Report

How Visual Stimuli Activate Dopaminergic Neurons at Short Latency

See allHide authors and affiliations

Science  04 Mar 2005:
Vol. 307, Issue 5714, pp. 1476-1479
DOI: 10.1126/science.1107026

Abstract

Unexpected, biologically salient stimuli elicit a short-latency, phasic response in midbrain dopaminergic (DA) neurons. Although this signal is important for reinforcement learning, the information it conveys to forebrain target structures remains uncertain. One way to decode the phasic DA signal would be to determine the perceptual properties of sensory inputs to DA neurons. After local disinhibition of the superior colliculus in anesthetized rats, DA neurons became visually responsive, whereas disinhibition of the visual cortex was ineffective. As the primary source of visual afferents, the limited processing capacities of the colliculus may constrain the visual information content of phasic DA responses.

Sensory stimuli that are biologically salient because of their novelty, intensity, or reward value elicit a stereotyped phasic (short-latency <100 ms; short-duration ∼100 ms) increase in firing rate of midbrain DA neurons in a variety of mammals (13). If not reinforced, responses to novel stimuli become habituated rapidly. The responses to rewarding stimuli also decline if stimuli can be predicted. When reward is signaled by an arbitrary stimulus, the phasic DA response shifts from the primary reward to the predicting stimulus. If, under these circumstances, a predicted reward fails to materialize, there is a brief pause in the ongoing activity of DA neurons. These findings have led to the influential suggestion that DA neurons provide the brain's reinforcement learning mechanisms with a “reward prediction error” signal that may be used to adjust future behavioral response probabilities (46). However, DA neurons exhibit robust responses to a wider class of stimuli than those unambiguously related to reward (2, 7); this suggests that the phasic DA signal may have a broader role than reward alone (8). An important strategy for decoding the phasic DA signal would be to identify and then to elucidate the perceptual properties of the sensory pathways providing input to DA neurons. Surprisingly, very little is known about the source(s) of the short-latency phasic sensory input to DA neurons. A candidate structure is the superior colliculus, a retino-recipient nucleus in the dorsal midbrain with direct efferent projections to dopamine-containing regions of the ventral midbrain (9). The experimental rationale of the present study was based on a recent report (10) that, in the deep layers of the superior colliculus, which project directly to DA neurons (9), visual sensitivity is suppressed by anesthesia and can be restored temporarily by local injections of disinhibitory pharmacological agents.

Simultaneous electrophysiological recordings from the superior colliculus deep layers and electrophysiologically identified DA neurons in the substantia nigra (N = 18), or ventral tegmental area (N = 17), of anesthetized rats (11) revealed in all cases (N = 35) that neither the superior colliculus (Fig. 1A, top left) nor midbrain DA neurons (Fig. 1A, top right) responded to a whole-field light flash. However, after microinjection of the γ-aminobutyric acid type A (GABAA) receptor blocker, bicuculline, into the superior colliculus (directly adjacent to the recording electrode), local neurons became sensitive to the light flash (Fig. 1A, bottom left, and 1B, left). Following the onset of collicular responses to the light stimulus, 30 out of 35 (85.7%) DA neurons also exhibited a clear short-latency response to the light. More than half of the light-activated DA neurons (17 out of 30; 56.6%) responded with an initial excitatory phase, of which nearly half were polyphasic (8 out of 17; 47.0%) (Fig. 1A, bottom right). In contrast, the ongoing activity of other DA neurons was initially suppressed (13 out of 30; 43.3%) (Fig. 1B, right), with about half (6 out of 13; 47.1%) exhibiting further excitatory components. The response latencies of neurons in the superior colliculus were reliably shorter (40.3 ± 3.1 ms) than those of corresponding DA neurons (113 ± 14.2 ms) (t = 5.4; df = 29; P < 0.001). Control experiments (N = 4) involving a comparable enhancement of visual processing in the striate cortex (Fig. 1C, left, produced by direct application of bicuculline to the cortical surface) left DA neurons unresponsive to the light stimulus (Fig. 1C, right).

Fig. 1.

Disinhibition of collicular deep layers induced phasic visual responses locally and in DA neurons. (A) Initially, raster displays and peri-stimulus histograms show that collicular neurons and a simultaneously recorded DA neuron were unresponsive to a regular (0.5 Hz) light flash (vertical dotted line) (top graphs). After a collicular microinjection of bicuculline, both local neurons and the DA neuron were excited at short latency by visual stimulation (bottom graphs). (B) Example of a light-evoked inhibitory response of a DA neuron after collicular disinhibition. (C) DA neurons remained insensitive to light after bicuculline-induced facilitation of the flash-evoked field potential in visual cortex.

We proceeded to consider factors that could differentiate the DA neurons by their initial excitatory or inhibitory reactions to the light. (i) Variables associated with the injections of bicuculline had no apparent effect. Analysis of histologically verified coordinates of the injection sites, and the distribution of the neural activity marker c-Fos (12) (evoked by the direct excitatory action of bicuculline, fig. S1A), revealed no systematic differences between excited, inhibited, and nonresponsive DA neurons (11). (ii) Excitatory and inhibitory responses were observed in both the substantia nigra (9:5) and ventral tegmental area (8:8) (fig. S1B) (χ2 = 0.63; df = 1: P > 0.05). (iii) The difference in spontaneous firing rate in the 500 ms before each light flash for excited (1.41 ± 0.3 spikes) and inhibited neurons (2.0 ± 0.24 spikes) was not statistically reliable (t = 1.485; df = 28; P = 0.173). (iv) The mean spike widths of DA neurons exhibiting excitatory (4.17 ± 0.25 ms) and inhibitory (3.83 ± 0.23 ms) responses were not reliably different (fig. S1C) (t = 0.972; df = 28; P = 0.339). All but two (28 out of 30) of our putative DA neurons satisfied the recently proposed additional criterion for distinguishing DA neurons (duration to the first negative peak >1.1 ms) (13); the difference between the mean values of this parameter for excited (1.4 ± 0.09 ms) and inhibited (1.5 ± 0.07 ms) neurons was also not significant (t = 0.292; df = 28; P = 0.773). (v) No reliable differences were found between the mean latencies (excited = 121.18 ± 18.48 ms; inhibited = 105.38 ± 22.88 ms; t = 0.543; df = 28; P = 0.592) or response durations (excited = 189.41 ± 29.27 ms; inhibited = 192.46 ± 32.85 ms; t = 0.69; df = 28; P = 0.945) of light-responsive DA neurons. Differences in the initial response may therefore reflect differential activations of excitatory and inhibitory components of the tectonigral pathway (9) and their respective control of individual DA neurons.

An important aspect of our methodology was that the light stimulus was spatially and temporally predictable (every 2 s, for hundreds of trials). Measures of response magnitude throughout the experimental sessions (Fig. 2A) revealed no signs of the rapid habituation reported by others (3). Rather, the magnitude of the responses appeared more accurately to reflect the waxing and waning of bicuculline's effect in the superior colliculus (Fig. 2A).

Fig. 2.

Flash-evoked activation of collicular and DA responses fails to show habituation to predictable stimuli. (A) Measures of the response magnitude of an excited (black histogram) and inhibited (white histogram) DA cell, plus associated collicular multiunit activity (black and white symbols, respectively, and right scale), throughout a single trial. For each flash, the DA and collicular event count in the 300 ms preceding the stimulus was subtracted from the event count in the 300 ms post stimulus. Each bar/point represents the mean of 30 of these values, consecutively throughout the trial. (B) Electrochemical response magnitudes throughout the trial illustrated in Fig. 3A. After normalizing the baseline current to 0 at the time of stimulus onset, the oxidation current was recorded 200 ms after each light flash (see Fig. 3C left). Each bar represents the mean of five of these values, consecutively throughout the trial.

Because the light stimulus evoked both excitatory and inhibitory DA neuronal responses, coupled with the observation that the release of dopamine from terminals is not always closely related to the electrophysiological activity of DA neurons (14), we conducted experiments using fixed potential amperometry (15, 16) to measure light-evoked release of dopamine into target regions of the neostriatum. Without additional treatment, whole-field light flashes caused no detectable release of dopamine into the neostriatum of anesthetized rats (11). However, after local injections of bicuculline into the deep layers of the superior colliculus (N = 23), short-latency (mean 153.7 ± 25.1 ms), short duration (mean 331.4 ± 19.3 ms) electrochemical responses were recorded from the neostriatum in every case. Normally, this effect was evident only after signal averaging (Fig. 3, B and C). However, in some examples, each light flash evoked a clearly observable electrochemical response (Fig. 3A). In these cases, the magnitude of the electrochemical response to individual flashes also showed no evidence of rapid habituation to the temporally predictable stimuli (Fig. 2B). Injection of the selective dopamine reuptake blocker, nomifensine, in each case (N = 4) increased both the amplitude and duration of the light-induced electrochemical response (Fig. 3, A and B). Combined injections of the serotonergic and nor-adrenergic reuptake blockers, fluoxetine and desmethyl imipramine (N = 2), had no effect (Fig. 3C).

Fig. 3.

Electrochemical oxidation currents reveal light-evoked release of dopamine into the striatum. (A) Onset of changes in striatal dopamine oxidation current induced by light flashes (red) during a collicular microinjection of bicuculline, before (bottom trace) and after pretreatment with nomifensine (20 mg/kg) (top trace). (B) Peristimulus averaging of light-evoked electrochemical responses (red, N = 30) and interleaved averages of control data (blue, N = 30), before (left) and after (right) pretreatment with nomifensine (20 mg/kg). (C) Averages of light-evoked responses (red, N = 30) and interleaved control data (blue, N = 30), before (left) and after (right) pretreatment with a combination of fluoxetine (20 mg/kg) and desmethyl imipramine (20 mg/kg).

The present results provide complementary electrophysiological and electrochemical evidence for the phasic modulation of midbrain DA neurons by discrete visual stimuli. Light flashes only increased the phasic release of dopamine, whereas DA neurons showed both excitatory and inhibitory responses; these results may be partly explained by the supra-additive accumulation of dopamine in the forebrain when DA neurons switch to burst firing mode (17). However, phasic DA responses were observed only when neurons in the superior colliculus were released from inhibition associated with the anesthetic (10) and were themselves responsive to light stimuli. Comparable disinhibition of early cortical visual processing left DA neurones unresponsive (Fig. 1C), and later cortical processing capable of object recognition typically has latencies equal to or longer than those of DA neurones (18). In addition, other retino-recipient systems (e.g., pretectal and accessory optic nuclei) have been associated mainly with ocular reflexes or responses to photoperiod (19). Therefore, the superior colliculus could be the primary, if not exclusive, source of presaccadic information concerning the unexpected occurrence of biologically salient visual events.

In unanesthetized animals, neurons in the deep layers of the superior colliculus (2022) and midbrain DA neurons (3) are both exquisitely sensitive to unexpected novel visual events, but they become habituated rapidly if stimuli become predictable or are not maintained by association with the primary reward (3, 20). In the anesthetized preparation we used in these studies, visual sensitivity was observed only when the suppressive effects of anesthesia were relieved by local disinhibitory injections of bicuculline into the superior colliculus. Given that such disinhibition can block both the behavioral (23) and electrophysiological signs of habituation (Figs. 1, A and B, and 2), it is relevant that both measures of DA activation in the present study also showed consistent responses to predictable stimuli over hundreds of consecutive trials (Fig. 2). This suggests that the mechanisms responsible for mediating habituation (24) and reinforcement-related modulations of sensory processing in the superior colliculus (25) would be able to regulate the often-reported habituation and reward-related dishabituation of DA neurons to neutral visual stimuli (3).

In most vertebrate species, unpredicted visual events are represented by the superior colliculus in terms of a restricted range of stimulus dimensions according to a well-defined, spatially organized retinotopic map (2022). In mammals, most visually responsive cells in the superior colliculus are transiently activated 40 to 60 ms after the appearance, disappearance, or movement of a stimulus within a specific region of the visual field. Collicular neurons respond poorly, if at all, to the contrast, velocity, wavelength, or geometric configuration of visual stimuli. Most experimental studies designed to evaluate the reward prediction error hypothesis of DA function [e.g. (26, 27)] have used tasks that can be solved on the basis of luminance change and/or position of specific reward-related visual stimuli. It is interesting that both characteristics can be coded by neurons in the superior colliculus (2022). It is likely, therefore, that the capacity of DA neurons to distinguish various classes of stimuli at short latency depends largely on presaccadic visual processing in the superior colliculus and the extent to which it can be modulated by associations with reinforcing stimuli.

Supporting Online Material

www.sciencemag.org/cgi/content/full/307/5714/1476/DC1

Materials and Methods

Fig. S1

References and Notes

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article