10 April 2008. In a first for neuroimaging, a Princeton University research team has overcome technical obstacles to successfully measure blood oxygen level dependent (BOLD) responses in the ventral tegmental area (VTA). This small brainstem nucleus is of great interest to schizophrenia researchers—dopaminergic VTA axons project to the ventral striatum (VStr) to form the brain’s “reward circuit” (the mesolimbic dopamine system) and also to the prefrontal cortex (PFC; the mesocortical dopamine system), where they regulate functions such as working memory. Dysfunction in the dopaminergic systems originating in the VTA has been implicated in schizophrenia, not to mention addiction, depression, and other psychiatric disorders (see the Dopamine Hypothesis of Schizophrenia by Anissa Abi-Dargham).
As reported in the February 29 issue of Science, functional magnetic resonance imaging (fMRI) studies of dopaminergic circuits have so far been restricted to VStr and PFC because these structures are large enough to reliably produce BOLD signals using conventional fMRI techniques, which attain a spatial resolution of several millimeters per voxel. The VTA is only about 60 cubic millimeters, or two voxels, in volume. The imaging challenge presented by the small size of these nuclei is compounded by the several large, pulsatile arteries that supply the brainstem with blood. With each beat of the heart, the tiny brainstem nuclei shift sufficiently in space to create troublesome movement artifacts. Moreover, differences in brainstem anatomy across individuals have made it difficult to precisely align, or “register,” brainstem structures in data sets used for group analyses.
Complementary innovations at Princeton’s Neuroscience Institute (PNI) allowed first author Kimberlee D’Ardenne, PNI co-director Jonathan Cohen, Leigh Nystrom, and Samuel McClure, now at Stanford University, to precisely locate the VTA and to easily differentiate it from other nuclei, to control for cardiac-related artifacts, and to precisely register their data across subjects. The team observed clear BOLD signals that reflected the study participants’ expectations of rewards, a finding that dovetails with a “reward prediction error theory” of dopamine function derived from single-unit recording studies in nonhuman primates.
In this previous work on reward prediction error, Wolfram Schultz and colleagues trained monkeys in a classical conditioning procedure in which the monkeys learned to expect a juice reward at a fixed interval after the display of a visual cue (reviewed in Schultz et al., 1997). Before training, VTA neurons increased their firing immediately after the delivery of a reward; this was termed a “positive reward prediction error” because the reward was unexpected, and hence not predicted. After training, VTA neurons increased firing just after presentation of the visual cue, and subsided to baseline firing levels by the time the reward was provided; here, the reward was predicted correctly. However, if a reward was not presented at the expected interval, a “negative reward prediction error” occurred: VTA neurons increased their firing after the visual cue and returned to baseline, but then markedly reduced their firing just after the time the reward was expected. Schultz and colleagues concluded that “dopaminergic activity encodes expectations about external stimuli or reward.”
The Princeton group studied a group of thirsty humans in an adaptation of this procedure that allowed for measurement of both positive and negative reward prediction errors after training. By sometimes delaying the delivery of juice or water rewards to subjects, the group was able to create conditions in which an expected reward was delivered at an unexpected time, hence eliciting a negative prediction error followed by a positive prediction error.
In high-resolution proton-density weighted images, the bow-shaped substantia nigra, another brainstem dopaminergic structure, was clearly visible, and provided a landmark to accurately locate the triangular VTA along the brain’s midline for fMRI. Functional imaging was synchronized with the subjects’ pulse, and using a new normalization algorithm, BOLD signals from the VTA could be properly registered anatomically, despite the varying brainstem anatomy of the subjects.
The researchers found that the BOLD signal in the VTA was significantly related to positive, but not negative, reward prediction errors. Conversely, the BOLD signal in the VStr was significantly related to negative reward prediction errors. However, there was a positive correlation between the VTA and VStr BOLD response to unexpected rewards, indicating that the VTA positive prediction error signal influences the BOLD signal in VStr.
To ensure that these findings apply to different types of rewards, the researchers conducted a second set of experiments in which subjects were presented with a number from 0 to 10 and asked to guess if a subsequently presented number would be larger or smaller. For each correct guess, the subjects were given $1.00. Again, the VTA BOLD responses reflected a positive reward prediction error, and there was good anatomical overlap between the VTA regions activated by juice or water and those activated by a monetary reward.
The Princeton group says that their work is generalizable in other ways as well, for example, to other small brainstem nuclei that contain serotonergic and noradrenergic neurons, which bodes well for future fMRI research on the role of monoaminergic modulatory systems in psychiatric illness.—Peter Farley.
D’Ardenne K, McClure SM, Nystrom LE, Cohen JD. BOLD responses reflecting dopaminergic signals in the human ventral tegmental area. Science. 2008 Feb 29;319 (5867):1264-7. Abstract