Probabilistic learning and reversal deficits in patients with Parkinson’s disease or frontal or temporal lobe lesions: possible adverse effects of dopaminergic medication

Abstract

Three groups of patients with Parkinson’s disease (PD) — mild, unmedicated (UPD), mild, medicated (MPD) and severe, medicated (SPD) — and patients with lesions of the frontal lobe (FLL) or temporal lobe (TLL) were compared with matched controls on the learning and reversal of probabilistic and two-pair concurrent colour discriminations. Both of the cortical lesion groups showed reversal deficits, with no increase in perseverative responding. The UPD group, although impaired on a spatial recognition task, showed intact discrimination learning and reversal; the MPD and SPD patients showed non-perseverative reversal impairments on both reversal tasks. Two hypotheses — based on disease severity and possible deleterious effects of medication — are offered to explain the reversal impairments of the PD patients and the results are discussed in terms of the role of dopamine in reward-based learning.

Introduction

Introducing dopaminergic medication, in the form of the dopamine (DA) precursor levodopa, or DA agonists such as bromocriptine, pergolide or ropinirole, usually has a dramatic ameliorative effect on the motor symptoms of Parkinson’s disease (PD). The effects of such medication on the cognitive symptoms are much less consistent. Although many report a lack of effect, a number show improvement, particularly on tasks which utilise ‘frontal lobe functions’ [3], [22], [33], [48]. A few studies report impairment of cognitive function [21], [32], [51], [60]. Such a lack of consensus should not be surprising, given the differential effects of the disease process on various realms of cognitive function, how these vary throughout the course of the disease and the different modes of action of anti-Parkinsonian drugs. The present study attempts to examine specifically the performance of PD patients on tasks which are assumed to utilise particular brain regions and neurochemical pathways. These assumptions are explored by also including in the study patients with focal cortical lesions, contrasting medicated PD patients at different stages of the disease and comparing medicated with non-medicated patients.

Much of the work in recent years into the cognitive impairments in Parkinson’s disease (PD) has incorporated some assessment of ‘set-shifting’ ability. Such assessments employ tasks which require subjects to alter the rules currently guiding behaviour, where these rules are at the level of an attentional set (a learned pre-disposition to attend to one dimension — e.g. shape or colour — of multidimensional stimuli). This type of set-shifting usually takes the form of rejection of an ‘attend to A, ignore B’ rule in favour of a contradictory rule ‘attend to B, ignore A’. It is central to performance on the Wisconsin Card Sorting Test (WCST) and other sorting tests as well as the Odd-Man-Out task and the CANTAB attentional set-shifting (‘ID/ED’) task, all of which are sensitive to impairments in PD [3], [9], [16], [18], [45], [49]. A different form of cognitive flexibility is required in ‘reversal’ tasks, where the learned reinforcement contingencies of stimuli are reversed. It therefore requires the rejection of a learned rule ‘choose A, not B’ in favour of a contradictory rule ‘choose B, not A’. These two types of cognitive flexibility appear to be distinct both psychologically [63] and neurally [65] and can be simulated via separate mechanisms in computerised models of discrimination learning [31].

The ‘ID/ED’ attentional set-shifting task [55] allows reversal and attentional set-shift elements to be assessed independently within a visual discrimination paradigm. In this task, subjects have to learn a visual discrimination between stimuli made up of a shape, a line or both. At the start of the task, the two stimuli are composed of one dimension, shape or line; in later stages, the other dimension is added (superimposed) but it is initially irrelevant to the discrimination. These stages cue subjects to attend to the first dimension. At the crucial extra-dimensional (ED) shift stage, subjects must shift their attention to the previously irrelevant dimension to solve the discrimination (i.e. a shift of attentional set). There are a number of reversal stages within the task, where the reward contingencies of the currently-relevant dimension are reversed.

Dias et al. [13] examined the ability of marmosets with discrete lesions of different prefrontal cortex (PFC) regions to inhibit responding on the basis of an acquired attentional set (ED shifting) vs that on the basis of learned stimulus-response-reward associations (reversal) using a version of the ID/ED task. A double dissociation was revealed such that lateral, but not orbital, lesions impaired ED shifting whereas orbital, but not lateral, lesions impaired reversal. Evidence from human neuropsychology appears so far to support this dissociation. Early Huntington’s disease (HD, in which striatal damage is relatively limited to the dorsal caudate nucleus [23], which is connected to dorsolateral PFC [1]), is sensitive to the ED shift but not the simple reversal stages of the ID/ED task [35], whereas patients with dementia of the frontal type (DFT, which primarily affects the ventral PFC) show deficits in a number of reversal stages of this task but not at the ED shift stage [52]. In addition, a recent PET study has indicated that different brain regions participate in reversal and set-shifting operations [57].

PD patients (including unmedicated as well as mild and severe, medicated sub-groups) exhibit specific impairments at the ED shift stage of the ID/ED task [16], [45]. In this respect their performance parallels that of patients with lesions of the prefrontal cortex (PFC) and early HD patients [35], [47]. Impairments in these groups are not identical, however, and it seems that the shifting impairments in PD may be less easily accounted for by perseveration (of attention to the previously-relevant stimulus dimension) than those in other patient groups [34], [46].

The purpose of the present study was to examine the performance of PD patients on reversal tasks to determine whether cognitive flexibility impairments in these patients are restricted to the level of attentional set or may be generalised to include the level of stimulus-reinforcement associations, and to relate this to the stage of disease and possible effects of dopaminergic medication.

In mild PD, dopamine depletion within the caudate nucleus is limited to the most dorsal rostral head of this nucleus (whereas depletion is severe in all portions of the putamen [29]). It may be expected therefore that in mild PD, as in early HD, set-shifting will be more vulnerable than reversal. To an extent, this has been borne out experimentally, with no reversal impairments having been identified among PD patients using the ID/ED task [16], [45]. However, one difficulty with drawing such a conclusion is that, for control subjects at least, the reversal stages of this task are very easy when compared with the ED shift stages in that they may often be accomplished with as few as one error. It is highly unlikely that an ED shift may be achieved in so few trials; rather, it is usually necessary to integrate information from a number of trials in order to identify both the relevant dimension and the correct exemplar within that dimension. This obviously calls for the greater availability of resources such as focused attention and working memory than does a reversal. Consequently, we have developed two new probabilistic and concurrent discrimination reversal tasks. These tasks have recently been used with early HD patients, who exhibited perseverative impairments in the probabilistic task [34]. Reversal in these tasks is less likely to be accomplished in just one trial than in the original task; this should help to eliminate a ceiling effect in the data and allow a possible subtle impairment of reversal to be identified in PD patients.

To address the possible neurochemical basis of the underlying processes, we divided the PD group into 3 sub-groups, based on clinical severity and medication status. Comparing groups at different stages of the disease is particularly relevant in the context of these tasks because of the hypothesis that separate cortico-striatal loops, whose function may be disturbed at different disease stages, may mediate different cognitive processes (i.e. reversal and set-shifting). The effect of dopaminergic medication on the reversal tasks may be assessed by comparing the performance of unmedicated and medicated patients; the results of this comparison may then be compared with the seemingly beneficial effects of medication on set-shifting tasks [3], [33]. Gotham et al. [21] showed that medication may not be beneficial to performance on all cognitive tasks, even those thought to be subserved by the prefrontal cortex such as conditional associative learning and subject-ordered pointing. They put forward the hypothesis that in a relatively intact dopamine-dependent system, the medication which is given to treat the motor symptoms of Parkinson’s disease may effectively produce an ‘overdose’ effect, with levels being too high to allow normal function. Dopaminergic activity may be expected to be especially relevant to reversal tasks because of its apparent importance in the neural mechanisms of reward-based learning, including visual discrimination learning [24].

To provide evidence on the neural specificity of a potential impairment, two groups of patients with focal cortical — frontal or temporal — excisions were also tested. Temporal lobe structures, including rhinal cortex and inferior temporal cortex, are implicated in visual memory processes including storage of visual representations and associations with reward [10], [42], but appear to be less crucial in reversal learning and attentional set-shifting [11], [47]. Impaired learning of the original discriminations may therefore be expected in this group. As already stated, orbital and ventrolateral regions of PFC have been implicated in the learning of discrimination reversals, with errors being predominantly perseverative in nature [12], [14], [25], [27]. Cells in orbitofrontal cortex appear to code the reward value of stimuli; they are able to alter their pattern of responding very rapidly in response to changing reinforcement criteria, such as occurs in reversal tasks [58]. A sub-group of frontal lobe lesion patients with orbital or ventrolateral damage may therefore be expected to have difficulty with these reversal tasks, whereas those with damage restricted to medial or dorsolateral regions may show relatively intact performance.

The caudate nucleus, depleted of dopamine in PD, is part of cortico-striatal loops which converge on both ITC and PFC [1], [39]. Comparing the performance of the cortical excision and PD patients may help to illuminate the nature of functional interactions of cortical structures with the striatum during the learning and reversal of probabilistic and concurrent visual discriminations.