figures 1 and 4) between the two locations in opposite hemifields was alternated in blocks of trials (120 to 220). The location of attention was cued by a few instruction trials (10 to 15) with a single sample stimulus before each block started. Selective attention was controlled between the two locations using both different reward sizes and different probabilities of test trials for the two sides. The reward size ratio for the valid location (attended) over the invalid location was 3.0–3.4 (mean ± 1 SEM reward size, monkey S, valid contralateral to recorded LC, 767 ± 30 μl, invalid ipsilateral to recorded LC, 317 ± 30 μl, valid ipsilateral to recorded LC, 780 ± 29 μl, invalid contralateral to recorded LC, 242 ± 19 μl; monkey P, valid contralateral to recorded LC, 661 ± 16 μl, invalid ipsilateral to recorded LC, 185 ± 4 μl, valid ipsilateral to recorded LC, 702 ± 11 μl, invalid contralateral to recorded LC, 169 ± 5 μl). The ratio of valid and invalid test stimulus probabilities for the locations was 3.0–3.4 (mean, monkey S, 3.4 [interquartile values, 3, 4]; monkey P, 3 [interquartile values, 3, 3]). Five different orientation changes (monkey S, 10°, 20°, 28°, 40°, 90°; monkey P, 10°, 17°, 28°, 48°, 90°) were used on the valid side (cued by instruction trials), and one intermediate orientation (28°) was used on the invalid side.

Both the test stimulus probability and reward size contributed to reliably control the spatial behavioral sensitivity (d’) between the stimulus locations. Varying reward size was specifically effective in controlling the perceptual decision criteria close to zero. In previous studies, we used with or without reward size differences to control behavioral d’ and criteria.^5,15 Furthermore, neuronal modulation at least in the visual cortex due to changes in behavioral d’ remained identical irrespective of the factors with or without varying reward size that motivated the monkey to shift their d’ during spatial attention task.¹⁵ Expectation of high reward size or high expectation of a test stimulus might be equivalent to a common valency (product of the test stimulus probability and reward size) associated with a stimulus location that drives the motivation to shift attention.⁵¹ Previous studies have shown that LC neurons are not sensitive to the absolute reward size; rather, their responses are sensitive to the mobilization of cognitive energy necessary to match the task demand.⁵²

Behavioral control of non-selective attentional effort

Spatially non-selective attentional effort (or intensity) (Figure 5) was controlled between low and high values in alternate blocks of trials (monkey S, 192 trial; monkey P, 200 trial) by varying task difficulty either easy or difficult at the two locations in opposite hemifields as described in previous studies.¹⁵ On each trial, orientation change was selected randomly from one of two values, a probe orientation change (monkey S, 40% of the trials; monkey P, 33% of trials) or contextual orientation change monkey S, 60% of the trials; monkey P, 67% of trials). Monkeys’ non-selective behavioral d’ was high (high nonselective attentional intensity) when the task difficulty locations were high (Δcontext orientation, 18° for monkey S and 16° for monkey P). A low nonselective behavioral d’ (low non-selective attentional effort) was obtained by making the task difficulty easy at both locations (Δcontext orientation, 90° for both monkeys S and P). Reward values for correct behavior responses and the test stimulus probability were always the same across blocks on both sides and fixed. On random half of the probe trials, the sample stimulus was paired with the optogenetic stimulation of LC (Figure 5).

Neurophysiological recordings

Neuronal signals from an extracellular multielectrode linear array (16 channel V-probes with 2 optic fiber channels; Plexon Inc.) were amplified, bandpass filtered (250 to 7,500 Hz), and sampled at 30 kHz using a data acquisition system (Cerebus, Blackrock Microsystems) (Figures 1 and 2). We simultaneously recorded from multiple single units and small multiunits over 51 sessions (29 sessions for monkey S; 22 sessions for monkey P). Spikes from each electrode were sorted offline (Offline Sorter, Plexon Inc.) by manually well-defining cluster boundaries using principal component analysis and waveform features. Well-isolated clusters were classified as single units from multiunits based on the isolation quality of unit clusters. The degree to which unit clusters were separated in two-dimensional (2D) spaces of waveform features (first three principal components: peak, valley, and energy) was measured by multivariate analysis of variance (MANOVA) F statistic using Plexon Offline Sorter (Plexon Inc.). A unit cluster of MANOVA P < 0.05 was considered a single unit that indicates that the unit cluster has a statistically different location in 2D space and that the cluster is statistically well separated.

Viral vectors

An excitatory opsin (channelrhodopsins, hChR2) was expressed unilaterally in NE-expressing neurons in the LC using a two-virus strategy.²² The first vector delivered Cre-recombinase (AAV9-DBH-Cre-P2A-mCherry-SV40pA, titer of 2.18×10¹³ vg/ml; Lot#17–502, Virovek, Inc.) under the control of the dopamine b-hydroxylase promoter (DBH; plasmid pEMS1113, catalog# 29042, Addgene). The second virus (AAV5-EF1a-DIO-hChR2(H134R)-EYFP, titer of 2.7×10¹³ vg/ml; SKU# VB4652, Vector Biolabs) delivered a Cre-recombinase-dependent hChR2 construct. A similar approach has been successfully used to target midbrain dopaminergic neurons in rhesus monkeys.²²

Virus injection

Injection sites were identified by electrophysiological recording of LC neurons characterized by waveform shape, sensitivity to arousing brief auditory noise stimuli, and were guided by neuronal response properties of other brain areas along the electrode trajectories leading to the LC, which included the superior colliculus, inferior colliculus, and the trochlear decussation in the brainstem.^17,18 We injected a volume of ~40 μl and ~50 μl of the two virus combination (1:1; titer 1×10¹² vg/ml) in monkey S and monkey P, respectively, in three separate unilateral locations in and around LC using custom-built injection cannulas. The injection cannula incorporated a glass pipette and stainless steel tubing with an insulated tungsten microwire for simultaneous neurophysiological recording and microinjection. We filled the injection cannula with the virus combination (~60 μl) and a calibrated glass tubing provided for visual confirmation of fluid injection. The virus-filled cannula was connected to a pneumatic pump (PV820, World Precision Instruments, customized to incorporate a constant holding pressure) using a patch-clamp microelectrode holder (AM Systems) for the microinjections. We injected at a rate of ~50 to 200 nl/min using a constant holding pressure. Behavior and electrophysiology started 10 weeks after the injection.

Optogenetic stimulation of LC-NE neurons

We used a 120 mW, 457 nm laser (Laserglow Technologies) with power at the end of the optic fiber in the range of ~50 mW. The power and timing were regulated by a Pockels cell (ConOptics) through the same software used for behavioral task. Laser light was delivered either as brief trains (e.g., 400 ms train of 10 ms pulses at 20 Hz for behavioral experiments) or 100 ms continuous pulse (for characterizing light-evoked excitability). For optogenetic stimulation sessions (Figures 3 and 4), we used custom made optrodes containing a 100 μm diameter optic fiber (NA, 0.66; Doric Lenses Inc.) attached to a tungsten sharp electrode (FHC Inc.). The sample stimulus was paired with the optogenetic stimulation of unilateral LC on random half of the valid trials (Figure 3) and probe trials (Figure 4).

Immunostaining

We performed histological confirmation in one of the experimental monkeys (monkey S; Figure 3C), as we are still conducting additional experiments on monkey P. First, the animal was deeply anesthetized and then euthanized, and then perfused with phosphate buffer solution (PBS) followed by 4% paraformaldehyde in PBS. The brain was removed, cryoprotected (in graded sucrose solutions, 10, 20 and 30%) and blocked. Sections of 50 μm were cut using a cryostat and stored in PBS. Then, the sections were washed in PBS containing 1% triton (1% PBS-T) for 30 min followed by a wash in PBS containing 0.3% triton (0.3% PBS-T) for 5 min. Then, the sections were placed sequentially in blocking buffer (5% goat serum in 0.3% PBS-T, Cat# 01–6201, ThermoFisher) for 3 h and in primary antibody solution (DBH antibody, 1:500, Cat# 22806, Immunostar; mCherry Monoclonal Antibody, 1:2000, Cat# 16D7, ThermoFisher; GFP Monoclonal Antibody, 1:1000, Cat# GF28R, ThermoFisher; in blocking buffer) for 48 h at 4°C. Then, the sections were washed in 0.3% PBS-T and incubated in secondary antibody solution (goat anti-rabbit IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 647, 1:500 (ThermoFisher Cat# A-21245); goat anti-rat IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor^™ 594, 1:500 (ThermoFisher Cat# A-11007); goat anti-mouse IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor^™ 514, 1:500 (ThermoFisher Cat# A-31555); ThermoFisher, in 0.3% PBS-T) for 4 h at room temperature. Following sequential washes in 0.3% PBS-T and PBS, sections were mounted with DAPI mounting media (Cat# P36981, ThermoFisher). Two days later, the sections were imaged with a confocal microscope.

QUANTIFICATION AND STATISTICAL ANALYSIS

Behavioral analysis

All completed trials were included in our analysis. Behavioral sensitivity (d’) and perceptual decision criteria (c) at a spatial location were measured from hit rates within nonmatch trials and false alarm (FA) rates within match trials using 1-dimensional signal detection theory^36,53 as d^′ = Φ⁻¹(H) − Φ⁻¹(F) and $c=\frac{1}{2}\left[{\Phi }^{-1}\left(H\right)+{\Phi }^{-1}\left(F\right)\right]$, where Φ^–1 is the inverse normal cumulative distribution function and H and F are the rates of hits and false alarms, respectively. The trial dynamics of behavioral d’ were computed by measuring the block-averaged d’ values for each attention condition (attend contra or ipsi to the recorded LC) within sliding windows of 30 trials (shifted by 1 trial) across sessions from both monkeys (Figure S4A).

Pupil area

All pupil area measurements were measured binocularly at 500 Hz while monkeys maintained fixation in the absence of a luminosity change using an infrared camera (EyeLink 1000, SR Research). Raw pupil areas were z-scored for each session and each eye separately. The mean pupil area was measured by averaging the z scored pupil area during the 400 ms after sample appearance. Unstimulated and optogenetically stimulated (LC) trials were compared by peak normalization of the trial-averaged z-scored pupil area (Figure 3C).

Detection of microsaccade

Microsaccades were detected using procedures described in previous studies^24,25 (Figure S5). Briefly, the horizontal and vertical eye positions were smoothed using a 10 ms sliding window. The eye positions were then differentiated with time to get velocities, which were further smoothed using a 10 ms moving window. The onset of a microsaccade was detected whenever the velocity exceeded a threshold for at least 10 ms. The velocity threshold was calculated on every trial independently for vertical and horizontal velocities as 6 times the medians of velocity distributions on that trial. Additionally, we visually inspected the quality of the microsaccades. We also analyzed the microsaccades using two other velocity thresholds (4 and 8 times the median) and durations (8 and 12 ms). The primary results regarding microsaccade distributions and directions remained identical. The proportion of microsaccades directed towards the sample stimuli were measures by the proportion of microsaccade directions within 15° (± 7.5°) solid angle towards sample stimulus over 200 ms periods immediate before (−200 to 0 ms) and after (0 to 200 ms) the onset of sample stimuli.

Estimation of shifts of central fixation eye position and fixation-matched trials

The center of fixation on each trial was measured by the average eye positions over a 300 ms period starting from the sample stimuli onset. Fixation positions across all trials within a session were binned into 10 equal spatial bins. Then the greatest common distribution of fixation centers across different experimental conditions and spatial bins was computed. The fixation center distribution of an experimental condition was matched to the common distribution by randomly selecting a different subset of trials (500 times) at every spatial bin.

Neuronal response

Spike counts in 2 ms bins were smoothed using a half-Gaussian kernel (standard deviation of 15 ms, rightward tail) for PSTHs. Neuronal responses were classified according to their average responses (greater than 250 ms pre-event period values over at least 20 consecutive bins; p < 0.05) to the sample stimulus (50 to 300 ms from sample onset; phasic), saccade (−50 to 200 ms from saccade onset; saccade responsive neurons) and fixation period (0 to 400 ms from fixation; non-phasic neurons). Spike trains were converted into z-scores (normalized with respect to mean and standard deviation of spike rates over 250 ms pre-event duration) to construct population PSTHs. Neuronal modulation with spatial attention was measured using neuronal sensitivity as follows: ${d^{’}}_{\mathit{\text{neuron}}}=\left({\mu }_{\mathit{\text{contra}}}-{\mu }_{\mathit{\text{ipsi}}}\right)/\sqrt{\frac{1}{2}\left({\sigma }_{\mathit{\text{contra}}}^2+{\sigma }_{\mathit{\text{ipsi}}}^2\right)}$, where μi and σi are the average and standard deviation of spike counts within 50 to 300 ms and −400 to 0 ms from sample stimuli onset, respectively, for phasic and non-phasic neurons (i, attended location, contralateral or ipsilateral to the recorded LC hemisphere).

Spike response latency

Spike rate latency was measured by the time to half peak response.⁵⁴ Mean spike rate PSTHs (aligned to sample stimulus onset) of each unit were smoothed with a Gaussian filter (sigma, 8 ms) for different attention and stimulation conditions. The SEM of the baseline response was estimated from the pre-sample (−200 to 0 ms from sample onset) responses during fixation. The peak response between 0 and 300 ms from the sample onset was measured with the requirement that it exceeded the mean baseline by 3.72*SEM. Units were not considered if the above criteria were not met. The response latency was computed as the time the PSTH reached half the difference between the peak response and mean baseline response.

Demixed principal component analysis

We examined how well LC neuronal activity during the presentation of sample stimuli could decode the focus of selective attention, and whether the response on a trial would be correct. These parameters were isolated from other stimuli and task variables using demixed principal component decompositions²¹ (Figure S3; Figure 2). All responsive LC units from both monkeys were separated into phasic and non-phasic neurons. Details of the demixed principal component decompositions have been described by Kobak et al.²¹ Briefly, neuronal population activity patterns were decomposed into a linear combination of specific components, each of which carries information of a single task variable. Mean-subtracted and trial-averaged spike trains of each neuron were decomposed into the sum of marginalized averages, each corresponding to a task variable and a noise term. This marginalization process ensures that the individual components are uncorrelated. A loss function that penalizes the difference between the marginalized data and the reconstructed full data is minimized using the least-square method. The reconstructed data are the full data projected with the decoders onto a low-dimensional latent space and then reconstructed with the encoders. Such a decomposition method offers reliable decoding accuracy measures even for relatively small explained variances by individual variables (Figure S3).²¹ Trials were classified into two attention conditions (attend contra versus ipsi) and two stimulus detection conditions (correct (hit, CR) versus error (miss, FA)), yielding four different trial configurations (Figures 2A and 2C). Single-trial spike rates were filtered with a half Gaussian kernel (σ = 30 ms) and subsampled at 100 Hz. We analyzed spike rates over 500 ms (50 time points, starting 100 ms before sample onset) for phasic responsive units and over 1000 ms (100 time points, starting 800 ms before sample onset) for non-phasic responsive units. Decomposition into demixed components was performed on training datasets (leave-one-out, 1000 repetitions). Attended location and detection were then decoded on the remaining cross-validated test trials using the top three components to estimate the decoding accuracy (Figures 2B and 2D). To assess the trial dynamics of decoding accuracy, spike trains from phasic LC neurons were analyzed over a 250 ms period (100 to 350 ms from sample stimuli onset) (Figure S4B). The spike counts were decomposed into demixed components using training datasets (leave-one-out, 1000 repetitions) within sliding windows of 10 trials (shifted by 1 trial) across blocks. Decoding accuracies were estimated using the top three components.

Statistical analysis

Unless otherwise specified, we used a paired t test and multifactor ANOVA to compare normally distributed datasets. Normality was checked using a Kruskal-Wallis test.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and virus strains
AAV9-DBH-Cre-P2A-mCherry-SV40pA	Virovek, Inc.	Custom AAV9 (Lot#17–502)
AAV5-EF1a-DIO-hChR2(H134R)-EYFP	Vector Biolabs	SKU# VB4652
Plasmid
pEMS1113 (DBH)	Addgene	Cat# 29042
Antibodies
DBH antibody (1:500)	Immunostar	Cat# 22806
mCherry Monoclonal Antibody, (1:2000)	ThermoFisher	Cat# 16D7
GFP Monoclonal Antibody (1:1000)	ThermoFisher	Cat# GF28R
goat anti-rabbit IgG (H+L) highly cross-adsorbed secondary antibody, Alexa Fluor 647 (1:500)	ThermoFisher	Cat# A-21245
goat anti-rat IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor™ 594 (1:500)	ThermoFisher	Cat# A-11007
goat anti-mouse IgG (H+L) cross-adsorbed secondary antibody, Alexa Fluor™ 514 (1:500)	ThermoFisher	Cat# A-31555
Deposited data
Data	This study	https://github.com/ghoshsupriya/LC-spike-modulation-associated-with-visual-spatial-attention.git
Software and algorithms
Lablib (software for behavioral task)	Custom	https://github.com/MaunsellLab/Lablib-Public-05-July-2016

Highlights.

• LC-NE neurons selectively spike to attended contralateral visual stimulus

• LC spike modulation is associated with correct perceptual detection

• Unilateral LC activation boosts contralateral perceptual d’, but not motor criteria

• The LC contribution to selective attention is distinct from arousal