In the present studies we set out to test this third hypothesis. If it is the case that cognitive symptoms of PD are related to the disease process then animal models that mimic specific aspects of the pathophysiology of PD should also show impairments in cognitive function. Despite this prediction, cognitive aspects of PD have been difficult to study pre-clinically since animal models generated using toxin-based methods often have severe motor impairments which prevent the normal mobility required to perform typical cognitive tests in animals (Bove, Prou et al. 2005). Thus the three above hypotheses have not been tested in animal models. It would be of great interest to have a model of cognitive aspects of PD to test potential therapeutics and investigate the neurobiology of cognitive symptoms of PD. Ideally, such a model would recapitulate important aspects of PD, but subjects would not be severely motor impaired so as to enable them to perform common behavioral tasks.

Over the last decade several genes have been identified that may play a role in dopamine neuron development as well as the pathophysiology of PD including Nurr1, Parkin, and Pitx3. Recently, we and others have shown that aphakia (ak) mice, which are deficient in the bicoid-related, homeodomain-containing transcription factor, Pitx3, fulfill major criteria for an animal model of PD. Pitx3 is restrictively expressed in the lens and midbrain and has been shown to play a key role in dopamine neuron development (Smidt, Smits et al. 2004). Consistent with an important role of Pitx3 in dopamine neuron development, Pitx3- deficient ak mice are the first genetic model to recapitulate selective nigrostriatal (A9) dopamine neuron loss (Hwang, Ardayfio et al. 2003; Nunes, Tovmasian et al. 2003; van den Munckhof, Luk et al. 2003; Smidt, Smits et al. 2004). Greater than 90% loss of A9 dopamine neurons occurs in ak mice, whereas A10 and other areas are less affected. Further validitation of the PD-like phenotype in this genetic model is that ak mice show impaired performance on select motor tests, which is reversible with L-Dopa (Hwang, Fleming et al. 2005). Table 1 summarizes the phenotypic findings reported in ak mice and their similarity to PD. Since Pitx3 is involved in lens development, ak mice are also blind. The key feature of ak mice however, is that despite select PD-like motor impairments, ak mice can ambulate well enough to perform many behavioral tasks. This makes ak mice a unique A9 dopamine deficiency model to study neuropsychiatric and non-motor aspects of PD. We report here that Pitx3 deficiency results in impairment in striatum-dependent procedural and associative learning tasks but not a non-striatal dependent associative learning task. These findings suggest that PD related pathophysiology rather than stress or medications contribute to the cognitive impairments observed in patients with PD.

The swimming t-maze is another procedural memory task; however unlike the rotarod it requires egocentric navigation for mice to find an escape platform. This task can be performed by mice with retinal degeneration and is sensitive to striatal damage (Van Raamsdonk, Pearson et al. 2005). As shown in Figure 2, Top Panel, wt and rd1 mice over the 5 days became increasingly faster at finding the platform. In stark contrast to the decrease in latency in the wt and rd1 mice, ak mice actually increased their latency to reach the platform over the 5 days. Analysis of the learning curve showed a main effect of day F_(4,76)=3.10, p=0.02 and genotype F_(2,76)=8.88, p=0.0019 and a significant interaction between day and genotype F_(8,76)=6.19, p<0.0001 in the latency to reach the platform. Since the ak mice swam with similar speed to wt and rd1 on day 1, the genotype effect is likely related more to impairment of procedural learning in ak mice, rather than impaired motor function per se. If ak mice were unable to swim, day 1 latencies would likely have been higher than wt or rd1 mice. Consistent with this idea there was a main effect of genotype on mean number of errors (Figure 2, Bottom Panel) committed F_(2,76)=8.28, p=0.0026 and day F_(4.76)=4.46, p=0.00026. Interestingly, by the third day of testing most wt and rd1 mice learned the swim task well enough to run most trials without any errors. ak mice in contrast continued to make errors. Our criterion for successfully learning the task was 3 consecutive days (9 trials) without an error. The number of mice that achieved this criterion was lower (p=0.0319) in ak mice (1 out of 7) than in the wt (7 out 9) or rd1 (4 out of 6) groups.

Due to the strenuous nature of the swimming T-maze we examined the behavior of ak mice on a dry, food motivated T-maze in order to rule out the effects of stress and fatigue. As shown in Figure 3 wt and rd1 mice showed robust decrease in latency between day 3 and 4, which is indicative of procedural learning and memory. The apparently elevated latency in wt animals compared to rd1 and ak mice is likely due to the observation that the sighted wt mice were engaging in more visual exploration of the apparatus. The improvement between day 3 and 4 was not observed in ak mice. Rather the curve for the ak mouse was nearly flat, similar to their performance in the rotarod. (F_(10,560) = 2.33, p = 0.01). In comparing the change in latency between day 1 and day 5 ak mice had a significantly less difference score than wt or rd1 mice, indicative of impaired procedural memory (F_(2,28.377) = 8.3, p = 0.003). In contrast to the latency, errors in the dry T-maze however did not significantly differ. Together the data from the rotarod swim and dry T-maze suggests procedural memory impairments in ak mice.

Inhibitory avoidance learning is an operant associative learning task that blind mice can readily perform (Dyer, Hammond et al. 1975; Farr, Banks et al. 2002). In addition inhibitory avoidance is altered by lesions of the caudate nucleus as well as manipulations of the prefrontal cortex, hippocampus and amygdala. These brain regions may be involved in cognitive dysfunction in PD. In the acquisition trial of the inhibitory avoidance task there was no difference between wt, rd1 and ak mice in the latency step into the shock paired side (Figure 4, Top Panel). The day 1 latencies can be used as a rough measure of baseline locomotor activity, since long or short latencies can indicate motor impairment or hyperactivity respectively. Thus the lack of difference across the groups on day 1 latencies is important because it indicates that ak mice show no evidence of gross locomotor activity impairment or hyperactivity on this task. As evidence of learning on day two (Figure 4, Middle Panel), 24 hrs after receiving the shock, wt and rd1 mice showed increased latencies, often more than 10 fold greater than the latencies on day 1. In contrast the latencies of ak mice increased only slightly if at all on day 2, indicating impaired memory for the shock presented 24hrs earlier. When comparing all three groups on the change in inhibitory avoidance, which is the change in latency due to learning and which factors in baseline latency, ak mice were significantly different from wt (p<0.05) and rd1 (p<0.001, One-Way ANOVA, Tukey’s Post-hoc) mice (Figure 4, Bottom Panel). Wild-type and rd1 mice did not differ on any measures. To rule out differences in shock sensitivity we measured vocalizations (>2s) and found no statistical differences in vocalizations (wt = 2.5, rd1 = 1.3, ak = 2.2).

Given the well known detrimental effects of stress on mood and cognition, one potential cause of cognitive impairments and neuropsychiatric symptoms in PD is the psychological burden or stress of living with a chronic medical illness (Duff, Mold et al. 2007). Under this premise, PD-related cognitive symptoms are not specific to PD but would be expected to occur in any disabling chronic medical illness. The second potential cause of cognitive symptoms in PD is the medication used to treat PD (Papapetropoulos, Argyriou et al. 2005; Cools 2006) . Dopamine (DA) pharmacotherapy could potentially impair cognition given the important role of DA in executive function, and the harmful effects of excess DA on cognition (Papapetropoulos, Argyriou et al. 2005). Anti-cholinergics are another class of PD treatment that could feasibly cause the cognitive impairment observed in PD. The last hypothesis regarding the etiology of cognitive symptoms in PD is that it could be a result of specific pathophysiological events intrinsic to the disease, such as loss of nigrostriatal dopamine or other neurotransmitters altered in PD (Zgaljardic, Foldi et al. 2004). Indeed, PD involves disruption of several key signaling molecules that are important in learning and memory including DA, acetylcholine, norepinephrine and glutamate (Guttman 1987; Di Chiara and Morelli 1993). It cannot be ruled out that behavioral differences observed in ak mice are related to disruption of neurotransmitter systems other than DA, however given the preponderance of evidence suggesting DA is important in striatum-dependent cognition and the strength of the DA related phenotype in ak mice, the DA system seems a likely candidate.

Secondly this provides to our knowledge the first genetic model of cognitive impairment in PD. The major advantages of this model are the capacity to probe cognition and other non-motor aspects of DA function in severely DA-deficient (>90% DA loss) subjects. Such testing is difficult to conduct in other PD DA-deficiency models since similar bilateral lesions of DA using toxins such as 6-OHDA results in severely motor impaired, unmotivated and unhealthy animals. Furthermore, though the myriad DA receptor knockout mice have provided invaluable information on the function of DA receptors, there are few pathway specific models that can be used to dissociate A9 vs. A10 related contributions. As is the case with all experimental animal models, ak mice cannot be said to model all aspects of PD. In fact, ak mice do not appear to model the progressive neurodegeneration . They do however recapitulate severe nigrostriatal DA loss (Hwang, Ardayfio et al. 2003), which is a hallmark feature of PD and likely responsible for much of the behavioral symptoms in PD. Additionally, we have previously reported that ak mice mimic other key features of PD such as L-Dopa responsivity, nigrostriatal motor defect and dopamine supersensitivity (Hwang, Fleming et al. 2005). Our recent analysis also showed that chronic treatment with L-DOPA and dopamine receptor agonists induces novel dyskinetic behaviors in ak mice, which was attenuated by anti-dyskinetic agents (Ding, Restrepo et al. 2007). Altogether the ak mice can be said to model an impressive number of features of PD.

We favor the view that the differences reported here are cognition related, however, another potential interpretation of these data is that the impairments that we observe in ak mice are artifacts of genotype differences in locomotor activity or sensory capacity. There is little support for this view however, since ak mice showed no evidence of global neurological deficits or severe motor disturbance in any of the paradigms tested here, which is consistent with our previous studies. ak mice did not differ from wt or rd1 mice on day 1 baseline performance of any of the tasks tested. Instead the main difference observed between ak mice and the wt and rd1 mice is that ak mice failed to demonstrate experience-dependent change in behavior. This is consistent with the idea of select striatum-dependent learning and memory impairment in ak mice.

Our finding of cognitive impairment in ak mice adds to the growing literature implicating an important role for the basal ganglia in cognition and neuropsychiatric symptoms. Deficits of procedural and associative learning have similarly been observed with other mouse models of striatal dysfunction including the YAC128 and R6/2 mouse models of Huntington’s disease (Lione, Carter et al. 1999; Van Raamsdonk, Pearson et al. 2005). For example YAC128 which have reduced dopamine in the striatum are impaired in the swimming t-maze (Van Raamsdonk, Pearson et al. 2005). It is interesting that in contrast to Parkinson’s disease, the cognitive impairment and neuropsychiatric features of Huntington’s disease are well recognized and an active area of both clinical and pre-clinical research. Furthermore, a recent elegant study showed that the heterozygous mice of Nurr1, another key transcription factor for dopamine neuron development (Wallen and Perlmann 2003; Smidt and Burbach 2007), display a behavioral pattern indicative of potential relevance for symptoms of schizophrenia (Rojas, Joodmardi et al. 2007).