Higher synthesis coupled with reduced transporter activity may act to elevate synaptic dopamine concentrations in older adults, potentially reflecting compensatory mechanisms for counteracting losses in receptor density. However, research to date has not supported a clear role of elevated synthesis in benefiting cognition . Indeed, inverted-U-shaped relationships are observed for dopamine synthesis , suggesting higher synthesis levels may, in some cases, be detrimental. Therefore, it may not be sufficient to consider aging as simply a hypodopaminergic state, but instead may be a dysregulated state characterized by a loss of balance between presynaptic and postsynaptic components of the system. Similar conceptualizations of dopamine dysregulation may be applied to psychological and psychiatric disorders including schizophrenia, ADHD, and addiction . This dysregulation may critically affect the precision of dopamine signaling, increasing the variability of its temporal dynamics and noise . There is emerging evidence that regulation of dopamine receptors and age-related decline in receptor density is not spatially homogeneous . Rieckmann and colleagues report age-related reductions in interregional correlations in D1 BPND using [11C]SCH23390. Their findings suggested that aging is associated with a dissociation in D1 receptor regulation between nigrostriatal and mesocortical/mesolimbic pathways. Seaman and colleagues evaluated regional differences in estimated rates of percentage of change in D2/3 BPND using [18F]fally pride . Their findings revealed tremendous heterogeneity across regions in the estimated rate of decline, which was variably linear or curvilinear. Even after correcting for differences in gray matter volume, procona system they estimated the most extreme reductions in subgenual frontal cortex and superior temporal gyrus and less pronounced reductions in ventral striatum, pallidum and hippocampus .
Altogether, findings in aging indicate that changes in the dopamine system vary across individuals, vary across different presynaptic and postsynaptic components, and vary spatially across the brain. In the following three sections, we highlight instances in which accounting for these three sources of variability may shed light on reported effects in valuebased decision-making.Loss of dopamine has been linked with propensity to avoid punishment rather than approach reward. The probabilistic selection task is a well-known decision-making task that has been used to study these questions and has been applied to healthy adults and a variety of patient populations including those with Parkinson’s disease and schizophrenia . It ostensibly taps into processes of modelfree reinforcement learning, with some evidence suggesting it is preferentially associated with striatal rather than PFC dopamine function . Despite links to dopamine, there have been largely null findings in previous studies examining consistent group-wise effects of age on performance. Here we provide background on the PST, and models of dopamine’s role via the D1-mediated “direct” and D2-mediated“indirect” pathways. We propose that any measure that captures individual differences in dopamine function may useful in clarifying these null effects. Further, we discuss specific ways in which D1 and D2/3 receptor imaging can be leveraged to contribute to our basic understanding of the role of these pathways in human decision-making as well as individual differences in performance in aging. The PST is composed of both a probabilistic learning phase and a choice phase. Its design aims to assess potential biases in choice action to approach reward versus choice action to avoid punishment.
Briefly, three sets of Japanese hiragana characters are presented in randomized order and are associated with differing probabilities of reward or punishment . During the learning phase, participants learn to select the stimuli associated with reward. However, this learning could be driven by positive reinforcement from reward or negative reinforcement from punishment . The choice phase of the PST is designed to dissociate these possibilities. The stimuli are presented in novel pairs to reveal underlying choice biases in incentive motivation to either approach reward or avoid punishment . The bias to avoid punishment was demonstrated in Parkinson’s patients tested off medication, but was reversed when patients were tested on medication in presumably dopamine-replete states . Since the original description of the PST task, there has been significant interest in understanding how age-related changes in dopamine function may affect biases in decision-making. A simple hypodopaminergic account of aging would predict that age effects mimic those observed in Parkinson’s disease, but to a lesser degree given the relative sparing of dopaminergic function . Such biases, if they produce inoptimal choice behavior, would be a prime target for intervention in aging. Behavioral evidence of age-related biases in choice selection is mixed, with greater individual differences reported for older adults. A bias to avoid punishment has been shown in an older subset of older adults, but not in a younger subset . Some studies report greater individual differences in the balance between positively and negatively motivated choices in aging , or have shown selective reduction in positive learning but not negative learning . Other aging studies have found no effects of valence . We posit that accounting for individual differences in the decline of dopamine function using specific dopamine targets in striatum or neuromelanin-sensitive MR will clarify these null and mixed results. Previous research using MR approaches for assessing midbrain dopaminergic nuclei suggest a benefit of greater structural integrity for reward learning .
We predict that the subgroup of older adults with a positive choice bias to approach reward will have greater midbrain neuromelanin MR signal than older adults with a negative choice bias .The proposed mechanisms of dopamine’s involvement in approaching reward versus avoiding punishment involve the weighting of two circuits, the direct pathway versus the indirect pathway, involving substantia nigra, striatum, globus pallidus, and thalamus. Activation of the direct “go” pathway leads to the disinhibition of thalamus, facilitating outputs to cortex. Conversely, activation of the indirect “no-go” pathway suppresses thalamic output to cortex. The direct pathway is associated with reward-based learning and approach, which may be mediated by stimulation of striatal D1 expressing medium spiny neurons . The indirect pathway is associated with aversion-based learning and avoidance, which may be mediated by stimulation of striatal D2 expressing medium spiny neurons . Frank and colleagues suggest that low tonic dopamine amplifies learning through D2 negative reinforcement mechanisms and accounts for biases to avoid punishment in Parkinson’s patients tested off medication . A recent PETstudy in healthy young adults directly probed relationships between individual differences in D1 and D2/3 BPND and propensity to approach reward versus avoid punishment in the PST . However, higher D2/3 BPND was not clearly related to a bias to avoid punishment in this young adult sample. It is currently not known whether D2/3 effects would emerge if these healthy subjects were tested in a dopamine depleted state . It is possible that individual differences in relative ratios of D1 and D2/3 receptor densities may underlie between-subject variability in PST performance in older adults, or variability in performance on other tasks, such as tasks involving risky gambles, which tap into approach versus avoidance mechanisms. There is some evidence, though limited, suggesting D1 and D2/3 receptor densities decline at different rates across the lifespan . Estimated rates of decline are numerically greater for D1 receptors than D2/3 receptors for between-subject comparisons of PET data and within-subject analyses of postmortem tissue . However, additional research is needed to establish whether there are asymmetric effects of aging on D1 versus D2/3 receptors. It is unclear what the underlying physiological mechanism might be for the relative vulnerability of D1 receptors or relative resilience of D2/3 receptors in aging. D1 and D2-expressing medium spiny neurons have distinct morphological and electrophysiological properties which may confer unique susceptibilities. In culture, D2, rather than D1 receptors, procona valencia buckets may be more vulnerable to excitotoxic insults . However, observations in early Parkinson’s disease reveal upregulation of D2/3 receptors, but not D1 receptors . This leaves open the possibility that differences in the capacity for receptor upregulation may underlie reductions of the ratio between D1 and D2/3 in healthy aging. Regardless of mechanism, the direction of these effects observed in aging is generally consistent with the view that aging shifts choice behavior toward a bias to avoid punishment rather than approach reward.There is, however, significant inter individual variability in the trajectory of age-related changes in the dopamine system, which would warrant examination of within-subject ratios of D1 to D2/3 receptor densities and their relationship with performance. For the PST, evaluation of individual subject performance has revealed subgroups of older adults with a “positive” choice bias to approach reward and subgroups with a “negative” bias to avoid punishment . Subjects with losses in D1 receptors but relative preservation of D2/3 receptors may show reduced choice behavior to approach reward , and greater choice behavior to avoid punishment . Subjects with losses in D2/3 receptors but relative preservation of D1 receptors could be expected to show the opposite pattern of results .
In vivo PET imaging could resolve the underlying neural basis of these individual differences in positively and negatively motivated choices in aging. Further, complementary studies in animal models could test whether selective knock-down of D1 versus D2 receptors generates a similar pattern of results.While the density of dopamine receptors declines with age, there is evidence for counteracting increases in dopamine synthesis and decreases in dopamine reuptake via reduced transporter BPND . If imbalance in the tuning of pre and postsynaptic components occurs in aging, this dysregulation may lead to reduced precision of RPEs implicated in value-based reinforcement learning . Such dysregulation would be expected to result in slower model-free reinforcement learning. Behavioral evidence suggests that while older adults perform comparably to young adults when cue reward contingencies are deterministic older adults show impaired performance in situations in which outcomes are probabilistic or require learning from feedback . Though behavioral evidence indicates impaired reward-based learning in older adults, it is not clear that neural activity associated with reward anticipation or reward outcome is systematically altered in aging. Ventral striatum/nucleus accumbens activation in response to reward-predicting cues is the same in young and older adults . Further, responses to rewarding outcomes in ventral striatum and medial PFC have been shown to be similar in young and older adults . These measures of BOLD activation suggest there are no systematic differences in reward responsivity in aging. However, these measures typically rely on averages across many trials, and may not capture age-differences in trial-to-trial variability. Fruitful lines of research in aging have linked reward-based learning with neurocomputational approaches to examine age differences in RPE-like BOLD signal that rely on trial-based estimates. Together, these studies suggest that aging reduces correlations between RPEs derived from reward learning tasks and BOLD activation in ventral striatum/ nucleus accumbens and ventromedial PFC . Few studies have linked age-related reductions in these correlations with alteration in dopamine function in aging. In one notable exception, Chowdhury, Guitart-Masip, Lambert, Dayan, et al. pharmacologically manipulated dopamine to examine its effects on learning and RPEs. This study demonstrated that treatment of older adults with levodopa increased both RPE-like signals in ventral striatum and rates of learning. A recent study using the same task probed relationships between striatal D1 BPND and nucleus accumbens RPE’s in young and older adults . Surprisingly, they did not find striatal RPE-like responses in either young or older adults. D1 BPND was not correlated with performance for either group but was positively related to ventromedial PFC signal associated with reward anticipation. Moving forward, it will be valuable to consider how dopamine changes in aging may fundamentally alter the reliability in dopamine signaling to affect reward-based learning. This can be achieved in studies in animal models that examine the effects of age on the amplitude and timing of phasic responses. In humans, foundational studies could investigate whether neuromelanin-sensitive MR measures of midbrain dopamine function are related to the strength of correlations between RPEs derived from reward learning tasks and striatal BOLD signal in older adults. Higher correlations may be predicted in older adults with higher midbrain neuromelanin MR contrast-to-noise ratios. Applying PET methods, future studies could consider relationships between presynaptic and postsynaptic components within subject. For example, individual differences in the ratio of dopamine transporter availability and D2/3 receptor availability within subject may correlate with rates of learning . One may predict an inverted-U-shaped relationship between a presynaptic/ postsynaptic composite measure and learning suggesting that an optimal balance in dopamine receptor binding and reuptake is associated with more precise RPEs and more efficient learning rates in aging.