Through experience with the world, brains learn to predict the sensory feedback generated by movement. For instance, learning the precise relationship between motor commands and the resulting visual feedback is the basis for both feedback control of movements as well as distinguishing self-generated from externally generated sensory feedback. One way that the brain may learn to do this is via predictive processing (Keller and Mrsic-Flogel, 2018; Rao and Ballard, 1999). In a sensorimotor version of this framework, the brain learns an internal model which transforms motor activity into a prediction of the sensory consequences. The framework postulates that specific neurons are responsible for computing discrepancies – known as prediction errors – between the prediction and the actual sensory inputs received. The neocortex has been shown to compute signals consistent with sensorimotor prediction errors (Eliades and Wang, 2008; Keller and Mrsic-Flogel, 2018; Schneider et al., 2018). In the primary visual cortex (V1), excitatory neurons in layer 2/3 compute mismatches between visual flow speed and locomotion speed (Attinger et al., 2017; Jordan and Keller, 2020; Keller et al., 2012; Zmarz and Keller, 2016). Such sensorimotor prediction error responses in the cortex depend on ongoing and developmental experience with sensorimotor coupling (Attinger et al., 2017; Schneider et al., 2018; Vasilevskaya et al., 2022; Widmer et al., 2022), and therefore the computation underlying them is actively learned. In predictive processing, one main function of prediction errors is to drive corrective plasticity in the internal model that generates predictions (Keller and Mrsic-Flogel, 2018).

During the first sensorimotor experience in life, prediction errors should strongly drive learning of an internal model of the world, while later in life, prediction errors should update internal models primarily when overall prediction errors are abundant (for instance, in novel or volatile situations). One way to implement such a developmental or contextual shift in the amount of plasticity induced by a particular prediction error would be a gating signal. The characteristics of a gating signal are that it is activated by prediction errors and that it permits increases in plasticity. In addition, because it is not driving plasticity, but simply gating it, it can be low dimensional (i.e. the same for all neurons). Prime candidates for a gating signal are neuromodulatory systems. Dopaminergic neurons in the midbrain signal reward prediction errors (Kim et al., 2020; Schultz, 2016) and action prediction errors (Greenstreet et al., 2022), which are thought to gate plasticity in downstream targets to support learning of stimulus-reward associations and goal directed behavior (Cox and Witten, 2019; Flagel et al., 2011; Reynolds et al., 2001; Steinberg et al., 2013). Indeed, the catecholamines – dopamine and noradrenaline – regulate and gate plasticity in the cortex (Choi et al., 2005; He et al., 2015; Seol et al., 2007). However, the midbrain dopaminergic system projects only sparsely to most of the sensory neocortex of the mouse, including V1 (Nomura et al., 2014), where visuomotor predictions are thought to be learned (Attinger et al., 2017; Widmer et al., 2022). Instead, the major catecholaminergic input to the cortex is the locus coeruleus (LC). LC activity has primarily been investigated in the context of classical stimulus presentations and reinforcement tasks. Such studies have shown that the LC responds to many different events, including unexpected task outcomes (Bouret and Sara, 2004; Breton-Provencher et al., 2022), novelty (Takeuchi et al., 2016; Vankov et al., 1995), and unexpected sensory stimuli in general (Aston-Jones and Bloom, 1981; Deitcher et al., 2019; Foote et al., 1980; Hervé-Minvielle and Sara, 1995). In zebrafish, noradrenergic neurons increase their activity during the prolonged visuomotor mismatches of failed swim attempts to drive ‘giving up’ behavior (Mu et al., 2019). We hypothesized that if the mouse LC is activated during similar visuomotor mismatches, this could represent visuomotor prediction error signaling that would function to gate prediction error driven plasticity in output circuits like the cortex.

To test fundamental aspects of this idea, we imaged calcium activity in LC axons in the cortex during visuomotor mismatches, and optogenetically manipulated LC activity levels during visuomotor experience to probe for experience-dependent changes in cortical activity. We found that LC axons in both sensory and motor cortical regions respond to unsigned visuomotor prediction errors, that is, unpredicted visual motion or unpredicted visual halts during locomotion. We then combined optogenetic LC axon stimulation with two-photon calcium imaging of layer 2/3 neurons in V1, to show that the LC greatly facilitates a form of plasticity consistent with predictive visuomotor learning: learned suppression of nasotemporal visual flow responses during forward locomotion. These results support the idea that a key function of the LC is to facilitate prediction-error-driven cortical plasticity.

We first assessed whether locus coeruleus (LC) axons convey sensorimotor prediction errors to the cortex. To record LC axonal activity, we expressed axon-targeted, Cre-dependent GCaMP6s in noradrenergic neurons of the LC via stereotactic viral vector injections in 13 NET-Cre mice (NET = norepinephrine transporter; Wagatsuma et al., 2018). We confirmed that these injections labelled only cells in the LC, and that GCaMP6s positive cells were also immunoreactive for tyrosine hydroxylase, a marker for catecholaminergic neurons (Figure 1—figure supplement 1A). For two-photon imaging, we implanted a 4 mm imaging window either over the right V1 and surrounding cortical regions (posterior cortex, 9 mice), or over the right primary and secondary motor cortices (anterior cortex, 4 mice). Prior to imaging, mice were habituated to a virtual reality (VR) system consisting of a virtual tunnel, with walls patterned with vertical sinusoidal gratings (Figure 1A). In the closed loop condition, locomotion was coupled to visual movement of the tunnel, and in open loop conditions, the walls moved independent of mouse locomotion. After habituation to head-fixation, we began imaging across up to five non-overlapping fields of view (FoV) in each imaging window, yielding a total of 31 posterior cortex FoVs and 9 anterior cortex FoVs. Axonal fluorescence showed strong correlations with pupil diameter (Figure 1—figure supplement 1B) typical for LC axons (Reimer et al., 2016), rapid responses to air puffs (Figure 1—figure supplement 1C) as previously reported (Breton-Provencher et al., 2022; Deitcher et al., 2019), and increases in fluorescence at locomotion onset that decayed during the locomotion bout (Figure 2—figure supplement 1) consistent with earlier work (Reimer et al., 2016). We found that within a FoV, individual axon segments exhibited highly correlated activity (Figure 1B and Figure 1—figure supplement 1D–F). Given that the algorithms that identify axons (i.e. determine whether two axon segments belong to the same axon) rely on temporal correlations of activity (Leinweber et al., 2017; Mukamel et al., 2009), axon segmentation is difficult in these data. Thus, to prevent oversampling of the data, we pooled data from all axon segments within a FoV for further analysis.

Figure 1 with 1 supplement see all

Download asset Open asset

Noradrenaline broadcasts visuomotor prediction errors across the dorsal cortex

Individual LC neurons can preferentially target different cortical areas (Chandler et al., 2019; Kebschull et al., 2016), but it is unclear whether these projection-specific neurons have differing response profiles. To assess the heterogeneity of LC signaling across the cortex, we first compared the calcium responses of LC axons between two imaging window locations: the anterior cortex (the primary and secondary motor cortex), and the posterior cortex (V1 and surrounding regions; Figure 2A). We found very similar responses to visuomotor mismatches, playback halts (stationary) and visual flow onsets (stationary; Figure 2B–C), as well as air puffs and locomotion onsets at these two locations (Figure 2B and D), and there were no significant differences in the responses between the two locations. These results are consistent with the idea that the LC sends similar visuomotor and locomotion related signals to distinct regions of the cortex, in line with recent findings showing that LC axons in different cortical regions show responses to the same sensory stimuli (Deitcher et al., 2019).

Figure 2 with 1 supplement see all

Download asset Open asset

Locus coeruleus axon stimulation has only a small effect on stimulus responses in layer 2/3 of V1

What could be the function of prediction errors signaled by the LC to the cortex? One possibility is that cortical prediction errors, like mismatch responses in V1 (Keller et al., 2012), are enhanced by LC output in order to augment error driven learning. Indeed, a prevalent idea is that noradrenaline is a mediator of gain modulation in the cortex (Ferguson and Cardin, 2020), including the gain increase in V1 during locomotion (Polack et al., 2013). To test whether the LC may enhance prediction error responses in cortical layer 2/3, we assessed the impact of transient LC axon stimulation on cortical responses using an optogenetic strategy (Figure 3A). First, we injected an AAV vector encoding Cre-dependent excitatory red-light activated opsin, ChrimsonR, into the right LC of 8 NET-Cre mice, while injecting a vector encoding jGCaMP8m into the right V1 to drive non-specific neuronal expression. An imaging window was then implanted over the right V1. Mice underwent habituation to the virtual reality system before calcium imaging of V1 neurons during (1) a closed loop condition with mismatches and (2) an open loop condition with 1 s fixed-speed visual flow stimuli. During imaging, we stimulated ChrimsonR by directing a 637 nm laser through the objective into the imaging window (27 mW/mm², in 15ms pulses presented at 20 Hz). This stimulation was presented (a) alone in closed loop conditions, (b) coupled to a random 50% of mismatches (Figure 3B), and (c) coupled to a random 50% of open loop visual flow stimuli. To assess the visual effects of optical stimulation of ChrimsonR (Danskin et al., 2015) in absence of the effects of optogenetically stimulating LC axons, we also imaged from a second group of six mice treated in an identical manner, but without ChrimsonR expression in the LC (control mice). Two mice were excluded from the LC-injected group due to low density of axonal labelling with tdTomato in V1, leaving 6 ChrimsonR-expressing mice in the dataset (see Methods).

Figure 3 with 2 supplements see all

Download asset Open asset

Optogenetic stimulation, when presented in isolation, resulted in dilations of the right pupil in mice expressing ChrimsonR in a manner that depended on the degree of axon labelling (Figure 3C and Figure 3—figure supplement 1A), but not in control mice. This likely resulted from antidromic activation of the LC, direct stimulation of which is known to produce strong ipsilateral pupil dilation (Liu et al., 2017), and increased our confidence that we were indeed driving LC axons with the parameters selected. Onset of the optogenetic stimulation laser caused a small calcium response across the population in ChrimsonR-expressing mice, which was indistinguishable from that caused in control mice (Figure 3D). This was likely due to a visual effect of the stimulation laser. The lack of a difference in this stimulation response between ChrimsonR-expressing mice and control mice indicated that optogenetic stimulation of LC axons did not have a direct measurable effect on calcium activity of L2/3 neurons in V1.

We then analyzed the effect of optogenetic activation of LC axons on mismatch and visual flow responses of the layer 2/3 population. To take account of visual effects of the laser, we compared the effect of the laser in ChrimsonR-expressing mice with that in control mice. Responses to visual flow onsets (i.e. at the beginning of the visual flow stimulus, or at the offset of mismatch) appeared transiently more pronounced during laser stimulation in ChrimsonR-expressing mice only (Figure 3E–F). This was most notable when plotting the difference in average response between trials with concurrent laser stimulation and those without: there were transient increases in response during laser-on trials at visual flow onset, and this effect was significantly different between ChrimsonR-expressing and control groups only during brief time-windows at the onset/resumption of visual flow (Figure 3E–F). However, this effect was small and not evident in average responses. Averaging across the full response duration, we found no significant effects of laser stimulation, both when comparing trials with and without laser stimulation in the ChrimsonR-expressing group, and when comparing the difference in response size between ChrimsonR-expressing and control mice (Figure 3E–F). No relationship could be found between the degree of LC axonal labelling and the effect of stimulation on average response size, so ChrimsonR expression-level was likely not the limiting factor (Figure 3—figure supplement 1B).

In V1, there is electrophysiological and molecular evidence for at least three different layer 2/3 pyramidal neuron types, including negative prediction error (NPE) neurons which respond to visuomotor mismatches, positive prediction error neurons (PPE) which respond to unexpected visual flow (Jordan and Keller, 2020; O’Toole et al., 2022), and an intermediate group. It is therefore conceivable that LC output could have differential effects on these groups that are masked when analyzed at the population level. To quantify potential cell-type-specific effects of LC axon stimulation, we thus split the population of neurons into three groups based on locomotion onset responses (see Figure 3—figure supplement 2 and Methods): the group with the strongest visuomotor mismatch responses (NPE), the group with the strongest visual flow responses (PPE), and an intermediate population (Other). We found no evidence for an effect of optogenetic stimulation of LC axons on average mismatch and visual flow responses across any of these cell groups (Figure 3G–H). We thus concluded that transient LC axon stimulation in the awake state has only a minor direct effect on the responses of neurons in layer 2/3 of V1. These results are largely in line with recent work demonstrating a lack of effect of LC stimulation on baseline firing in cortical pyramidal neurons, and an enhancement of transient responses to afferent input (Vazey et al., 2018).

Phasic LC output enhances sensorimotor plasticity in layer 2/3 of V1

If transient LC activity in the awake state does have a large impact on average layer 2/3 responses directly, what is the function of visuomotor prediction errors signaled by the LC in the cortex? The LC is thought to be involved in cortical plasticity (Kasamatsu et al., 1985; Kasamatsu and Pettigrew, 1976; Martins and Froemke, 2015; Shepard et al., 2015). Catecholamine receptors in V1 are known to modulate synaptic plasticity by activating intracellular signaling cascades and interacting with eligibility traces (Choi et al., 2005; He et al., 2015; Nomura et al., 2014; Seol et al., 2007). One plausible function of LC sensorimotor prediction errors is to gate the plasticity that underlies learning of sensorimotor predictions, for example, the prediction of visual feedback during locomotion. Primary sensory cortices can indeed learn to suppress the sensory consequences of locomotion, driving selective reductions of responses to locomotion-coupled inputs (Schneider et al., 2018; Widmer et al., 2022).

While our optogenetic stimulation of LC axons had only small measurable effects on mismatch or visual flow response size (Figure 3), we did notice one striking difference between ChrimsonR-expressing and control mice. In V1, there is a known increase in the size of visual responses during locomotion compared to stationary periods (Bennett et al., 2013; Erisken et al., 2014; Niell and Stryker, 2010; Pakan et al., 2016; Polack et al., 2013; Zmarz and Keller, 2016). Analyzing only trials where the optogenetic stimulation laser was not on, we found that this canonical gain increase was present for control mice but reversed in sign in ChrimsonR-expressing mice: Visual responses during locomotion were suppressed on average relative to stationary periods (Figure 3I). This resulted in a significantly lower locomotion modulation index (see Methods) for ChrimsonR-expressing relative to control mice (Figure 3I). Since episodic LC axon stimulation occurred throughout the whole imaging session (i.e. across 10 min in the closed loop condition, prior to the presentation of open loop visual flow stimuli), it is possible that the reversal of locomotion modulation index reflects a result of plasticity induced by earlier LC axon stimulation.

We wanted to directly test the idea that transient LC activation enhances plasticity that leads to a reduction in locomotion modulation index over the course of visuomotor coupling experience. To do this, we designed an experiment to assess the impact of LC axon stimulation on neuronal learning about visuomotor coupling (Figure 4A). We expressed jGCaMP8m in neurons in V1 of 19 mice. Twelve mice also had AAV vector injections into the LC to express ChrimsonR in NET-positive neurons, while the remaining seven underwent no injection into the LC to serve as controls for the non-optogenetic effects of the protocol. Mice were habituated to the VR as before and we then proceeded to two-photon imaging in layer 2/3 of V1. During the imaging session, mice would first be exposed to 5 min of 1 s fixed-speed open loop visual flow presentations, to quantify the gain-change of visual flow responses during locomotion (Figure 4A). Mice were then exposed for 5–10 min to either (a) a closed loop condition where visual flow and locomotion are coupled, or (b) an open loop replay of visual flow generated from a previous closed loop session. During this session, LC axons would be periodically optogenetically stimulated as before (15ms pulses, presented at 20 Hz for 1 s every 7 s on average) to simulate phasic LC responses (Figure 4A). Afterwards, the open loop visual flow presentations would be repeated to quantify how the previous LC axon stimulation affected visual responses. To assess the stimulus specificity with which visual responses changed, we presented both nasotemporal and temporonasal visual flow (mice see the former during forward locomotion in the closed loop condition; Figure 4A). If LC axon stimulation results in enhanced plasticity of locomotion based predictions of visual flow, we would expect the following observations: (1) Suppression of visual flow responses by locomotion is enhanced after stimulation of LC axons in closed loop conditions, (2) this suppression is specific to the type of visual flow seen during forward locomotion (nasotemporal), (3) the effect depends on ChrimsonR expression in the LC, and (4) the effect is absent when LC axons are instead stimulated during open loop conditions (i.e. the effect depends on visuomotor coupling).

Figure 4 with 2 supplements see all

Download asset Open asset

Optogenetic stimulation caused a degree of pupil dilation that was strongly predictable from axonal ChrimsonR expression levels in V1 quantified postmortem (R²=0.89, p<10^–5, 12 mice, linear regression) (Figure 4B). We split the LC-injected dataset into high ChrimsonR expression (six mice) and low ChrimsonR expression (six mice) according to V1 axonal expression level (Figure 4B and Figure 4—figure supplement 1A). Low ChrimsonR-expressing mice had optogenetically evoked pupil dilations that were not statistically distinguishable from controls, while high ChrimsonR-expressing mice had overt pupil dilations (Figure 4B and Figure 4—figure supplement 1B). For the following analyses, we excluded 8 (5) of 49 (25) sessions from the ChrimsonR (control) dataset due to lack of locomotion during those sessions (see Methods). At the beginning of the imaging session, prior to optogenetic stimulation, neurons in layer 2/3 showed a pronounced positive locomotion modulation index for both the nasotemporal and temporonasal visual flow responses, both in high and low ChrimsonR-expressing mice and in control mice (Figure 4C–E and Figure 4—figure supplement 1D). Quantifying visual flow responses in high ChrimsonR-expressing mice after optogenetic stimulation during closed loop visuomotor experience revealed a significant reduction in the locomotion modulation index for nasotemporal visual flow responses (Figure 4C). This originated from a reduction in the response to visual flow during locomotion, while responses during stationary periods remained stable (Figure 4G). This effect was specific to the direction of visual flow seen during forward locomotion, since locomotion modulation index was preserved for temporonasal visual flow responses (Figure 4D), with negligible changes in response size both during locomotion and stationary periods (Figure 4H). The effect was also specific to high ChrimsonR-expressing mice, as locomotion modulation index did not significantly change in both control (Figure 4—figure supplement 1D) and low ChrimsonR-expressing mice (Figure 4E and I).

The selective reduction in locomotion modulation index of nasotemporal visual flow responses in high ChrimsonR-expressing mice could be due to enhanced learning to suppress reafferent visual feedback seen during forward locomotion, or it could simply be due to general exposure to nasotemporal visual flow during optogenetic stimulation (e.g., enhanced adaptation). In the latter case, response size reduction should be the same if optogenetic stimulation takes place during open loop replay of the same visual flow seen during closed loop conditions. To assess this, we repeated the same experiment, but instead of stimulating LC axons during the closed loop condition, we stimulated LC axons during open loop replays of visual flow from previous closed loop sessions of the same mice (Figure 4A). Instead of a stimulus selective reduction in locomotion gain, this paradigm drove a nonsignificant increase in locomotion modulation index for both nasotemporal (Figure 4F and Figure 4—figure supplement 1F) and temporonasal visual flow (Figure 4—figure supplement 1E) responses. This confirmed that the reduction in locomotion modulation index after stimulation of LC axons in the closed loop condition (Figure 4C) depends on visuomotor coupling experience and is not simply due to exposure to nasotemporal visual flow.

We conclude that LC axon stimulation over the course of minutes increases locomotion driven suppression of visual flow responses in layer 2/3 in a manner that is specific to reafferent stimulus properties and dependent on both ChrimsonR expression and visuomotor coupling (Figure 4J). Importantly, the effect could not be explained by differences in locomotion behavior between any of the groups analyzed (Figure 4—figure supplement 2).

LC axon stimulation over minutes recapitulates visuomotor plasticity seen over days

In the virtual reality environment, mouse movement and visual flow are constrained to the forward-backward dimension, and the visuomotor gain is constant: changes in visuomotor gain that would usually occur during changes in environment geometry are absent. Mice head-fixed in the VR therefore experience a vastly reduced diversity of visuomotor coupling compared to freely moving conditions. The reduction in locomotion modulation index resulting from LC axon stimulation (Figure 4C) may therefore result from overlearning of this simplified visuomotor coupling.

If this interpretation is correct then extensive training within the VR should also lead to visual flow response suppression during locomotion even in mice without artificial LC axon stimulation. To assess this idea, we reanalyzed two-photon calcium imaging data from V1 layer 2/3 in coupled trained mice reared with visual experience entirely constrained to that of the visuomotor coupling in the VR across days (and otherwise dark reared) (Attinger et al., 2017; Widmer et al., 2022; Figure 5A). We compared this data to similar data from normally reared mice that also have experience of the virtual reality system, but that had experience of normal visuomotor coupling in freely moving conditions (Vasilevskaya et al., 2022; Zmarz and Keller, 2016; Figure 5A). Analyzing responses to nasotemporal visual flow onsets during open loop replay, we found that visual flow responses were amplified during locomotion relative to stationary periods in normally reared mice but were suppressed during locomotion in coupled trained mice (Figure 5B). This resulted in a significantly lower locomotion modulation index in coupled trained mice compared to normally reared mice (Figure 5C). The effect was not due to dark rearing, as non-coupled trained mice (also dark reared, but with only open loop virtual reality experience Figure 5A; Attinger et al., 2017) showed a positive locomotion modulation index on average that did not significantly differ from that of normally reared mice (Figure 5C).

Figure 5 with 1 supplement see all

Download asset Open asset

Qualitatively, the comparison of locomotion modulation index between normally reared and coupled trained mice (Figure 5B–C) was similar to the change in locomotion modulation index in ChrimsonR-expressing mice for nasotemporal visual flow after LC axon stimulation (Figure 4C). This suggests that the increase in locomotion driven suppression after LC axon stimulation is caused by a similar form of plasticity to that seen when visuomotor experience is constrained to the virtual reality across days. If this is the case, we should see similarities in the properties of the locomotion driven suppression. First, we looked at the locomotion speed dependence of the suppression. If the negative locomotion modulation index results from learning to suppress visual responses according to a locomotion driven prediction of visual flow, the degree of suppression should depend on locomotion speed. As locomotion speed increased, visual flow responses were increasingly suppressed in coupled trained mice (Figure 5D), while this was not evident in normally reared mice (Figure 5E). For nasotemporal visual flow responses in ChrimsonR-expressing mice, locomotion modulation index was positive and did not change with locomotion speed at the start of the imaging session, but after stimulation of LC axons in closed loop conditions, locomotion modulation index became increasingly negative as locomotion speed increased (Figure 5F–G). Thus, the locomotion driven suppression of visual flow emerging after stimulating LC axons in closed loop conditions was locomotion speed dependent in a similar manner to the locomotion driven suppression seen in coupled trained mice. Note that in high ChrimsonR-expressing mice, suppression of visual flow was seen at high locomotion speeds only (Figure 5G), but in coupled trained mice, some level of suppression was evident across all speeds (Figure 5E). This could be due to differences in visual flow speed presented: for coupled trained mice, the stimuli analyzed are visual flow onsets during open loop replay; thus, the visual flow speed will cover a similar range as the locomotion speed distribution of the mice. For high ChrimsonR-expressing mice, the stimuli are presented at a fixed visual flow speed that often exceeds the locomotion speed of the mouse. Thus, the fact that suppression only appears at the highest locomotion speed in these mice is consistent with a suppression of visual flow input based on a locomotion speed dependent expectation. In addition to the similarities in locomotion speed dependence of suppression in the two groups, the pattern of change in locomotion modulation index across different functional layer 2/3 cell groups also appeared similar between high ChrimsonR-expressing and coupled trained mice (Figure 5—figure supplement 1).

Overall, the pattern of differences in locomotion modulation was remarkably similar for ChrimsonR-expressing (Figure 5G, and Figure 5—figure supplement 1D) and coupled trained mice (Figure 5E and Figure 5—figure supplement 1B). This indicates that the functional consequence of plasticity evoked by minutes of LC axon stimulation during visuomotor coupling is similar to that seen after experience is developmentally constrained to visuomotor coupling for days.

The locus coeruleus (LC) has long been hypothesized by computational models to report quantities such as unexpected uncertainty (Yu and Dayan, 2005) or high level state-action prediction errors (Sales et al., 2019). What these quantities have in common is that they essentially are metrics for the inaccuracy of the brain’s internal models. One way to implement such a measure would be to integrate prediction errors from across the brain and then broadcast this signal back across the brain. Indeed, we found that the LC sends unsigned visuomotor prediction errors non-specifically across the dorsal cortex (Figures 1 and 2). Models hypothesizing that the LC reports the global inaccuracy of the brain’s internal models suggest that the purpose of this is to increase learning rates in output circuits like the cortex, enabling internal models to be modified more rapidly. In line with this, we found that stimulating cortical LC axons during closed loop visuomotor coupling over minutes reduces and even reverses the canonical locomotion related gain of visual flow responses in layer 2/3 of V1, in a manner specific to the direction of the movement-coupled stimulus (Figure 4). Surprisingly, this enhanced plasticity seems to happen even with only a small direct effect of LC axon stimulation on average calcium responses (Figure 3). This plasticity, on the timescale of minutes, recapitulates the effect of restricting visuomotor experience to the closed loop coupling in virtual reality across days (Figure 5), suggesting that it reflects an acceleration of a slower form of visuomotor plasticity (i.e., an increased learning rate).

The LC responds to unexpected stimuli of various modality, including reward prediction errors, punishment prediction errors, and novel or intense unexpected sensory stimuli (Aston-Jones and Bloom, 1981; Basu et al., 2022; Bouret and Sara, 2004; Breton-Provencher et al., 2022; Deitcher et al., 2019; Foote et al., 1980; Hervé-Minvielle and Sara, 1995; Su and Cohen, 2022; Takeuchi et al., 2016; Vankov et al., 1995). The responses we found in LC axons to the various visual flow stimuli were proportional to the change in the degree of error between visual flow and locomotion speeds (Figure 1). This supports a model in which the LC signals prediction errors, and combined with previous findings, suggests that prediction errors of all modalities are a general feature of LC signaling. Both positive and negative visuomotor prediction errors appear to be computed in layer 2/3 of the primary visual cortex (Jordan and Keller, 2020) and activate the anterior cingulate cortex (Heindorf and Keller, 2022) – an area that projects directly to the LC. Thus, the LC likely inherits prediction errors computed upstream in areas like the cortex. Based on the multimodality and unsigned nature of LC responses, and the integrate and broadcast nature of its input-output anatomy (Schwarz et al., 2015), our working model of LC function is as follows: The LC integrates prediction errors of all modalities to compute a measure of how inaccurately internal models are currently predicting the global sensory inputs of the animal. The LC does not signal what type of prediction error occurred, but what the global rate of prediction errors currently is (i.e. a measure of surprise). The resulting LC output does not directly drive model updating in cortical output areas, but it modulates the rate at which local cortical prediction errors drive plasticity in internal models. Thus, the specificity of how internal models are changed depends on ongoing prediction errors in the cortex, and the LC input acts to control the learning rate, modulating the extent to which these cortical prediction errors drive plasticity. This would allow the LC to be a controller of plasticity in internal models: when prediction error rates are high (i.e. during developmental, novel, or volatile situations), updating of internal models will be more rapid as a result of increased LC output. To support this hypothesis, it is important for future work to expand the range of LC stimulation parameters tested to identify the function between LC activation level and cortical plasticity, as well as cortical representations.

While the idea of a brain wide average over sensory prediction errors can explain many of the observed LC responses, other responses are less easily explained by prediction error. LC calcium responses to locomotion onsets, which have been shown previously in LC axons (Deitcher et al., 2019; Reimer et al., 2016), were generally larger than visual flow or visuomotor mismatch responses (Figure 2 and Figure 2—figure supplement 1). This could be the result of a direct movement related input unrelated to prediction errors, or the result of prediction errors that are not driven by differences between forward locomotion and backward visual flow. Increases in activity occurring after locomotion onset could reflect vestibulomotor prediction errors that result from the fact that the mouse is head-fixed, or visuomotor prediction errors for motor attempts we do not record, like turns or head movements. We do find that visuomotor prediction errors contribute to locomotion onset responses, as locomotion onset responses in the open loop condition are smaller during visual flow, when locomotion evoked visuomotor error is lower (Figure 2—figure supplement 1B). However, LC calcium activity also rises prior to locomotion onset, creating the additional possibility of an internally generated movement onset response. Other kinds of movement are also correlated with LC firing, including pup retrievals (Dvorkin and Shea, 2022) and lever pressing (Breton-Provencher et al., 2022). These movement onset responses could also reflect other forms of prediction error, including action prediction errors (Greenstreet et al., 2022) or temporal difference reward prediction errors (in cases where the movement is associated with reward expectation) similar to those documented in the dopaminergic system (Kim et al., 2020). Future work will need to investigate these possibilities to build a complete general model of LC computation.

Depleting noradrenaline has been shown to prevent forms of sensory-guided developmental plasticity in the cortex, including ocular dominance plasticity in V1 (Kasamatsu and Pettigrew, 1976) and changes in the tonotopic map in the auditory cortex (Shepard et al., 2015). Loss of function experiments can suffer from the potential confound that catecholamines may simply be necessary for normal neuronal activity, and the loss of plasticity may be a secondary effect of a disruption in neuronal function. Indeed, acute noradrenergic receptor block can silence neurons in layer 2/3, and massively reduce the amplitude of visual inputs (Polack et al., 2013). In gain-of-function experiments, electrical stimulation of the LC has been shown to partially restore ocular dominance plasticity in adult cats (Kasamatsu et al., 1985). Here, we also performed a gain of function experiment: we show that a few minutes of optogenetic LC axon stimulation in V1 of adult mice during visuomotor coupling can recapitulate the developmental effects of constrained exposure to visuomotor coupling in the virtual reality (Figures 4 and 5). While our optogenetic stimulation is likely far more constrained to LC neurons than previous electrical stimulations, we cannot rule out a systemic effect of the LC axon stimulation. We know that we are likely antidromically activating the LC with our optogenetic stimulation owing to the resulting pupil dilation (Figures 3C and 4B). Given that LC neurons are electrically coupled (McKinney et al., 2022) and have widely divergent axons, our stimulation likely has effects on other LC output targets, such as other neuromodulatory systems, and other areas of the cortex. Thus, the plasticity effect that we measure may not entirely result from LC output in V1 alone. However, since the LC will never be acting in isolation and broadcasts its output to different cortical areas, the functional impact of LC axon stimulation in this experiment is strongly suggestive of a role in facilitating plasticity. Future work will be needed to understand the circuit locus of plasticity generated by this paradigm, and its dependence on particular neuromodulators (e.g. dopamine and noradrenaline). The fact that the plasticity occurs despite only minor effects of LC stimulation on neuronal activity (Figure 3) indicates that the LC can either enhance plasticity without affecting activity levels, for instance via the molecular cascades or eligibility traces activated by GPCRs (He et al., 2015; Hong et al., 2022; Seol et al., 2007), and/or that a part of the relevant circuit outside of layer 2/3 is more strongly modulated by LC activation.

Locus coeruleus activation has been shown to facilitate learning rates at the behavioral level (Glennon et al., 2023; Glennon et al., 2019; Martins and Froemke, 2015). The behavioral relevance of the LC-gated plasticity in our study is yet to be demonstrated, but we can conjecture some perceptual functions of the plasticity based on its nature: learned suppression of reafferent visual flow during locomotion. This theoretically would be useful at the perceptual level for detection of external moving objects during movement, or for detecting changes in the geometry of the environment through which the animal is moving. Novel behavioral paradigms for assessing these percepts could be used in future studies to demonstrate the behavioral impact of the plasticity that we see.

Prediction-error-driven catecholamine release is thought to gate cortical synaptic plasticity during reinforcement learning (He et al., 2015; Roelfsema and Holtmaat, 2018). The results presented here expand this idea from reward associations to sensorimotor associations. We propose that a general function of catecholamines is to gate the plasticity underlying predictive learning across all modalities.

All data and analysis code are available at https://data.fmi.ch/PublicationSupplementRepo/, in the section for the group of Georg Keller, under the title 'The locus coeruleus broadcasts prediction errors across the cortex to promote sensorimotor plasticity'. Other previously published datasets used in this study (Figure 5) are located in the same repository, listed under the same titles as the published papers in which they were first used (Attinger et al., 2017; Vasilevskaya et al., 2022; Widmer et al., 2022; Zmarz and Keller, 2016). Data and analysis code are also available at Zenodo under https://doi.org/10.5281/zenodo.8006972. Core analysis and imaging code are available at https://sourceforge.net/projects/iris-scanning/ (copy archived at Widmer and Keller, 2023).

1. 1. Jordan R
   2. Keller GB

   (2023) Zenodo

   The locus coeruleus broadcasts prediction errors across the cortex to promote sensorimotor plasticity.

   https://doi.org/10.5281/zenodo.8006972

1. 1. Aston-Jones G
   2. Bloom FE

   (1981) Activity of norepinephrine-containing locus coeruleus neurons in behaving rats anticipates fluctuations in the sleep-waking cycle

   The Journal of Neuroscience 1:876–886.

   https://doi.org/10.1523/JNEUROSCI.01-08-00876.1981

   • PubMed
   • Google Scholar
2. 1. Attinger A
   2. Wang B
   3. Keller GB

   (2017) Visuomotor coupling shapes the functional development of mouse visual cortex

   Cell 169:1291–1302.

   https://doi.org/10.1016/j.cell.2017.05.023

   • PubMed
   • Google Scholar
3. Preprint

   1. Basu A
   2. Yang JH
   3. Yu A
   4. Glaeser-Khan S
   5. Feng J
   6. Krystal JH
   7. Li Y
   8. Kaye AP

   (2022) Prefrontal Norepinephrine Represents a Threat Prediction Error under Uncertainty

   bioRxiv.

   https://doi.org/10.1101/2022.10.13.511463

   • Google Scholar
4. 1. Bennett C
   2. Arroyo S
   3. Hestrin S

   (2013) Subthreshold mechanisms underlying state-dependent modulation of visual responses

   Neuron 80:350–357.

   https://doi.org/10.1016/j.neuron.2013.08.007

   • Google Scholar
5. 1. Bouret S
   2. Sara SJ

   (2004) Reward expectation, orientation of attention and locus coeruleus-medial frontal cortex interplay during learning

   The European Journal of Neuroscience 20:791–802.

   https://doi.org/10.1111/j.1460-9568.2004.03526.x

   • PubMed
   • Google Scholar
6. 1. Breton-Provencher V
   2. Drummond GT
   3. Feng J
   4. Li Y
   5. Sur M

   (2022) Spatiotemporal dynamics of noradrenaline during learned behaviour

   Nature 606:732–738.

   https://doi.org/10.1038/s41586-022-04782-2

   • PubMed
   • Google Scholar
7. 1. Chandler DJ
   2. Jensen P
   3. McCall JG
   4. Pickering AE
   5. Schwarz LA
   6. Totah NK

   (2019) Redefining noradrenergic neuromodulation of behavior: impacts of a modular locus coeruleus architecture

   The Journal of Neuroscience 39:8239–8249.

   https://doi.org/10.1523/JNEUROSCI.1164-19.2019

   • PubMed
   • Google Scholar
8. 1. Choi SY
   2. Chang J
   3. Jiang B
   4. Seol GH
   5. Min SS
   6. Han JS
   7. Shin HS
   8. Gallagher M
   9. Kirkwood A

   (2005) Multiple receptors coupled to phospholipase C gate long-term depression in visual cortex

   The Journal of Neuroscience 25:11433–11443.

   https://doi.org/10.1523/JNEUROSCI.4084-05.2005

   • Google Scholar
9. 1. Cox J
   2. Witten IB

   (2019) Striatal circuits for reward learning and decision-making

   Nature Reviews. Neuroscience 20:482–494.

   https://doi.org/10.1038/s41583-019-0189-2

   • PubMed
   • Google Scholar
10. 1. Danskin B
    2. Denman D
    3. Valley M
    4. Ollerenshaw D
    5. Williams D
    6. Groblewski P
    7. Reid C
    8. Olsen S
    9. Blanche T
    10. Waters J

    (2015) Optogenetics in mice performing a visual discrimination task: measurement and suppression of retinal activation and the resulting behavioral artifact

    PLOS ONE 10:e0144760.

    https://doi.org/10.1371/journal.pone.0144760

    • PubMed
    • Google Scholar
11. Preprint

    1. Deitcher Y
    2. Leibner Y
    3. Kutzkel S
    4. Zylbermann N
    5. London M

    (2019) Nonlinear Relationship between Multimodal Adrenergic Responses and Local Dendritic Activity in Primary Sensory Cortices

    bioRxiv.

    https://doi.org/10.1101/814657

    • Google Scholar
12. 1. Dombeck DA
    2. Khabbaz AN
    3. Collman F
    4. Adelman TL
    5. Tank DW

    (2007) Imaging large-scale neural activity with cellular resolution in awake, mobile mice

    Neuron 56:43–57.

    https://doi.org/10.1016/j.neuron.2007.08.003

    • PubMed
    • Google Scholar
13. 1. Dvorkin R
    2. Shea SD

    (2022) Precise and pervasive phasic bursting in locus coeruleus during maternal behavior in mice

    The Journal of Neuroscience 42:2986–2999.

    https://doi.org/10.1523/JNEUROSCI.0938-21.2022

    • Google Scholar
14. 1. Eliades SJ
    2. Wang X

    (2008) Neural substrates of vocalization feedback monitoring in primate auditory cortex

    Nature 453:1102–1106.

    https://doi.org/10.1038/nature06910

    • PubMed
    • Google Scholar
15. 1. Erisken S
    2. Vaiceliunaite A
    3. Jurjut O
    4. Fiorini M
    5. Katzner S
    6. Busse L

    (2014) Effects of locomotion extend throughout the mouse early visual system

    Current Biology 24:2899–2907.

    https://doi.org/10.1016/j.cub.2014.10.045

    • PubMed
    • Google Scholar
16. 1. Ferguson KA
    2. Cardin JA

    (2020) Mechanisms underlying gain modulation in the cortex

    Nature Reviews. Neuroscience 21:80–92.

    https://doi.org/10.1038/s41583-019-0253-y

    • PubMed
    • Google Scholar
17. 1. Flagel SB
    2. Clark JJ
    3. Robinson TE
    4. Mayo L
    5. Czuj A
    6. Willuhn I
    7. Akers CA
    8. Clinton SM
    9. Phillips PEM
    10. Akil H

    (2011) A selective role for dopamine in stimulus-reward learning

    Nature 469:53–57.

    https://doi.org/10.1038/nature09588

    • PubMed
    • Google Scholar
18. 1. Foote SL
    2. Aston-Jones G
    3. Bloom FE

    (1980) Impulse activity of locus coeruleus neurons in awake rats and monkeys is a function of sensory stimulation and arousal

    PNAS 77:3033–3037.

    https://doi.org/10.1073/pnas.77.5.3033

    • PubMed
    • Google Scholar
19. 1. Glennon E
    2. Carcea I
    3. Martins ARO
    4. Multani J
    5. Shehu I
    6. Svirsky MA
    7. Froemke RC

    (2019) Locus coeruleus activation accelerates perceptual learning

    Brain Research 1709:39–49.

    https://doi.org/10.1016/j.brainres.2018.05.048

    • PubMed
    • Google Scholar
20. 1. Glennon E
    2. Valtcheva S
    3. Zhu A
    4. Wadghiri YZ
    5. Svirsky MA
    6. Froemke RC

    (2023) Locus coeruleus activity improves cochlear implant performance

    Nature 613:317–323.

    https://doi.org/10.1038/s41586-022-05554-8

    • PubMed
    • Google Scholar
21. Preprint

    1. Greenstreet F
    2. Vergara HM
    3. Pati S
    4. Schwarz L
    5. Wisdom M
    6. Marbach F
    7. Johansson Y
    8. Rollik L
    9. Moskovitz T
    10. Clopath CM
    11. Stephenson-Jones M

    (2022) Action Prediction Error: A Value-Free Dopaminergic Teaching Signal That Drives Stable Learning

    bioRxiv.

    https://doi.org/10.1101/2022.09.12.507572

    • Google Scholar
22. 1. He K
    2. Huertas M
    3. Hong SZ
    4. Tie X
    5. Hell JW
    6. Shouval H
    7. Kirkwood A

    (2015) Distinct eligibility traces for LTP and LTD in cortical synapses

    Neuron 88:528–538.

    https://doi.org/10.1016/j.neuron.2015.09.037

    • PubMed
    • Google Scholar
23. Preprint

    1. Heindorf M
    2. Keller GB

    (2022) Reduction of Layer 5 Mediated Long-Range Cortical Communication by Antipsychotic Drugs

    bioRxiv.

    https://doi.org/10.1101/2022.01.31.478462

    • Google Scholar
24. 1. Hervé-Minvielle A
    2. Sara SJ

    (1995) Rapid habituation of auditory responses of locus coeruleus cells in anaesthetized and awake rats

    Neuroreport 6:1363–1368.

    https://doi.org/10.1097/00001756-199507100-00001

    • PubMed
    • Google Scholar
25. 1. Hong SZ
    2. Mesik L
    3. Grossman CD
    4. Cohen JY
    5. Lee B
    6. Severin D
    7. Lee HK
    8. Hell JW
    9. Kirkwood A

    (2022) Norepinephrine potentiates and serotonin depresses visual cortical responses by transforming eligibility traces

    Nature Communications 13:3202.

    https://doi.org/10.1038/s41467-022-30827-1

    • Google Scholar
26. 1. Jordan R
    2. Keller GB

    (2020) Opposing influence of top-down and bottom-up input on excitatory layer 2/3 neurons in mouse primary visual cortex

    Neuron 108:1194–1206.

    https://doi.org/10.1016/j.neuron.2020.09.024

    • PubMed
    • Google Scholar
27. 1. Kasamatsu T
    2. Pettigrew JD

    (1976) Depletion of brain catecholamines: failure of ocular dominance shift after monocular occlusion in kittens

    Science 194:206–209.

    https://doi.org/10.1126/science.959850

    • PubMed
    • Google Scholar
28. 1. Kasamatsu T
    2. Watabe K
    3. Heggelund P
    4. Schöller E

    (1985) Plasticity in cat visual cortex restored by electrical stimulation of the locus coeruleus

    Neuroscience Research 2:365–386.

    https://doi.org/10.1016/0168-0102(85)90047-1

    • PubMed
    • Google Scholar
29. 1. Kebschull JM
    2. Garcia da Silva P
    3. Reid AP
    4. Peikon ID
    5. Albeanu DF
    6. Zador AM

    (2016) High-Throughput mapping of single-neuron projections by sequencing of barcoded RNA

    Neuron 91:975–987.

    https://doi.org/10.1016/j.neuron.2016.07.036

    • PubMed
    • Google Scholar
30. 1. Keller GB
    2. Bonhoeffer T
    3. Hübener M

    (2012) Sensorimotor mismatch signals in primary visual cortex of the behaving mouse

    Neuron 74:809–815.

    https://doi.org/10.1016/j.neuron.2012.03.040

    • PubMed
    • Google Scholar
31. 1. Keller GB
    2. Mrsic-Flogel TD

    (2018) Predictive processing: a canonical cortical computation

    Neuron 100:424–435.

    https://doi.org/10.1016/j.neuron.2018.10.003

    • Google Scholar
32. 1. Kim HR
    2. Malik AN
    3. Mikhael JG
    4. Bech P
    5. Tsutsui-Kimura I
    6. Sun F
    7. Zhang Y
    8. Li Y
    9. Watabe-Uchida M
    10. Gershman SJ
    11. Uchida N

    (2020) A unified framework for dopamine signals across timescales

    Cell 183:1600–1616.

    https://doi.org/10.1016/j.cell.2020.11.013

    • PubMed
    • Google Scholar
33. 1. Leinweber M
    2. Ward DR
    3. Sobczak JM
    4. Attinger A
    5. Keller GB

    (2017) A sensorimotor circuit in mouse cortex for visual flow predictions

    Neuron 95:1420–1432.

    https://doi.org/10.1016/j.neuron.2017.08.036

    • Google Scholar
34. 1. Liu Y
    2. Rodenkirch C
    3. Moskowitz N
    4. Schriver B
    5. Wang Q

    (2017) Dynamic lateralization of pupil dilation evoked by locus coeruleus activation results from sympathetic, not parasympathetic, contributions

    Cell Reports 20:3099–3112.

    https://doi.org/10.1016/j.celrep.2017.08.094

    • Google Scholar
35. 1. Martins ARO
    2. Froemke RC

    (2015) Coordinated forms of noradrenergic plasticity in the locus coeruleus and primary auditory cortex

    Nature Neuroscience 18:1483–1492.

    https://doi.org/10.1038/nn.4090

    • PubMed
    • Google Scholar
36. Preprint

    1. McKinney A
    2. Hu M
    3. Hoskins A
    4. Mohammadyar A
    5. Naeem N
    6. Jing J
    7. Patel SS
    8. Sheth BR
    9. Jiang X

    (2022) Cellular and Circuit Organization of the Locus Coeruleus of Adult Mice

    bioRxiv.

    https://doi.org/10.1101/2022.03.01.481656

    • Google Scholar
37. 1. Meijering E
    2. Jacob M
    3. Sarria JCF
    4. Steiner P
    5. Hirling H
    6. Unser M

    (2004) Design and validation of a tool for neurite tracing and analysis in fluorescence microscopy images

    Cytometry 58A:167–176.

    https://doi.org/10.1002/cyto.a.20022

    • Google Scholar
38. 1. Mu Y
    2. Bennett DV
    3. Rubinov M
    4. Narayan S
    5. Yang CT
    6. Tanimoto M
    7. Mensh BD
    8. Looger LL
    9. Ahrens MB

    (2019) Glia accumulate evidence that actions are futile and suppress unsuccessful behavior

    Cell 178:27–43.

    https://doi.org/10.1016/j.cell.2019.05.050

    • PubMed
    • Google Scholar
39. 1. Mukamel EA
    2. Nimmerjahn A
    3. Schnitzer MJ

    (2009) Automated analysis of cellular signals from large-scale calcium imaging data

    Neuron 63:747–760.

    https://doi.org/10.1016/j.neuron.2009.08.009

    • PubMed
    • Google Scholar
40. 1. Niell CM
    2. Stryker MP

    (2010) Modulation of visual responses by behavioral state in mouse visual cortex

    Neuron 65:472–479.

    https://doi.org/10.1016/j.neuron.2010.01.033

    • PubMed
    • Google Scholar
41. 1. Nomura S
    2. Bouhadana M
    3. Morel C
    4. Faure P
    5. Cauli B
    6. Lambolez B
    7. Hepp R

    (2014) Noradrenalin and dopamine receptors both control cAMP-PKA signaling throughout the cerebral cortex

    Frontiers in Cellular Neuroscience 8:247.

    https://doi.org/10.3389/fncel.2014.00247

    • PubMed
    • Google Scholar
42. Preprint

    1. O’Toole SM
    2. Oyibo HK
    3. Keller GB

    (2022) Prediction Error Neurons in Mouse Cortex Are Molecularly Targetable Cell Types

    bioRxiv.

    https://doi.org/10.1101/2022.07.20.500837

    • Google Scholar
43. 1. Pakan JM
    2. Lowe SC
    3. Dylda E
    4. Keemink SW
    5. Currie SP
    6. Coutts CA
    7. Rochefort NL

    (2016) Behavioral-state modulation of inhibition is context-dependent and cell type specific in mouse visual cortex

    eLife 5:e14985.

    https://doi.org/10.7554/eLife.14985

    • PubMed
    • Google Scholar
44. 1. Polack PO
    2. Friedman J
    3. Golshani P

    (2013) Cellular mechanisms of brain state-dependent gain modulation in visual cortex

    Nature Neuroscience 16:1331–1339.

    https://doi.org/10.1038/nn.3464

    • PubMed
    • Google Scholar
45. 1. Rao RP
    2. Ballard DH

    (1999) Predictive coding in the visual cortex: a functional interpretation of some extra-classical receptive-field effects

    Nature Neuroscience 2:79–87.

    https://doi.org/10.1038/4580

    • PubMed
    • Google Scholar
46. 1. Reimer J
    2. McGinley MJ
    3. Liu Y
    4. Rodenkirch C
    5. Wang Q
    6. McCormick DA
    7. Tolias AS

    (2016) Pupil fluctuations track rapid changes in adrenergic and cholinergic activity in cortex

    Nature Communications 7:13289.

    https://doi.org/10.1038/ncomms13289

    • PubMed
    • Google Scholar
47. 1. Reynolds JNJ
    2. Hyland BI
    3. Wickens JR

    (2001) A cellular mechanism of reward-related learning

    Nature 413:67–70.

    https://doi.org/10.1038/35092560

    • PubMed
    • Google Scholar
48. 1. Roelfsema PR
    2. Holtmaat A

    (2018) Control of synaptic plasticity in deep cortical networks

    Nature Reviews. Neuroscience 19:166–180.

    https://doi.org/10.1038/nrn.2018.6

    • PubMed
    • Google Scholar
49. 1. Sales AC
    2. Friston KJ
    3. Jones MW
    4. Pickering AE
    5. Moran RJ

    (2019) Locus coeruleus tracking of prediction errors optimises cognitive flexibility: an active inference model

    PLOS Computational Biology 15:e1006267.

    https://doi.org/10.1371/journal.pcbi.1006267

    • PubMed
    • Google Scholar
50. Preprint

    1. Saravanan V
    2. Berman GJ
    3. Sober SJ

    (2020) Application of the Hierarchical Bootstrap to Multi-Level Data in Neuroscience

    bioRxiv.

    https://doi.org/10.1101/819334

    • Google Scholar
51. 1. Schneider DM
    2. Sundararajan J
    3. Mooney R

    (2018) A cortical filter that learns to suppress the acoustic consequences of movement

    Nature 561:391–395.

    https://doi.org/10.1038/s41586-018-0520-5

    • PubMed
    • Google Scholar
52. 1. Schultz W

    (2016) Dopamine reward prediction error coding

    Dialogues in Clinical Neuroscience 18:23–32.

    https://doi.org/10.31887/DCNS.2016.18.1/wschultz

    • PubMed
    • Google Scholar
53. 1. Schwarz LA
    2. Miyamichi K
    3. Gao XJ
    4. Beier KT
    5. Weissbourd B
    6. DeLoach KE
    7. Ren J
    8. Ibanes S
    9. Malenka RC
    10. Kremer EJ
    11. Luo L

    (2015) Viral-genetic tracing of the input-output organization of a central noradrenaline circuit

    Nature 524:88–92.

    https://doi.org/10.1038/nature14600

    • PubMed
    • Google Scholar
54. 1. Seol GH
    2. Ziburkus J
    3. Huang S
    4. Song L
    5. Kim IT
    6. Takamiya K
    7. Huganir RL
    8. Lee HK
    9. Kirkwood A

    (2007) Neuromodulators control the polarity of spike-timing-dependent synaptic plasticity

    Neuron 55:919–929.

    https://doi.org/10.1016/j.neuron.2007.08.013

    • PubMed
    • Google Scholar
55. 1. Shepard KN
    2. Liles LC
    3. Weinshenker D
    4. Liu RC

    (2015) Norepinephrine is necessary for experience-dependent plasticity in the developing mouse auditory cortex

    The Journal of Neuroscience 35:2432–2437.

    https://doi.org/10.1523/JNEUROSCI.0532-14.2015

    • Google Scholar
56. 1. Steinberg EE
    2. Keiflin R
    3. Boivin JR
    4. Witten IB
    5. Deisseroth K
    6. Janak PH

    (2013) A causal link between prediction errors, dopamine neurons and learning

    Nature Neuroscience 16:966–973.

    https://doi.org/10.1038/nn.3413

    • PubMed
    • Google Scholar
57. Preprint

    1. Su Z
    2. Cohen JY

    (2022) Two Types of Locus Coeruleus Norepinephrine Neurons Drive Reinforcement Learning

    bioRxiv.

    https://doi.org/10.1101/2022.12.08.519670

    • Google Scholar
58. 1. Takeuchi T
    2. Duszkiewicz AJ
    3. Sonneborn A
    4. Spooner PA
    5. Yamasaki M
    6. Watanabe M
    7. Smith CC
    8. Fernández G
    9. Deisseroth K
    10. Greene RW
    11. Morris RGM

    (2016) Locus coeruleus and dopaminergic consolidation of everyday memory

    Nature 537:357–362.

    https://doi.org/10.1038/nature19325

    • PubMed
    • Google Scholar
59. 1. Vankov A
    2. Hervé-Minvielle A
    3. Sara SJ

    (1995) Response to novelty and its rapid habituation in locus coeruleus neurons of the freely exploring rat

    The European Journal of Neuroscience 7:1180–1187.

    https://doi.org/10.1111/j.1460-9568.1995.tb01108.x

    • PubMed
    • Google Scholar
60. Preprint

    1. Vasilevskaya A
    2. Widmer FC
    3. Keller GB
    4. Jordan R

    (2022) Locomotion-Induced Gain of Visual Responses Cannot Explain Visuomotor Mismatch Responses in Layer 2/3 of Primary Visual Cortex

    bioRxiv.

    https://doi.org/10.1101/2022.02.11.479795

    • Google Scholar
61. 1. Vazey EM
    2. Moorman DE
    3. Aston-Jones G

    (2018) Phasic locus coeruleus activity regulates cortical encoding of salience information

    PNAS 115:E9439–E9448.

    https://doi.org/10.1073/pnas.1803716115

    • PubMed
    • Google Scholar
62. 1. Wagatsuma A
    2. Okuyama T
    3. Sun C
    4. Smith LM
    5. Abe K
    6. Tonegawa S

    (2018) Locus coeruleus input to hippocampal CA3 drives single-trial learning of a novel context

    PNAS 115:E310–E316.

    https://doi.org/10.1073/pnas.1714082115

    • PubMed
    • Google Scholar
63. 1. Widmer FC
    2. O’Toole SM
    3. Keller GB

    (2022) Nmda receptors in visual cortex are necessary for normal visuomotor integration and skill learning

    eLife 11:e71476.

    https://doi.org/10.7554/eLife.71476

    • PubMed
    • Google Scholar
64. Software

    1. Widmer F
    2. Keller GB

    (2023) Iris, version swh:1:rev:1f1b5616aaaf6cf862d2f4b467b65b479a2b0416

    Software Heritage.

    https://archive.softwareheritage.org/swh:1:dir:f7082d123de339338a4a2585f1b1609fac3f23cc;origin=https://svn.code.sf.net/p/iris-scanning/code;visit=swh:1:snp:698c5e65f00c35df665a9a370afc49323890cf86;anchor=swh:1:rev:1f1b5616aaaf6cf862d2f4b467b65b479a2b0416
65. 1. Yu AJ
    2. Dayan P

    (2005) Uncertainty, neuromodulation, and attention

    Neuron 46:681–692.

    https://doi.org/10.1016/j.neuron.2005.04.026

    • PubMed
    • Google Scholar
66. 1. Zmarz P
    2. Keller GB

    (2016) Mismatch receptive fields in mouse visual cortex

    Neuron 92:766–772.

    https://doi.org/10.1016/j.neuron.2016.09.057

    • PubMed
    • Google Scholar

Author details

1. Rebecca Jordan

   Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland

   Present address

   Simons Initiative for the Developing Brain, University of Edinburgh, Edinburgh, United Kingdom

   Contribution

   Conceptualization, Data curation, Software, Formal analysis, Funding acquisition, Investigation, Visualization, Methodology, Writing – original draft, Writing – review and editing

   For correspondence

   rjordan3@ed.ac.uk

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-4871-6265

2. Georg B Keller

   1. Friedrich Miescher Institute for Biomedical Research, Basel, Switzerland
   2. Faculty of Sciences, University of Basel, Basel, Switzerland

   Contribution

   Resources, Supervision, Funding acquisition, Methodology, Writing – original draft, Project administration, Writing – review and editing

   Competing interests

   No competing interests declared

   "This ORCID iD identifies the author of this article:" 0000-0002-1401-0117

• 11,308

  views

• 416

  downloads

• 51

  citations

Views, downloads and citations are aggregated across all versions of this paper published by eLife.