We used novel probes of oculomotor decision-making to demonstrate relative insensitivity to reward in an individual with apathy following bilateral GPi lesions. Our TLT (Adam et al., 2012
) requires reward sensitivity and motivation or effort to succeed, combined with fast reaction times and the ability to update behaviour in response to positive and negative feedback. A reactive response – simply waiting for the green light – is less well rewarded than an anticipatory response prepared in advance of the green signal. KD initially made very few anticipatory responses compared with age-matched controls. However, dopaminergic therapy, first with levodopa and then with ropinirole, increased anticipatory responses to within the normal range.
The directional saccade reward-sensitivity task, originally developed for the study of reward sensitivity in macaque monkeys (Hong and Hikosaka, 2008
), demonstrated that KD had SRTs within the normal range but showed no speeding to the rewarded side (RS), unlike healthy volunteers. Treatment with levodopa led to reward sensitivity, with speeding of responses to the RS and slowing to the unrewarded side (US) compared to baseline. Off medication, the difference in SRTs to rewarded and unrewarded targets became non-significant, while subsequently on ropinirole, a direct dopamine D2/D3 receptor agonist, KD again demonstrated reward sensitivity, as well as generalized speeding.
These effects on dopaminergic medication were associated with clinical improvement – reduction of apathy and increased motivation to find work and in social interactions – most prominently while on the dopamine agonist. The findings demonstrate a causal relationship between basal ganglia function and motivation or willingness to make an effort for reward. They provide proof-of-concept data for the treatment of apathy which is increasingly recognized to be a key component of several neurological disorders (Bonelli and Cummings, 2008
; Marin, 1991
; Chow et al., 2009
; Starkstein, 2009
Unlike other tasks involving risk, such as the Iowa Gambling Task (Bechara et al., 1994
) or the Cambridge Gamble Task (Clark et al., 2004
), our TLT requires participants to take risks by making anticipatory responses. Many other paradigms place certain and risky options on an equal footing with the same amount of effort required for both choices. This has the benefit of establishing risk preferences independently of effort but tends to favour a careful, deliberative response strategy. The traffic lights paradigm imposes time constraints on decisions and rewards behaviour that might be considered ‘functionally impulsive’ (Dickman, 1990
): on this task, it can be functionally useful to make anticipatory responses because these can lead to greater rewards, analogous to many situations in real life. It is possible that KD's lack of anticipatory responses on this task reflects risk aversion
, rather than lack of motivation or unwillingness to make an effort for rewards. However, it is less easy to explain how such a mechanism might account for behaviour on the directional saccadic task, where there was no risk of incurring a penalty.
How did dopamine reverse apathy and reward insensitivity? Substantial evidence links dopamine to reinforcement learning (Schultz, 2007
). However a growing body of research also implicates dopamine in effort-based decision-making, generating the motivation and vigour to overcome costs of initiating actions (Niv et al., 2007
; Kurniawan et al., 2011
). The progressive improvement of KD's performance on the TLT immediately post l
-dopa (B) is suggestive of dopaminergic enhancement of learning. However, during the drug holiday period such learning was radically reversed (C), suggesting that if this effect was solely due to a reinforcement learning effect of l
-dopa it had not been completely consolidated. Dopamine was still required to maintain it.
On the directional reward-sensitivity task, l
-dopa also had a dramatic effect after its introduction, speeding saccades to the RS (). During the drug holiday, however, there was no longer any significant reward-sensitivity but saccades were generally faster than before treatment, suggesting there were some general, non-specific effects of practice on the task. The time course of action on reward-sensitivity and its reversal during the drug holiday makes it unlikely that dopaminergic effects on synaptic plasticity and learning were the only mechanism of action. Instead, it might also have had an effect on response vigour or overcoming costs of effort (Niv et al., 2007
; Kurniawan et al., 2011
Dopamine could act directly on brain systems left intact after stroke, but perhaps disconnected because the major outflow from the basal ganglia is via the GP. Alternatively, because the GPi lesions were not complete in KD, it is possible that his lesions led to imbalance in cross-talk between striatal regions which could be ameliorated by dopamine therapy. It has been demonstrated that parallel corticostriatal loops through the basal ganglia need not operate in isolation but can instead communicate with each other, e.g., via spiralling striato-nigro-striatal connections (Haber et al., 2000
) which allow ventral striatal regions to influence more dorsal striatal areas. Moreover, the nigrostriatal system is not the only dopaminergic modulator of basal ganglia function; the intra-striatal dopaminergic system is complex and can alter with denervation (Smith and Kieval, 2000
). Finally, it is important also to consider the possibility that the effects of dopamine observed in KD might arise from indirect, knock-on effects on other neurotransmitter systems, e.g., there is evidence of interactions between dopaminergic and noradrenergic systems (Hara et al., 2010
) as well as several other neurotransmitters (see Steiner and Tseng, 2010
, for reviews).
In macaques, using the directional reward saccade task, Hong and Hikosaka (2008)
found that saccades to the RS with shorter latency than to the US, with reward-related speeding being associated with activity in GPi neurons which project to the lateral habenula. If a homologous circuit operates in the human brain, it is likely to have been partially disrupted in KD in whom both GPi were damaged. However, the lateral habenula remained intact, together with the caudate and putamen. Furthermore, SPECT imaging of the DAT demonstrated that the nigrostriatal dopaminergic pathway was intact as there was good signal bilaterally in the caudate and putamen of KD. Thus one locus of dopaminergic drug action is potentially the intact caudate, putamen or even surviving parts of the GP complex.
Another potential site of action of dopamine is prefrontal cortex. The OFC, in concert with basal ganglia structures, is considered to have a special role in the processing of reward signals (Schultz, 2000
; Kringelbach and Rolls, 2004
; Wallis, 2007
). Projection of KD's lesion onto the known topography of the pallidal trans-thalamic connections to the cortex, determined using diffusion-weighted tractography (Draganski et al., 2008
), suggests that the connections to the VMPFC and OFC have most likely been disrupted (). OFC neurons not only respond selectively to reward or aversive stimuli, but also signal relative preference for rewards and may integrate different types of information to compute a representation of value (Thorpe et al., 1983
; Tremblay and Schultz, 1999
; Padoa-Schioppa and Assad, 2006
; Wallis and Kennerley, 2010
). Consistent with these neurophysiological findings in macaque monkeys, imaging studies in humans have described activations in OFC and VMPFC which correlate with behavioural measures of stimulus value (O'Doherty, 2004
; Plassmann et al., 2007
; Rangel and Hare, 2010
; Haber and Knutson, 2010
; Glascher et al., 2009
; Blair et al., 2006
Lesions of the OFC in humans lead to impaired decision-making about the expected outcome of choices (Bechara et al., 1998
) while alterations in striatal dopamine binding in drug addicts is associated with hypoactivity in OFC (Volkow et al., 2009
). Dopaminergic neurons are known to innervate prefrontal cortex, including OFC (Williams and Goldman-Rakic, 1993
). Although these arise from midbrain dopaminergic populations, partial disconnection of OFC neurons from trans-thalamic pallidal inputs – as is likely in KD – might disrupt dopaminergic reward signals within OFC. This view is compatible with recent functional imaging evidence that dopamine agonists might alter decision-making and risk-taking in susceptible individuals with Parkinson's disease via actions on OFC (van Eimeren et al., 2009
Intriguingly, previous work also suggests that a dopaminergic deficit might be an important contributory factor to apathy in Parkinson's disease, which occurs in up to 60% of cases (Oguru et al., 2010
). Patients who undergo STN deep brain stimulation (DBS) often require reduction or withdrawal of dopaminergic therapy because of improvements in motor control following surgery. Czernecki et al. (2008)
reported that apathy occurred after dopamine withdrawal in some of these cases, but importantly it could be reversed with ropinirole. More recently, a PET study has demonstrated greater mesocorticolimbic dopaminergic denervation involving the OFC in Parkinson's disease patients who develop postoperative apathy compared to those who do not (Thobois et al., 2010
Regardless of the precise locus of drug action in KD, it is clear that his lesions rendered him apathetic but this could be ameliorated by dopaminergic modulation. Alteration in reward-sensitivity mirrored clinical changes, suggesting apathy in this case is associated with lack of motivation to obtain rewards. Animal learning theory has proposed that rewards might in fact constitute the basic goals of voluntary behaviour (Dickinson and Balleine, 1994
). From this perspective, the absence of sensitivity to rewards would be expected to have devastating consequences for goal-directed action, just as one observes in apathy. But note that although this view might account for behaviour in our particular case, apathy is most likely to be a syndrome that is multidimensional (Cummings, 1993
; Levy and Dubois, 2006
). In different clinical contexts, it could potentially result from deficits in other cognitive components of the decision-making process. Further studies are required to delineate these components and which specific deficits occur in different clinical conditions. Our study represents progress towards understanding one component of apathy – namely, relative reward insensitivity.
Although cases such as KD with bilateral GPi lesions are rare, apathy is common in Parkinson's disease (Oguru et al., 2010
; Pedersen, et al., 2009
; Starkstein, 2009
), as well as in other neurodegenerative disorders, including Huntington's and Alzheimer's disease (Bonelli and Cummings, 2008
; Chow et al., 2009
; Starkstein et al., 2006
; Marin, 1991
). These conditions often involve disruption of cortico-striato-thalamo-cortical loops (Alexander et al., 1986
) but the mechanisms underlying apathy when there is widespread neurodegeneration has been difficult to study. Focal lesion cases such as KD provide important information about the neural substrates underlying apathy and modulation of this behavioural state with neuropharmacological intervention.