|Home | About | Journals | Submit | Contact Us | Français|
Cholinergic neurons originating from the basal forebrain innervate the entire cortical mantle. Choline-sensitive microelectrodes were used to measure the synaptic release of cortical acetylcholine (ACh) at a sub-second resolution in rats performing a task involving the detection of cues. Cues that were detected, defined behaviorally, evoked transient increases in cholinergic activity (at the scale of seconds) in the medial prefrontal cortex (mPFC), but not in a non-associational control region (motor cortex). In trials involving missed cues, cholinergic transients were not observed. Cholinergic deafferentation of the mPFC, but not motor cortex, impaired cue detection. Furthermore, decreases and increases in pre-cue cholinergic activity predicted subsequent cue detection or misses, respectively. Finally, cue-evoked cholinergic transients were superimposed over slower (at the time scale of minutes) changes in cholinergic activity. Cortical cholinergic neurotransmission is regulated on multiple time scales to mediate the detection of behaviorally significant cues and to support cognitive performance.
Attentional capacities and mechanisms, such as the sustained readiness for input processing, the ability to monitor and discriminate between multiple stimulus sources and modalities, and associated executive processes such as response selection, error detection and effortful control, collectively determine the efficacy with which stimuli control behavior. Ascending neuronal projection systems, particularly the cholinergic and noradrenergic projections arising from basal forebrain and brainstem areas, respectively, have been proposed to contribute to attentional performance by modulating the processing of information in the fronto-parietal attentional network (Aston-Jones and Cohen, 2005; Everitt and Robbins, 1997; Hasselmo and McGaughy, 2004; Mesulam, 1990; Posner and Dehaene, 1994; Sarter et al., 2005a, 2006). The persistent attentional impairments that result from lesions of the basal forebrain or the selective removal of the cortical cholinergic input system indicated the necessary role of this neuromodulator for attentional performance (Chiba et al., 1995; Dalley et al., 2004; McGaughy et al., 1996, 2000, 2002; Muir et al., 1992, 1994; Turchi and Sarter, 1997; Voytko et al., 1994). Furthermore, studies measuring acetylcholine (ACh) release, using microdialysis, revealed increases in cortical ACh release specifically in association with demands on attentional processes but not with the basic behavioral operations associated with cognitive task performance (Arnold et al., 2002; Dalley et al., 2001; Himmelheber et al., 2000; McGaughy et al., 2002; Passetti et al., 2000).
However, the precise cognitive operations supported by changes in cortical cholinergic activity have remained unknown. The low temporal resolution of measures of ACh release using microdialysis (minutes) limits the attribution of changes in cholinergic neurotransmission to specific behavioral or cognitive operations. Moreover, such measures of ACh release supported the traditional notion that this neuromodulator system acts at a time scale of minutes to influence cortical “arousal” states. However, the presence of a highly potent metabolizing enzyme for the neurotransmitter, acetylcholinesterase (AChE) and fast ionotropic nicotinic acetylcholine receptors (nAChRs) suggest that the functions of the forebrain cholinergic system are not sufficiently described by such notions.
Cortical cholinergic inputs, particularly to prefrontal regions, have been hypothesized to mediate the detection of cues (Sarter et al., 2005a). The term “detection” refers to multiple cognitive processes involving “…the entry of information concerning the presence of a signal into a system that allows the subject to report the existence of the signal by an arbitrary response indicated by the experimenter” (Posner et al., 1980). This definition further implies that “detection” involves response preparation and response timing, response outcome expectation and the timing of such outcome. The hypothesis that the cortical cholinergic input system mediates cue detection is consistent with neurophysiological evidence indicating that ACh augments the processing of thalamic inputs (Ashe et al., 1989; Kilgard and Merzenich, 1998; Tremblay et al., 1990; Weinberger, 2003) and that the effects of lesions of the cortical cholinergic input system on attention performance selectively manifest in trials in which cues are presented, while sparing the animals’ ability to reject non-cue events (McGaughy et al., 1996). However, direct evidence indicating that the cholinergic system is selectively active during cue detection has not been available, due largely to the absence of methods for the monitoring of cholinergic activity at a sufficiently high temporal resolution.
To test the hypothesis that cholinergic activity in the PFC mediates cue detection, we employed, for the first time in task-performing animals, ceramic-based multi-site microelectrode arrays for the electrochemical measurement of synaptic ACh release at a sub-second resolution (Burmeister and Gerhardt, 2003; Burmeister et al., 2003; Parikh et al., 2004). The measurement scheme underlying this technique is illustrated in Figure S1 (Supplemental Materials). Our previous experiments indicated the validity of this technique in terms of measuring choline resulting from acetylcholinesterase (AChE)-induced hydrolysis of newly released ACh (Parikh et al., 2004, 2006; Parikh and Sarter, 2006). Cholinergic activity was recorded in the mPFC and a non-associational control region (motor cortex) of animals performing a cued-appetitive response task (Figure 1). This task involves the presentation of a cue predicting subsequent reward delivery and therefore evoking attentional shifts from ongoing behavior to the monitoring of the two reward ports (“detection)”. Although this task involves less well defined demands on attentional processes than operant procedures involving computerized control of levers and reward delivery devices, it allows for manual operation of task events and thus is devoid of sources of static energy that were found to interfere, despite extensive shielding, with the recording of small currents (picoamperes). The collective results from these experiments indicate that the cortical cholinergic input system acts on multiple time scales (at the scales of seconds, tens of seconds, and minutes) to support cue detection and attentional performance.
Animals reached criterion performance for each stage of learning of the cued appetitive response task within about 2 weeks of training. The latencies between cue presentation and reward retrieval decreased continuously during the two stages of task acquisition (Fig. 1C), as indicated by a significant effect of day (stage 1, 10-s cue followed by immediate reward: F(4,20)=27.25, p<0.001; stage 2: 1-s cue followed by reward 6±2s later: F(4,20)=8.98, p<0.001).
In sessions during which cholinergic activity was recorded, animals detected significantly more cues than they missed (58.7±2.3 % of the cues were detected; 25 trials/session; t(10)=5.03, p<0.001; Fig. 2A). As would be expected, the latencies between reward delivery and reward retrieval were longer in trials in which cues were missed (t(10)=2.26, p=0.048; Fig. 2B).
Details concerning electrode preparation, in vitro calibration and electrode properties in vivo following completion of the recording experiments are described in Supplemental Materials. Amperometric recordings of cholinergic activity in the mPFC, but not motor cortex (Supplemental Materials), revealed transient increases that were evoked by cues that were detected (Fig. 2C–G). Cue-evoked cholinergic signal amplitudes were significantly higher for detected cues when compared with missed cues (highest choline signal levels observed during the 6±2s cue-reward interval; t(10)= 4.21, p=0.002; Fig. 2G). The time required for cholinergic signal amplitudes to decrease by 50% from peak (t50) was 3.17±0.27 s. As will be further substantiated below, during trials involving missed cues, cholinergic activity remained unchanged from pre-cue levels (Fig. 2D,F).
Additional analysis indicated that reward delivery and retrieval did not evoke cholinergic activity. First, choline signal levels recorded for 2 s prior to and 5 s following reward delivery did not differ by trial type (detected/missed; t(10)=1.18, p=0.27). Second, in trials involving missed cues, choline signal levels recorded for 5 s following the (missed) cue and following reward delivery did not differ (t(10)=2.17, p=0.10; Fig. 2F). The conclusion that reward-related processes did not confound cholinergic activity is further supported by the demonstration of regular cue-evoked cholinergic transients in catch trials not involving reward delivery, and by the absence of such transients early into the acquisition of the task (for these results see Supplemental Materials).
As the definition of detection involves the initiation of a behavioral response that indicates the entrance of a behaviorally significant cue into the processing stream (Introduction), the onset of the cue-evoked behavioral response was expected to correlate with the onset of the increase in cholinergic activity. Such increase in cholinergic activity was defined as the time point, relative to cue presentation, when cholinergic activity increased by 25% over pre-cue levels. As illustrated in Figure 2H, the time of onset of the choline spike correlated significantly with the onset of the behavioral shift (Pearson’s r=0.79, p<0.001).
In this task, the efficacy of the cue detection process is indicated by response latencies. Choline signal amplitudes correlated significantly with the latencies between cue presentation and reward retrieval (Pearson’s r=−0.37, p=0.045). Analysis of the regression between these two variables indicated that an increase in choline signal amplitude by 1 μM was associated with a decrease of 1.75 s in response latency.
The evidence described above was based on recordings in the mPFC of rats performing the cued appetitive response task involving a 6±2 s interval between cue and reward delivery (Fig. 1A). Cholinergic activity was recorded in a separate group of animals trained to perform the cued appetitive response task involving a shorter (2±1 s; Fig. 1A) interval, in order to test the following hypothesis. If cue-evoked cholinergic transients merely reflect the sensory encoding of the cue, the timing of cue-evoked cholinergic activity should be insensitive to variation of the interval between cue and reward delivery. In contrast, if variation of this interval causes variation of the timing of the cue-evoked cholinergic transients, such a finding would indicate that cholinergic transients reflect a shift in the timing of cue-evoked cognitive operations that collectively define detection (Introduction). As illustrated in Figure 3, the latency from cue presentation to the (detected) cue-evoked choline signal peak amplitude was significantly shorter in animals performing the task involving the shorter cue-reward interval (t(53)=9.26, p<0.001; Fig. 3B). The amplitudes of the cholinergic transients did not differ between the two task versions (t(9)=1.72, p>0.12). As was the case for recordings from the mPFC of animals performing the task involving the longer cue-reward interval, cholinergic activity evoked by detected cues was significantly higher when compared with missed cues (t(8)=6.97, p<0.001). Cholinergic activity in trials involving missed cues and reward delivery-evoked port approach remained at pre-trial levels (Fig. 3D; see below for statistical results).
Based on the choline signal population data for detected trials from both task versions, over the entire 16 s period (see Fig. 2E and and3C),3C), the effects of the variation of the cue-reward interval were indicated by a significant interaction between the effects of time (data across 16 s) and cue-reward interval (long, short) on choline signal levels (main effect of time: F(1,31)=13.28, p<0.001; main effect of interval: F(1,53)=21.38, p<0.001; time × interval: F(31,1643)=10.72, p<0.001). In the analysis of choline signal levels recorded during trials in which the cue was missed, neither an effect of time or interval nor an interaction between these two factors were found (both p>0.05), reflecting the absence of changes in cholinergic activity (Fig. 2F, ,3D).3D). Cue-evoked cholinergic transients were not observed in separate experiments in which cholinergic activity was recorded in the motor cortex (Supplemental Materials).
In the analysis of cholinergic signal levels across trials involving cue detection and misses, respectively, systematic relationships between pre-cue trends in cholinergic signal levels in the mPFC and trial outcome (detection or miss) were discovered. For a systematic analysis of this relationship, data from a 20-s period prior to the cue was boxcar-filtered and the slope of the linear regression was determined (see Supplemental Methods). As illustrated in Figure 4A, in 80% of trials involving cue detection, mPFC pre-cue cholinergic activity showed a negative trend; conversely, 83% of misses were preceded by increases in cholinergic activity (χ2=24.15, p<0.001). Moreover, for trials with detected cues, steeper decreases in pre-cue cholinergic activity correlated with greater amplitudes of cue-evoked cholinergic activity (Pearson’s r=−0.32, p=0.01).
A similar result was found in the analysis of cholinergic activity recorded in the motor cortex (76% and 72%, respectively; χ2=9.70, p=0.002; Fig. 4B). The magnitude of these trends did not differ between mPFC and motor cortex (Fig. 4C; decreases preceding cue detection: t(41)=0.038, p=0.97; increases preceding misses: t(41)=0.93, p = 0.36).
In order to confirm that the demonstration of evoked cholinergic activity, measured by choline-sensitive microelectrodes, requires the presence of cholinergic terminals, cholinergic activity was recorded following the unilateral removal of cholinergic inputs to the recording region (see Methods). In contrast to bilateral cholinergic deafferentation of the mPFC (below), such restricted deafferentation is insufficient to impair attentional performance (Gill et al., 2000) and, likewise, did not affect the proportion of cues that was detected (t(9)=1.75, p=0.22). Detected cue-evoked cholinergic activity was not observed in the deafferented recording region, confirming the validity of the measure in terms of reflecting ACh released from cholinergic neurons (Fig. 5A,B).
Bilateral removal of mPFC cholinergic inputs decreased the proportion of detected cues (F(3,16)=8.68, p=0.001; Fig. 5C). Multiple comparisons indicated that this impairment was present during all three weeks of post-surgery training and testing (all p<0.025). The number of port approaches was recorded across test sessions (see Methods), regardless of whether such approaches were evoked by cue or reward delivery. The effects of the lesions on this measure were analyzed in order to reveal potential confounds based on general exploratory or activity changes. Although the lesion produced a significant effect on this measure (F(3,16)=3.46, p=0.041), multiple comparisons indicated that this was due to an increased frequency of port approaches observed during the second week after the infusions of the immunotoxin (Fig. 5D). Immunotoxin-induced deafferentation typically reaches asymptotic levels two weeks post-injection (Waite et al., 1994).
In contrast to the effects of bilateral cholinergic deafferentation of the mPFC, a similar deafferentation of the motor cortex did not affect cue detection rate (F(3,16)=0.55, p=0.67; see Supplemental Materials).
The transient increases in cholinergic activity that were recorded in the mPFC during trials involving detected cues were superimposed over more slowly changing (on the scale of minutes), or tonic, changes in cholinergic activity. Such tonic cholinergic activity was also observed in the motor cortex (Fig. 6). ANOVA confirmed that session-related changes in cholinergic activity occurred in both cortical regions (main effect of time: F(39,351)=2.13, p<0.001) and they did not differ in magnitude (main effect of region: F(1,9)= 0.32, p=0.59).
Performance-associated increases in mPFC tonic cholinergic signal levels were positively correlated with the amplitudes of cue-evoked cholinergic transients (Pearson’s r=7.21, p<0.001; Fig. 6B) and with a greater proportion of detected cues (analyzed over blocks of 5 trials each; r=0.46, p=0.01). Tonic signal levels recorded in the motor cortex were not correlated with performance (r=0.04, p=0.86). Furthermore, the total number of port approaches, a measure of task-related locomotor and exploratory activity, did not correlate with tonic levels of cholinergic activity recorded in mPFC or motor cortex (both p>0.05). Session-related tonic cholinergic activity corresponded with levels of ACh release measured by using microdialysis in both cortical regions (Supplemental Materials).
In animals trained to perform the task and placed into the test chamber without activating the task, no such tonic changes in cholinergic activity were observed, indicating that performance of the task is necessary to evoke such tonic changes, and that context alone and expectation of performance were not sufficient in evoking tonic increases in cholinergic activity (mPFC: F(5,17)=0.49, p=0.78; motor cortex: F(5,17)=0.83, p=0.55; Fig. 6A,C). Finally, session-related tonic changes in mPFC cholinergic activity were not observed following unilateral removal of cholinergic inputs to the recording region (F(5,29)=0.77, p=0.58).
The results from these experiments support the following main conclusions. Transient or “phasic” increases in PFC cholinergic activity are evoked by attended cues. In trials involving missed cues, the delivery of reward triggered port approach and reward retrieval; as these events did not evoke cholinergic transients, cholinergic transients mediate cue-evoked cognitive operations, but not port approach and reward retrieval. This conclusion is further supported by the evidence from catch trials not involving reward delivery, and from trials early into the acquisition of the task when rewards were delivered and retrieved but cues did not yet evoke a behavioral response. The demonstration of the shift in the timing of cholinergic transients in response to shorter cue-reward intervals is consistent with the hypothesis that these transients mediate a cognitive operation, as opposed to merely indicating the sensory processing of the cue. As removal of cholinergic inputs to the mPFC, but not motor cortex, impaired cue detection, cue-evoked cholinergic activity in the mPFC is necessary for cue detection. Performance session-related, tonic changes in cholinergic activity occur over minutes, with higher tonic levels predicting greater amplitudes of phasic signals and enhanced cue detection (as indicated by shorter cue-reward retrieval latencies). Finally, pre-cue increases or decreases in cholinergic activity, observed over tens of seconds prior to cue presentation, predict subsequent misses or cue detection, respectively.
Removal of cholinergic inputs to the mPFC, but not motor cortex, impaired cue detection. As pre-cue cholinergic trends as well as task session-related tonic changes in cholinergic activity were also recorded in motor cortex, and were also abolished as a result of deafferentation, mPFC cholinergic activity is necessary for cue detection (below). The significant correlation between the amplitudes of these transients and response latencies, and the temporal left-shift of these transients in response to shorter cue-reward intervals further substantiate this conclusion.
As discussed in the Introduction, cue detection involves a range of cognitive processes, including attentional shifts away from ongoing, task-irrelevant activities to task-related behavioral and cognitive processes, including reward port monitoring, response rule processing and response preparation, reward anticipation, and the timing of responses and reward delivery. The present evidence is consistent with the hypothesis that cue-evoked cholinergic transients mediate cue detection.
The present evidence collectively rejects the possibility that reward delivery and reward retrieval evoked transient increases in cholinergic activity. First, reward port approach and reward retrieval also occurred in trials involving missed cues; yet cholinergic transients were not evoked by these events. Second, in trials involving detected cues, reward delivery occurred during the decay of the cholinergic transient; therefore, potential reward delivery-associated cholinergic spikes would have been readily observed. Third, in catch trials not involving reward delivery, cue-evoked cholinergic transients were identical to those observed in regular trials, indicating that reward delivery and retrieval did not confound cue-evoked cholinergic transients. Fourth, early into training while cues did not yet control behavior but while rewards were delivered and effectively retrieved, cholinergic transients were not observed. Therefore, the presence or absence of transient cholinergic activity indicates the differences between the cognitive/behavioral operations elicited by the distal (cue) versus proximal (reward delivery-associated) conditioned stimuli. For spatially and temporally distal stimuli to guide behavior they need to trigger cognitive operations such as attentional shifts away from task-irrelevant activities toward anticipation and timing of the reward, port monitoring, response rule processing and the timing of the response (Holland, 1993; Holland and Gallagher, 1999). In contrast, stimuli that are spatially and temporally bound with reward delivery can elicit port approach and reward retrieval without requiring such cognitive operations.
As discussed earlier (Sarter et al., 2005a, 2006), detection represents a top-down process that requires representation of the presence of the cue and information about the associative significance of the cue. Consistently predictive cues evoke attentional shifts toward outcome-related behaviors and events and, as indicated by the present results, such shifts are necessarily mediated by transient increases in cholinergic activity in the mPFC. Increases in mPFC cholinergic neurotransmission are hypothesized to be necessary for recruitment of prefrontal neuronal assembles that orchestrate, top-down, the components of the detection process. Therefore, in the absence of cholinergic inputs to the PFC, cues are missed at a higher frequency and, in animals performing more demanding attention tasks, performance is persistently disrupted (references above).
Results from neurophysiological recordings of basal forebrain neuronal activity correspond with the present conclusions. First, evidence for phasic as well as tonic firing characteristics of basal forebrain neurons were described (Detari et al., 1999). Second, neurophysiological studies conducted in task-performing primates indicated that basal forebrain neuronal activity reflects decision-making processes, cue-evoked reward expectation and timing (Richardson and DeLong, 1990; Wilson and Rolls, 1990).
The processes underlying missed cues remain necessarily speculative. Given the parameters of cue presentation (1 s duration, ceiling-mounted), it is unlikely that the cue failed to stimulate the retina; rather, misses demand an explanation in terms of post-sensory, cognitive processes. This view is supported by the observation, based on videotape analyses, that missed cues triggered brief (<1 s) disturbances in the sequencing of grooming behavior but failed, by definition, to trigger termination of such behavior. Effective cue detection involves a state of readiness for input processing, meaning the allocation of attentional resources for input processing and the disengagement from ongoing behavior and task-irrelevant cognitive activity. A miss could be attributed to a low readiness for input processing and may be similar to phenomena described as inattentional blindness or attentional lapses (Simons and Chabris, 1999; Weissman et al., 2006).
The present evidence suggests that pre-cue decreases in cholinergic activity in the mPFC and motor cortex, and therefore perhaps cortex-wide, foster subsequent cue detection, while increases in pre-cue cholinergic activity were followed by misses (Fig. 4). Moreover, for recordings in the mPFC, steeper pre-cue decreases predicted greater cue-evoked cholinergic signal amplitudes and therefore faster response latencies. Therefore, pre-cue negative slopes in cholinergic activity are hypothesized to indicate, or even contribute to, a more effective manifestation of the brain resting default state, while positive slopes reflect a less effective suspension of task-related activity. This interpretation is consistent with findings from human studies indicating that attentional lapses are more likely if task-irrelevant cognitive activity prevents the return to the resting default state (Weissman et al., 2006). The hypothesis that trends in pre-cue cholinergic activity determines trial outcome requires research in which these trends are controlled experimentally by, for example, varying the duration of the ITI and thus by controlling the suspension of task-related processes.
Session-related, tonic increases in cholinergic activity recorded in the mPFC correlated with higher cue detection rates and with greater amplitudes of cue-evoked cholinergic transients. Furthermore, greater amplitudes predicted shorter response latencies. These findings suggest functionally significant interactions between the multiple components of cholinergic neurotransmission. Minute-based changes in mPFC cholinergic activity contribute to the general readiness for cortical input processing and therefore also influence the efficacy of the detection process.
Because lesions of the cholinergic input to the motor cortex did not affect the animals’ performance, the role of tonic cholinergic activity elsewhere in the cortex remains unclear. The performance of cognitive tasks involving multimodal stimuli and complex instrumental behaviors may generally be optimized by tonic cholinergic activity, including in the motor cortex to support skilled motor responses (Conner et al., 2005; Conner et al., 2003). As the present task did not tax such motor functions, the removal of cholinergic inputs to motor cortex was inconsequential.
The present evidence is consistent with a model that assumes multiple cholinergic modules and a regulation of cholinergic activity in modality- and cortical area-specific manner (Zaborszky, 2002). Moreover, our results suggest that performance-related cholinergic activity manifests on multiple time scales. The anatomical characteristics of the basal forebrain cholinergic system do not suggest a topographic organization that would readily explain the presence of such functional modules and multiple modes of action (Mesulam, 1990; Sarter and Bruno, 1997; Zaborszky et al., 1999). However, there is evidence to propose that the cholinergic inputs to the mPFC represent a critical component of neuronal circuits that consist of prefrontal projections to the basal forebrain and the nucleus accumbens (NAc), and from the NAc to the basal forebrain, suggesting that in addition to local, intra-PFC mechanisms contributing to the orchestration of cholinergic transients, larger loops involving mesolimbic circuitry influence mPFC cholinergic activity and therefore cue detection (Neigh et al., 2004; Zmarowski et al., 2005, 2007). It is intriguing to speculate that phasic dopamine signals recorded in the NAc in response to cues predicting reward (Day et al., 2007) contribute, via NAc projections to the basal forebrain, to the manifestation of mPFC cue-evoked cholinergic transients. Thereby, reward prediction may be integrated with prefrontally controlled attentional shifts and response processing, collectively giving rise to the cholinergically mediated detection of cues.
The findings that transient increases in cholinergic activity mediate cue detection and that the cholinergic system acts on multiple time scales to support cognitive performance form the basis for a significant expansion of hypotheses concerning the role of cholinergic dysregulation in the manifestation of the cognitive symptoms of neuropsychiatric disorders and the dementias (Mesulam, 2004; Sarter et al., 2005b). Specifically, abnormalities in the orchestration of cue-evoked cholinergic transients may precede more global and structural decline in cholinergic function. Dysregulated transients would be expected to disrupt the ability to utilize external stimuli in order to shift attentional resources toward goals. Indeed, such deficits have been extensively documented in patients with Alzheimer’s disease and attributed to dysregulation and loss of cholinergic neurons (Mesulam, 2004). Likewise, deficits in target detection represent a core cognitive symptom of schizophrenia (Braff and Light, 2004) and have been attributed to dysregulation in forebrain cholinergic systems (Sarter et al., 2005b). Future efforts designed to understand the role of cholinergic dysfunction in the manifestation of cognitive impairments and the usefulness of cholinergic treatments will need to dissociate between the regulation and functions of the multiple phasic and tonic components of forebrain cholinergic neurotransmission.
Adult male Fisher/Brown Norway hybrid rats (FBNF1; Harlan, Indianapolis, IN), weighing 250–300 g at the beginning of the experiments were used. Animals were individually housed in a temperature (23°C) and humidity-controlled (45%) environment and on a 12hr light/dark cycle (lights on at 06:30 a.m.). Food and water was available ad libitum until the commencement of behavioral training. Rats were mildly food-deprived by providing 30 g of lab chow in their home cages following each daily test session, thereby maintaining at least 85% of their free-feeding body weights. Water was always available ad libitum. All procedures were conducted in adherence with protocols approved by the University Committee on Use and Care of Animals (UCUCA) of the University of Michigan.
The test environment is described in Supplemental Materials. For two weeks, food-deprived animals were handled daily for 5 min and then placed in to the test chamber for an additional 30 min. Four pieces (12 mg each) of Kellogg’s Fruit Loop® were placed in the chamber to allow familiarization with the food used subsequently as reinforcement. Once animals rapidly consumed the pellets, they were then trained to accept the pellets presented by using a plastic tweezers inserted through one of the two food ports (random selection).
Training of the cued appetitive response task consisted of two stages. In the first stage, the light cue was illuminated for 10 s and a pellet was delivered immediately after cue offset (25 trials/day). The inter-trial interval (ITI) was 60±30 s. Animals were trained in this version until latencies between cue-onset and pellet retrieval were <13 s for at least 75% of the trials/session. During the second stage of training, cue duration was shortened to 1 s and the latency between cue presentation and pellet delivery was increased to 6±2 s or, in a separate group of animals, 2±1 s (Fig. 1A). In addition, the ITI was increased to 90±30 s. Individual training/test sessions lasted for approximately 40 min and included an 8-min waiting period between placing the animal into the chamber and the onset of the first trial (25 trials total). Training continued until the latencies between cue presentation and reward delivery were =9 s in at least 80% of the trials. Figure 1C depicts the learning of this response in terms of decreasing response latencies during the two stages of training.
Animals’ performance was videotaped for the off-line classification of trials by experimenters blind to the choline recording data. Trials involving cue detection were classified as such based on cue-evoked behavior, characterized by disengagement from ongoing behavior (typically grooming), orientation to, and monitoring of, the two reward ports. Trials involving a failure to detect the cue (“missed cue”) were characterized by the absence of cue-elicited changes in behavior (Fig. 1B). It is important to note that in trials involving missed cues, the salient auditory and visual stimuli associated with food delivery reliably evoked the animals’ approach to the baited port and food retrieval, albeit with longer latencies between cue presentation and food retrieval when compared with trials involving cue detection (see Results). Thus, trials involving missed cues served as an additional control for the test of the hypothesis that port approach and reward consumption and associated locomotor activity did not evoke transient cholinergic activity (see Results). On average, animals detected ~65% of the cues. After reaching stable criterion performance in the task, animals were habituated, for one additional week, to performing the task in a shielded test chamber used for subsequent electrochemical recordings. Animals were then prepared either for microelectrode implantation or lesion surgery (below).
During post-surgery re-training, which lasted 4–6 days/sessions, animals were placed into the chambers 90 min prior to task-onset to foster habituation to tethering (described in Supplemental Materials). Post-surgery training sessions, including sessions during which cholinergic activity was recorded, were video-taped. Trials were classified as having involved detected or missed cues off-line by experimenters blind to the recording data.
The following measures of behavioral performance were obtained or calculated from each test session; 1) the number and proportion of cues that were detected; 2) for trials involving detected cues, the latency between cue presentation and disengagement of ongoing behavior. To generate this measure, experimenters blind to the recording data rated the time of onset of cue-evoked change in behavioral activity, typically indicated by termination of grooming behavior; 3) the latency between food delivery and food retrieval; 4) general food-port approach behavior, independent of trial-related activity, was determined off-line by dividing the floor into 9 squares and counting the number of entries into the two squares underneath the ports throughout the session.
Ceramic-based, multi-site microelectrodes featuring four 15×333 μm Platinum- recording sites arranged in side-by-side pairs (Quanteon LLC, Nicholasville, KY; see Figure S1A) were prepared for enzyme coatings and calibrated in vitro. These methods, as well as surgical methods and procedures used for in vivo recording of cholinergic activity are described in detail in Supplemental Methods.
After completion of recording sessions, choline was infused through the guide cannula to determine the sensitivity of the microelectrode to choline. Additionally, and in order to confirm that the responses of the implanted microelectrode reflects choline resulting from the hydrolysis of endogenously generated ACh, the effect of neostigmine, an acetylcholinesterase (AChE) inhibitor, on potassium-evoked choline signals was determined (see Supplemental Materials for Methods and Results).
Methods used for self-referencing of choline signal recordings, the analysis of event-evoked cholinergic signals and session-related tonic changes in cholinergic activity, methods used for the microdialysis experiments and to compare session-related tonic changes in cholinergic activity with microdialysis release data, and the number of animals per group, are described in Supplemental Materials.
In order to confirm that changes in choline levels recorded in the mPFC of task-performing animals reflect choline resulting from hydrolysis of newly released ACh from cholinergic terminals, electrodes were implanted in the mPFC following cholinergic deafferentation of the recording area by infusing the immunotoxin 192 IgG-saporin (192-SAP; ATS, San Diego, CA). Animals (n=5) received unilateral infusions of 192-SAP (100 ng/0.5 μL) into the right mPFC using the following coordinates (AP: +3.2 mm, ML: −0.7 mm; DV: −3.5 mm). Infusions were made at a rate of 0.25 μL/min using a 1 μL Hamilton microsyringe; the needle remained in place for an additional 4 min following the infusion. Animals were returned to daily test sessions and microelectrodes were implanted three weeks later. Importantly, such unilateral, restricted deafferentation of the recording region does not affect the rats’ performance of attention-demanding tasks (Gill et al., 2000).
To determine whether cholinergic innervation of the mPFC is necessary for the performance of the cued appetitive response task, the hypothesis that bilateral removal of cholinergic inputs into the mPFC reduces cue detection rate was tested in a separate group of animals (n=5). These animals were trained to task criterion. In order to remove cholinergic inputs to the mPFC (including pre- and infralimbic region and anterior cingulate cortex), 192-SAP (100 ng/0.5 μL) was infused bilaterally at two sites per hemisphere (AP: +3.7/2.6; ML: ± 0.7 mm; DV: −3.5 mm). Following two days of post-surgery recovery with food and water ad libitum, the animals were returned to the deprivation regimen and daily test sessions. Animals were tested for three more weeks. Sessions were videotaped once a week for analysis (see Supplemental Materials for histological methods).
Statistical analyses were performed using SPSS/PC+ (V13.0; SPSS, Chicago, IL). Repeated measure mixed factor ANOVAs were used to analyze the effects of group (intact and unilateral lesion, two levels; pre-lesion (bilateral) and post-lesion; four levels), task (standard and shorter cue-reward interval; two levels) and trial blocks (five levels) on behavioral performance. Post-hoc multiple comparisons for analysis of significant main effects were performed using Least Significance Difference (LSD) test or independent t-tests. One way ANOVAs or planned multiple two-tailed unpaired t-tests were employed to test group differences with respect to the proportion of detected cues, reward retrieval latencies, and port approach frequencies. The effect of trial blocks on the proportion of cues that were detected was examined using one-way ANOVA (for more details see Supplemental Materials).
This research was funded by PHS Grants 2RO3 MH073600, 5RO1 NS37026, and 1KO2 MH01072 to M.S. We thank Drs. Greg Gerhardt, Francois Pomerleau and Peter Huettl (University of Kentucky) for their continued technical support. We also thank Drs. Terry Robinson, Stephen Maren and Josh Berke, and W. Matt Howe (University of Michigan) for reviewing drafts of this manuscript.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.