Developmental role of acetylcholinesterase in impulse control in zebrafish

Cellular and molecular processes that mediate individual variability in impulsivity, a key behavioral component of many neuropsychiatric disorders, are poorly understood. Zebrafish heterozygous for a nonsense mutation in ache (achesb55/+) showed lower levels of impulsivity in a 5-choice serial reaction time task (5-CSRTT) than wild type and ache+∕+. Assessment of expression of cholinergic (nAChR), serotonergic (5-HT), and dopamine (DR) receptor mRNA in both adult and larval (9 dpf) achesb55/+ revealed significant downregulation of chrna2, chrna5, and drd2 mRNA in achesb55/+ larvae, but no differences in adults. Acute exposure to cholinergic agonist/antagonists had no effect on impulsivity, supporting the hypothesis that behavioral effects observed in adults were due to lasting impact of developmental alterations in cholinergic and dopaminergic signaling. This shows the cross-species role of cholinergic signaling during brain development in impulsivity, and suggests zebrafish may be a useful model for the role of cholinergic pathways as a target for therapeutic advances in addiction medicine.


Introduction
The identification of endophenotypes, as quantifiable, core components of complex behavioral traits and disease phenotypes makes genetic analysis of the pathogenesis of neuropsychiatric disease more tractable in both humans and model organisms (Burmeister et al., 2008). One such potential endophenotype is impulsivity (Urcelay and Dalley, 2012). Impulsivity not only is the hallmark symptom of a number of neuropsychiatric disorders (ADHD, addiction) but, in the case of addiction, has been shown to predict patterns of relapse and compulsive drug seeking in rats (Belin et al., 2008).
Despite the well-established role in a number of neuropsychiatric disorders, the cellular, and molecular mechanisms that underlie impulsivity are not well-understood. The cholinergic system, in particular cholinergic projections from the PFC, has long been implicated in sustained attention (Sarter et al., 2001). For example, IgG-saporin lesions of cholinergic neurons in the basal forebrain reduce sustained attention (Mcgaughy and Sarter, 1998), while systemic administration of the nAChR agonist nicotine improves performance accuracy and reduces omissions on the 5-CSRTT (Blondel et al., 2000;Hahn and Stolerman, 2002;Young et al., 2004). In addition, infusions of scopolamine (mAchR antagonist) into the medial pre-frontal cortex (mPFC), and systemic mecamylamine (nAchR antagonist) reduce response accuracy (Robbins et al., 1998). The effects of chronic elevation of ACh, however, are less clear, although Grottick and Higgins (2000) found that improved performance accuracy is apparent with chronic nicotine exposure. The effects of genetic alteration of ACh activity have not previously been tested, particularly with respect to premature responding on the 5-CSRTT.
Notwithstanding their small size, low housing costs and prolific breeding, there now exists a number of genetic tools for zebrafish research, including N-ethyl-N-nitrosourea (ENU) mutagenized lines, extensive sperm libraries, and a number of GFP/RFP lines. Despite anatomical differences between the fish and their mammalian counterparts, key neurochemical pathways are well-conserved between the species (Guo, 2004); for example, the ascending and descending midbrain catecholeminergic pathways (Guo et al., 1999).
Here, we tested the performance of ache (the gene that codes for acetylcholinesterase; AChE) deficient (ache sb55/+ ) zebrafish, for performance characteristics on the 5-CSRTT, a task designed to test aspects of impulse control through examination of anticipatory responding. ache sb55 contain a point mutation close to the catalytic site of the enzyme resulting in a replacement of serine 226 by an asparagine. Serine 226 is conserved in all ache gene family members, and is important for catalytic activity (Behra et al., 2002). Chronic alterations in cholinergic signaling with the AChE inhibitor chlorpyrifos has previously been demonstrated to increase impulsivity, make cholinergic signaling an interesting target for inquiry into the molecular mechanisms underlying impulse control (Middlemore-Risher et al., 2010;Cardona et al., 2011;Oca et al., 2012). Zebrafish have previously been shown to respond well on the 5-CSRTT (Parker et al., 2012(Parker et al., , 2013(Parker et al., , 2014.

Ethics Statement
All experimental procedures, including drug dosing, and behavioral testing, were carried out under the Animals (Scientific procedures) Act (1984). The procedures carried out conformed both to local ethical guidelines and to the terms of a project license from the UK Home Office. In addition, all experiments were approved by the Queen Mary Animals Welfare and Ethical Review Board.

Subjects
Twenty-nine (n = 10 ache sb55/+ (Ninkovic et al., 2006), n = 19 Tubingen wild-type), adult zebrafish (age = 6 months; mixed sex) were selected for the first part of the study (5-s fixed interval PSI), and 12 adult zebrafish (age = 5 months; mixed sex; n = 5 ache sb55/+ ; n = 7 ache +/+ ) were selected for the second part (Variable PSI). All were sourced initially from the Sanger Institute (Cambridge, UK), and bred and reared in the aquarium facility at Queen Mary University of London according to standard protocols (Westerfield, 1993). During the entire experimental period, fish were fed artemia/bloodworm mix during testing trials, and this was supplemented with flake food/artemia in the evenings and at weekends. Figure 1 displays the 5-CSRTT tanks used in the study. The shell of the testing tanks was constructed from opaque acrylic, as were the central gates. The lights were LEDs (magazine light green, stimulus aperture lights yellow). The reinforcer used was artemia liquidized with bloodworm, suspended in aquarium-treated water (R-O water with added salts). The food was delivered via a plastic syringe fitted with a 1 mm diameter rubber catheter tube, which was driven by a linear stepper motor (Figure 1).

General Procedure
The main procedure is an extension and modification of the commonly used rodent 5-CSRTT, and has been described in detail elsewhere (Parker et al., 2012(Parker et al., , 2013(Parker et al., , 2014.

Pre-training
Prior to commencing training, all subjects were habituated to the test room for 1 week to acclimate to the conditions. All pretraining, training and testing was carried out Monday-Friday (0800-1800), with the exception of the final stage (Stage 8, see Table 1), which was also carried out Saturday and Sunday. Training was divided into eight distinct stages (see Table 1).
During stages 1-3 (pre-training) data were collected and examined to ensure that all animals were receiving food during training. Any that did not perform the task (e.g., froze in the tank or did not approach the lights; n < 2 on any given session) had their food supplemented immediately after the session. During acclimation (Stage 1), fish were placed individually into the test tanks for 30-mins. During this all lights were illuminated and the gate was open. Immediately after acclimation, the fish were trained to enter the food magazine (Stage 2). During this stage, the gate remained closed at all times. The magazine light was illuminated for 30-s intervals, during which entry to the magazine resulted in the light turning off, and a small delivery (∼20 µl) of artemia/bloodworm mix. In Stage 3 the fish were trained to approach the response apertures. Here, the gate opened to reveal all of the response apertures illuminated, and entry to any one of the apertures was conditionally reinforced with illumination of the magazine light. Subsequent entry to the food magazine was reinforced with artemia/bloodworm mix. During Stage 3 (response aperture orientation) only fish that completed 20 or more correct trials were taken forward to 5-CSRTT training.

Five-choice Serial Reaction Time Task: Phase 1
After a 2-min habituation period, the magazine light was illuminated, and entry to the food magazine initiated the trial sequence after an inter-trial interval (ITI) of 20-s 1 . This ITI always followed food delivery, and allowed the fish time to consume the reinforcer ration. After 20-s, the gate was raised, and one of the stimulus apertures was illuminated after a prestimulus interval (PSI). Entry to the correct aperture during the  stimulus illumination, or during a brief pause thereafter (limited hold; LH), were conditionally reinforced by illumination of the magazine light, and the trial ended when the fish collected the food. All training sessions lasted 30-mins. For the first 4 weeks (Stage 4) the fish were trained with 30-s stimulus duration, a PSI of 1-s and a 1-s limited hold period. At all times during training and testing, the magazine light remained illuminated for 30-s following a correct response, after which magazine entry was not reinforced. During the second stage of 5-CSRTT training (Stage 5) the stimulus duration was reduced to 10-s, the PSI was increased to 5-s and limited hold remained at 1-s. The criterion for moving from each stage to the next was that the fish had reached a steady-state response, operationalized as completing >20 trials per session over 5-consecutive sessions. Any fish not meeting this criterion were excluded from the subsequent stage.

Long PSI Stage
There were three long PSI sessions, during which the PSI was increased to 7-s. All other test parameters remained the same as during Stage 5 (stimulus duration = 10-s, limited hold = 1-s). The three long PSI sessions were interspersed by two baseline sessions (Stage 5; PSI = 5-s, stimulus duration = 10s, limited hold = 1-s). During the long PSI sessions, the length of the session was increased to 35 min. The criterion for a fish progressing to the long PSI phase of the experiment was that they reached steady state responding, again, operationally defined as having completed five sessions of >20 trials prior to testing. Any fish that did not meet this criterion were excluded from the testing phase.

Five-choice Serial Reaction Time Task: Phase 2
For the second phase of the experiment, we trained a group of experimentally naïve fish (n = 5 ache sb55/+ ; n = 7 ache +/+ ) in an identical manner to that described above for stages 1-4. For Stage 5, we introduced 5-s variable interval (VI) PSI. All other timings were the same as in Phase 1, Stage 5 (stimulus duration = 10-s, limited hold = 1-s). There was no Long-PSI stage in Phase 2.

Acute Exposure to AChe Antagonist, and nAChR and mAChR Agonists
Trained fish (wild-type from Phase 1) were selected for the drug administration phase. The exposure schedule was organized according to a full crossover design, with each fish receiving each of the drugs over a 1-week period. Fish were initially retrained (2-weeks) in the absence of drug to establish steadystate baseline performance (>20 reinforced trials/session, for 5 sessions). The 5-CSRTT was as before in Stage 5 (see above: stimulus-duration = 10-s, PSI = 5-s, LH = 1-s), except that in this phase we employed a variable interval (VI) 5s PSI. During the first experiment, there was no difference between the strains during the long PSI trials, but there was a difference during the earlier stages of training. Therefore, we chose to increase the complexity of the task by using a VI-PSI during the entire training period. Immediately prior to training, fish were immersed in a pre-treatment tank ( (Day et al., 2007;Brembs, 2009;Cardona et al., 2011). The dose of donepezil was also based on an initial assessment of brain levels of ACh and AChE following drug administration to determine a dose that best reflected the ACh and AChE levels in ache sb55/+ (Ninkovic et al., 2006). Brain activity of AChE and brain levels of ACh were assessed in wild-type fish exposed to 2.63 µM donepezil or aquariumtreated H 2 O for 20 mins using a fluorescence-based approach (George et al., 1961). Following exposure to drug fish were placed in a recovery tank for 5-min, and then killed by immersion in ice water. Brains were immediately removed, weighed and homogenized in ice-cold Tris-HCl (pH 8). Samples were then centrifuged (20-min at 13,000 rpm) and AChE and ACh was assessed from the resulting supernatant using Amplex Red Acetylcholine/Acetylcholinesterase assay kit (Molecular Probes, Invitrogen Detection Technologies, Paisley, UK) according to manufacturer's instructions. Briefly, AChE converts ACh into choline, which is then oxidized by choline oxidase to betaine and H 2 O 2 . Brain activity of AChE and brain levels of ACh were measured using 10-acetyl-3, 7-dihydroxyphenoxazine, a flourogenic probe for H 2 O 2 . All ACh and AChE samples were examined in duplicate against standards and fluorescence was measured on a fluorescence microplate reader (FLUOstar OPTIMA, BMG LABTECH, Cary, NC). Following exposure to 2.63 µM donepezil, the levels of ACh were found to be higher in the drug group [11.8 nM/g vs. 7.1 nM/g; t (8) = 2.81, P = 0.02], which was directly comparable to levels seen in the ache sb55/+ thus validating the dose used (Ninkovic et al., 2006).
The exposure schedule was as follows: Week 1: drug A, Week 2: recovery (no drug), Week 3: drug B, Week 4: recovery, Week 5: drug C. As stated, each fish was tested in the presence of each of the three drugs, the order of which was counterbalanced across weeks.

Gene Expression Changes in ache sb55/+
We collected embryos from 4 × ache sb55/+ in-crosses. All homozygous individuals were removed at 72 hpf (easily identifiable by morphological features and lack of motor activity) leaving petri dishes with ∼2/3 heterozygous individuals. We also collected embryos from 4 × ache +/+ in-crosses for comparison. Reference genes used were β-actin, ef1α, and rpl13α based on previous findings findings (Tang et al., 2007). Target genes used are listed in Table 2. All embryos were manually sorted to ensure all were at the same developmental stage over the first 72 hpf, and grown to 9 dpf in petri dishes (∼40/dish) in an incubator (28 • C). At 9 dpf embryos were terminally anesthetized in MS-222, and placed in RNAlater until assay (4 • C). Eight batches of n = 3 embryos per strain (ache sb55/+ and ache +/+ ) were lysed in 200 µl Lysis buffer with 2 µl Proteinase K for 30-45 min (55 • C). mRNA was isolated using 40 µl Dynabeads R Oligo(dT) 25 according to manufacturer's instructions. Ten adult (6 months) brains (n = 5 ache sb55/+ ; n = 5 ache +/+ ) were homogenized in 400 µl Lysis buffer with 4 µl Proteinase K for 30-min (55 • C). mRNA was isolated using 80 µl Dynabeads R Oligo(dT) 25 according to manufacturer's instructions. All qPCR reactions were carried out in triplicate. 1 µl of cDNA and 1.5 µl each of forward and reverse primers (see Table 2) were added to 5 µl SYBR R Green PCR Master mix and run in a 384-well plate format (Roche Diagnostics). Method reported in full elsewhere (Gemenetzidis et al., 2010) (Teh et al., 2013).

Data Analysis
5-CSRTT data were fitted to general linear models (fit by REML), with time (5-CSRTT phases 1-5) and strain (either ache sb55/+ vs. ache +/+ or ache sb55/+ vs. wild-type) as fixed effects. In the drug administration phase, drug (four-levels, nicotine, pilocarpine, donepezil, and control) was added as a fixed factor, with ID and day as random effects. In each case,  Post-hoc Tukey tests were carried out to examine main effects and interactions of 5-CSRTT data.
Finally, to test the difference between levels of mRNA expression in larvae and adult ache sb55/+ and ache +/+ siblings, we carried out a series of Mann-Whitney U-tests, with strain (ache sb55/+ vs. ache +/+ ) as the independent variable and target gene expression, relative to reference genes, as the dependent variables. For mRNA expression data, P values were estimated following Bonferroni correction for multiple comparison. Effect sizes for all differences in expression were also calculated using the Grissom and Kim (2012) method. Descriptive statistics are reported as mean ± SEM unless otherwise stated. A type-1 error rate of α = 0.05 was adopted for all statistical tests. All data were analyzed using IBM SPSS Statistics v.21 for Macintosh.

Discussion
The aim of this experiment was to test the hypothesis that developmental alterations in cholinergic signaling affect impulse control using a zebrafish model of the commonly used 5-CSRTT with a strain heterozygous for a missense mutation in ache (ache sb55/+ ). We found that ache sb55/+ showed a lower proportion of premature responding than ache sb55/+ siblings and wild-type zebrafish. There were no significant differences in either the number of correct responses, latency to respond, number of trials or the number of omissions, although the ache sb55/+ appeared to learn faster, collecting more reinforcers during pre-training. Acute reductions of AChE (donepezil) had no significant effects on premature responding, or other 5-CSRTT parameters, and acute administration of a nAChR agonist significantly increased performance accuracy, while having no effect on premature responding. Finally, ache sb55/+ have a down regulation of chrna2, chrna5, and drd2 mRNA expression at 9 dpf, but no difference in expression in any of the genes we examined in adulthood. Previous studies have shown that high levels of AChE inhibition during development (e.g., with the organophosphate weedkiller chlorpyrifous Middlemore-Risher et al., 2010;Cardona et al., 2011;Oca et al., 2012) increase impulsivity in later life. Collectively, these data suggest the intriguing theory that variation in AChE during development may follow a J-curve with respect to its effects on impulse control, potentially through downstream effects on cholinergic and dopaminergic pathways.
Lesion, neuropsychological, and pharmacological studies have demonstrated that cortical cholinergic projections to mid-brain regions are strongly implicated in sustained attention and in general top-down cognitive control (Sarter et al., 2001). In particular, during 5-CSRTT performance rats display elevated ACh release from the medial pre-frontal cortex (mPFC), and phasic increases in ACh release when a visual distracter was introduced to increase task complexity (Passetti et al., 2000). We did not see any differences in the number of correct responses in our version of the task, but more of the ache sb55/+ met criteria to move to the 5-CSRTT stage of training, and of those that met criteria, overall performance in terms of reinforcers gained was significantly greater than the wildtype. This finding replicates assessment of this strain's learning previously demonstrated in a T-maze task (Ninkovic et al., 2006). During this initial training stage, despite the strain difference, there was no day × strain interaction, suggesting that ache sb55/+ learnt at the same rate. It may be that the ache sb55/+ were more motivated to perform, or habituated faster than the wild-type. This effect was transient, however, disappearing once training started on the 5-CSRTT. We did, however, find evidence for the role of nAChR in task performance, with acute exposure to nicotine (nAChR agonist) increasing the proportion of correct responses in the task. This supports previous data from rodents (Blondel et al., 2000;Hahn and Stolerman, 2002;Young et al., 2004).
A potential mechanism for the observed differences in premature responding may relate to the role of nAChR during early brain development and patterning. nAChR subtypes, in particular α4, α5, α7, β2, and β4, are found early in brain development, and have been suggested to play a role in modulating and mediating early patterning, dendritic outgrowth and synaptogenesis (Hellström-Lindahl et al., 1998). It is possible therefore that reduction in AChE levels, as is characteristic of the ache sb55/+ heterozygotes, during early brain development alter the distribution of nAChRs thus causing differences in patterning and dendritic morphology. Indeed, in zebrafish, AChE enzymatic activity has been shown to be important for both axon outgrowth and synapse stability, albeit within the neuromuscular projections of the nervous system (Behra et al., 2002;Downes and Granato, 2004).
Chronic reductions of AChE in adult rats with donepezil increases expression of α4 and α7 nAChR (Kume et al., 2005), and ACh-modulated reductions in impulsive action in the 3-CSRTT are mediated by α4 nAChR (Tsutsui-Kimura et al., 2010). Although we did not observe differences either in chrna4 or chrna7 here, we did observe robust down regulation of chrna2 and chrna5 mRNA expression in the ache sb55/ + heterozygotes at 9 dpf, but no differences in adulthood. CHRNA2 and CHRNA5 variants have been shown to predict impulsive responding in response-inhibition in humans (Rigbi et al., 2008), and transgenic mice overexpressing the Chrna3, Chrna5, Chrnb4 gene cluster show a reduction in impulsivity (Viñals et al., 2012). However, the differences in behavior observed in the ache sb55/+ heterozygotes demonstrate haploinsufficiency of the AChE gene, and thus has implications for the impact of AChE mutations within the human population. Although we are yet to understand the mechanism, this may inform our exploration of potential targets for therapeutics in the future.
The functional properties of nAChRs on catecholaminergic (in particular, dopaminergic) axonal terminals alter during development, highlighting their role in the development of the dopamine system (Azam et al., 2007). It is clear that over-activation of nAChR during early development, e.g., from maternal smoking during pregnancy, can result in an increased risk for impulse control disorders (Button et al., 2007). In addition, as discussed above, excessive inhibition of AChE during development, resulting from exposure to the organophosphate insecticide chlorpyrifos, results in higher impulsivity (Middlemore-Risher et al., 2010;Cardona et al., 2011;Oca et al., 2012). Although this shows a clear link between developmental effects of cholinergic-system disruption and impulsivity, it is not clear at this stage the mechanisms by which subtle alterations, such as are seen with ache sb55/+ , subsequently reduces impulsivity. It is possible that this reflects species-specific differences in patterning during early brain ontogeny, although this seems unlikely based on documented similarities between fish and mammalian cholinergic system development (Xie et al., 2000;Behra et al., 2002). During development, AChE is transiently involved with aspects of neural patterning and hodological development. For example, during cortical synaptogenesis and development of thalamo-cortical pathways, AChE activity is recorded in various brain regions (Button et al., 2007). The cholinergic system interacts with mid-brain dopamine activity in a number of ways.
First, the nucleus accumbens (NAc) is densely innervated by cholinergic projection neurons (Meredith et al., 1989;Woolf, 1991). Second, cholinergic receptors [both muscarinic (mAChR) and nicotinic (nAChR)] are found on ventral tegmental area (VTA) dopamine neurons, suggesting dopaminergic control of cholinergic activity (Clarke and Pert, 1985). Third, mesolimbic cholinergic projection neurons are abundant with dopamine receptors, suggesting cholinergic mediation of dopamine activity (Gronier et al., 2000), creating a feedback loop. Rats characterized as high trait impulsivity based on baseline performance on the 5-CSRTT show a greater tendency for elevated cocaine self-administration (Dalley et al., 2007), increased compulsive cocaine seeking (Belin et al., 2008) and increased relapse to compulsive cocaine seeking following punishment-induced abstinence (Economidou et al., 2009). In addition, high impulsive rats show a reduction in DRD2/DRD3 in the ventral striatum, suggesting a potential biomarker for the addiction phenotype (Dalley et al., 2007). Interestingly, ache sb55/+ have previously been characterized as showing a decrease in conditioned place preference (CPP) for amphetamine (Ninkovic et al., 2006). It is well-established, through the therapeutic efficacy of dopamine agonists such as methylphenidate in reducing impulsivity in ADHD patients (Barkley, 1997), that impulsivity is, at least in part, related to a reduction in availability of dopamine (Li et al., 2006). It is possible that genetic impairment of AChE in ache sb55/+ -which results in higher levels of circulating ACh and desensitization of AChRs (Ninkovic et al., 2006) -may act to stabilize dopamine activity (Zhou et al., 2001), thus decreasing impulsive responding. However, although we observed downregulation in drd2 mRNA in 9 dpf ache sb55/+ embryos, there was no significant differences in the adults. This requires further exploration in order to elucidate the mechanism.
In rodents, low levels of premature responding in the 5-CSRTT are predictive of animals that show resistance to developing compulsive drug seeking (Belin et al., 2008) and relapse following abstinence (Economidou et al., 2009), and this has been interpreted as these animals showing low levels of trait impulsivity affecting top-down cognitive control (Dalley et al., 2011). The neural circuits of impulsivity are currently not wellunderstood (Brown et al., 2006;Chang et al., 2012), but these findings suggest that zebrafish, an established genetic model system, offer a means for exploration of this.
Gaining a better understanding of the etiology of psychiatric disease is currently a priority area of research (Campbell, 2010), and with current advances in neuroimaging and huge increases in genetic sequencing power this aim is beginning to be realized. For example, genome-wide association studies (GWAS) are making progress in this regard (Sullivan, 2010), but are limited by uncontrollable factors such as environmental influences and heterogeneity of diseases (Burmeister et al., 2008). Animal models have proved useful in terms of identifying molecular mechanisms of many psychiatric diseases, as symptoms consistent with DSM-IV (APA, 2000) diagnoses of psychiatric disorder have been characterized in many models (Gould and Gottesman, 2006). A better understanding of the molecular mechanisms will be helpful in tailoring treatment options for patients, but also for early identification of at-risk individuals to allow preventative measures to be adopted in the early stages of the disorder (Uhl et al., 2008). Progress in identifying molecular mechanisms, however, has remained slow. This study shows more evidence that zebrafish may be very useful in expediting this process.
In conclusion, this study has found that alterations in Ache reduce premature responding in zebrafish on the 5-CSRTT. This effect appears to relate specifically to developmental effects of reduced AChE, as acute exposure to an AChE antagonist had no effect on premature responding in the task. Molecular analyses suggest that the route of action may be through cholinergic interactions with midbrain dopamine systems during development. This study opens the door for potential largescale forward genetic population screening of mutagenized lines of zebrafish to identify novel alleles for phenotypes such as impulsivity, which is crucial in the search for novel therapeutics and individualized medicine (Jain et al., 2011).