After starting on medication, some people with Parkinson’s disease undergo a dramatic personality change, becoming impulsive, addiction-prone pleasure-seekers. In particular, a high rate of pathological gambling among these patients has led researchers to rethink how the neurotransmitter dopamine works in both diseased and normal brains.
Dopamine is depleted by Parkinson’s disease and resupplied by medications that treat the disease. Thus the behavioral shift in these cases would seem to be caused when medication overcompensates for a lack of dopamine in the brain’s reward and learning circuits. Recent research into these cases, however, has only deepened the mystery of dopamine’s role in reward and learning.
“It’s awfully complex, and there’s a lot that’s not yet known,” says Alain Dagher, a researcher at McGill University in Montreal and first author of a February 2009 review article on the subject in Neuron.
Parkinson’s medications do seem to restore dopamine levels to near normal in brain regions most affected by the disease, such as the dorsal striatum, Dagher says. These medications also may push dopamine levels above normal in other, less disease-affected centers such as the ventral striatum—one of the hotspots of the brain’s reward system.
Recent studies of people with Parkinson’s suggest that newer and more powerful dopamine “agonists,” which bind to dopamine receptors on brain cells, are more likely to produce this “ventral overdose” effect. A review in Movement Disorders in 2007 found that pathological gambling behavior has been diagnosed two to eight times more frequently among Parkinson’s patients who take these agonists than among the general population. Yet it is not easy to reconcile such an effect with current models of dopamine signaling.
According to these models, the perception of an unexpected reward stimulus leads to a short-term burst of excitement among dopamine-producing neurons in the midbrain, near the brainstem. The brief episode of increased firing results in a quick but sizeable surge of dopamine along these neurons’ output fibers and into target regions in the striatum. In the ventral striatum, this dopamine surge is widely thought to be a primary learning signal, effectively telling the brain how much to value the relevant stimulus.
However, this signal is known to be mediated mainly by dopamine “D1” receptors on striatal neurons. And as Dagher points out, “dopamine agonists that are used clinically don’t act on the D1 receptor. They act only on dopamine D2 and D3 receptors.”
People with Parkinson’s, perhaps because of their significant loss of midbrain, dopamine-supplying neurons, show an increased expression of D3 receptors on the striatal neurons that normally express D1 receptors.
But how these drugs could boost a person’s valuation of rewards is still unclear, says Dagher, because by flooding the system with dopamine stimulation, these agonists should interfere with any transient dopamine signal coming from the midbrain. The fact that patients who take such drugs are still capable of reward-based learning suggests that this signal is not as important for learning as currently thought. “So maybe the models we have are not completely correct,” says Dagher.
Impaired Reversal Learning
Tests of dopamine agonists in healthy people have not resolved these questions, because healthy brains do not respond like the brains of Parkinson’s patients. As Dagher points out, in healthy people, “one of the effects of dopamine agonists is actually to stop dopamine neurons from firing.”
Researchers believe this happens because midbrain and striatal dopamine-producing neurons normally express a kind of dopamine D2 receptor at the ends of their output fibers. These D2 “autoreceptors” form part of a negative feedback loop to restrict the flow of dopamine: As synaptic dopamine levels build up, the D2 autoreceptors are increasingly stimulated, and the flow of dopamine is increasingly suppressed. Pramipexole and other D2-stimulating agonists magnify this effect.
In Parkinson’s patients, however, dopamine agonists don’t seem to work via D2 autoreceptors. “In Parkinson’s those neurons are mostly dead,” says Dagher. For such patients, D2 stimulation by dopamine agonists may occur primarily on a different set of D2 receptors, which lie on the other side of striatal synapses.
These other D2 receptors are believed to be sensitive primarily to the brief reductions in dopamine that follow the omission of an expected reward or the occurrence of an unexpected punishment. Animal and human experiments suggest that by smothering this transient signal, D2 agonists impair a form of learning known as reversal learning, in which an aversion develops after a stimulus stops being rewarding and starts being punishing.
In 2007, a group at Cambridge University led by postdoctoral student Roshan Cools showed that in Parkinson’s patients the drug levodopa, the mainstay of dopamine-replacement treatment, disrupts normal reversal-learning-associated activity in the nucleus accumbens, a part of the ventral striatum, as measured using functional magnetic resonance imaging (fMRI). When the patients stopped taking the drug, the impairment and its associated fMRI-signal disruption went away.
“It seems to be particularly those forms of flexibility that require learning from punishment that are impaired by the medication,” says Cools, now a researcher at Radboud University in the Netherlands. She suspects that this impairment reduces a person’s ability to escape from once pleasurable but ultimately self-destructive behaviors, such as gambling despite mounting losses.
In more recent work, Cools and other researchers have found that the effect of D2 stimulation on reversal learning seems far from straightforward: In ordinary subjects this effect depends at least partly upon the background level of dopamine production in the striatum. Like Dagher, Cools acknowledges that researchers still have a great deal to learn about how dopamine drugs cause these striking behavioral changes in people with Parkinson’s.
Taking Off the Prefrontal Brakes
Other researchers have focused on the role of the prefrontal cortex in reward- and punishment-driven learning—and on dopamine’s ability to modulate that role.
“We tend to think about reversal learning as one of the more frontally associated processes,” says Luke Clark, a researcher at Cambridge University who studies addiction behaviors such as pathological gambling.
Clark cites evidence from animal and human studies that reversal learning activates parts of the frontal cortex that represent or process negative consequences; this activity tends to be impaired in drug addicts and pathological gamblers. People with trauma or stroke damage to their orbitofrontal cortex or the ventromedial prefrontal cortex also are known to perform poorly on gambling-type tasks and often show a striking “disinhibition” in their behavior, with reduced impulse control and relatively little ability to anticipate or learn from adverse consequences—very much like people with Parkinson’s who develop impulse control disorders.
How would dopamine agonists block the influence of these prefrontal areas and thereby impair reversal learning? According to one prominent hypothesis, based on work in rodents and published in 2005 in Nature Neuroscience, the influence of prefrontal signals on striatal neurons is regulated by striatal dopamine levels, via D2 receptors. “Dopamine can act as a traffic cop to decide which input has the most influence” over striatal neurons, Dagher says.
In this view, too much of a D2-agonist Parkinson’s medication would dial down the prefrontal influence, allowing behavior to become less restrained and less thoughtful. Understanding how this prefrontal-limbic balancing act can be disrupted is an area of intense research, Clark says.
What Does the Dopamine Signal Mean?
One lesson that could be drawn from this research is that dopamine does not appear to be all-important for learning, says Dagher: “Parkinson’s patients still do learn and are not severely impaired.” He adds that mice engineered to be dopamine-deficient also seem able to learn normally. “If dopamine is a learning signal, then how can an animal with no dopamine learn?”
Dagher suggests that there may be redundant networks for learning and cites evidence that background levels of dopamine stimulation can serve to amplify the reward signal, motivating behavior toward already learned rewards. If that’s true, then the higher background stimulation produced by a dopamine agonist in a person with Parkinson’s could promote an excessive wanting of already-learned pleasurable behaviors such as gambling. At the same time, the ability to learn from behavioral choices and to know which ones are better would remain somewhat intact.
“What happens in pathological gambling and in [drug] addiction is that people persist in the behavior despite negative consequences,” Dagher says. “But they’re aware of those negative consequences.”