These new analyses used generalized psychophysiological interactions , a technique that allows for the examination of connectivity between two brain areas as a specific function of the task of interest. gPPI was developed to allow for the use of more than two task conditions in the same model while also improving model fit . The evaluations were centered around two hypotheses. Hypothesis 1 is based on observations from fMRI studies of young individuals who did not have an AUD history prior to testing where regional BOLD differences were observed during placebo conditions. This prediction states that compared with high LR individuals, those with less intense response to modest alcohol doses will demonstrate lower levels of functional connectivity between the amygdala and PFC when processing emotion-laden faces. Further, we predict negatively valenced emotions will elicit a greater decrease in connectivity relative to positively valenced emotions . Given prior studies demonstrating changes to brain functional connectivity following alcohol administration, Hypothesis 2 states that low LR individuals will have lower levels of functional connectivity following alcohol for negatively valenced emotions compared with high LR individuals.In the original study from which the current data were extracted , a survey was distributed to randomly selected 18 to 25-year-old European American and white Hispanic students at the University of California, San Diego, using methods approved by the UCSD Institutional Review Board. This questionnaire gathered information on demography, substance use, and DSM-IV psychiatric disorders using questions extracted from the Semi-Structured Assessment for the Genetics of Alcoholism interview . To be included in the fMRI protocol described below, subjects were limited to those who were: right-handed; had no history of brain trauma or epilepsy; no history of alcohol or drug dependence; no current major psychiatric disorder; were not pregnant; and had no irremovable body metal. A laboratory-based alcohol challenge was then used to establish an individual’s LR.
After confirming zero baseline breath alcohol concentrations ,container for growing weed subjects drank alcohol over a 10-min period given as a 20% by volume solution in a room-temperature carbonated beverage, a protocol that produced approximately equivalent BrACs across sexes . We used a median split on self-report scores from the Subjective High Assessment Scale during the alcohol challenge to determine low versus high LR in both the prior study and the present secondary analysis of those data. Once the LR status was confirmed through the initial alcohol challenge, two MRI sessions were scheduled on nonconsecutive days within 1 week of each other whenever possible. Alcohol and placebo sessions were carried out in an MRI scanner where participants received, in random order, the same alcohol dose as in the laboratory session or placebo in sessions that included the Hariri emotional face recognition task . Complete data required for functional connectivity analyses were available from 216 fMRI scans across placebo and alcohol challenge sessions for 108 individuals. The subjects were comprised of 54 pairs of low- and high-LR participants who were matched on sex, demography, drinking frequency and usual quantities, as well as tobacco and cannabis use. The data presented here were extracted from the Hariri emotional face recognition task, which was presented to participants in the MRI scanner 60 min post-beverage consumption during placebo and alcohol fMRI sessions. The task was measured at a time close to the peak BrAC during the average alcohol session. At this point of the alcohol session, the alcohol levels were similar in the LR groups.We used a modified version of the Hariri emotional face-processing task . Participants were presented with a target face and two probe faces for 5 s, with instructions to match the probe and emotional expression of the target by pressing a button in a block design. Each block had six consecutive trials where faces were either angry, fearful, or happy. Additionally, a sensorimotor control condition was used where vertical or horizontal ovals or circles were presented for six consecutive trials in block with instructions to match the shape of the probe to the target. Each block of emotion condition and the sensorimotor control condition were presented three times in a pseudorandomized order. The task began with an 8-s fixation period and had interspersed 12-s fixation periods between each block.
There were 18 trials for each condition yielding 512 s total task time. Figure 1 illustrates the timing of the task with examples from the first two blocks . We recorded accuracy and reaction time during the task to confirm the task was carried out correctly .The Analysis of Functional Neuro Images software package was used for preprocessing data and gPPI analyses. All T1-weighted images were skull-stripped and warped into Talairach Tournoux atlas space using a 12-parameter affine transformation. Next, functional data were visually inspected for scanner artifacts. To correct for small movements over time, image repetitions were registered to a selected base volume via AFNI’s 3dvolreg program. The time series of the motion parameters were used in the linear regression analysis of individual data to control for spin history effects . Functional data were then transformed into the participant’s anatomical space and aligned to standard space using the previously computed anatomical transformation matrix and smoothed with a 4-mm FWHM Gaussian kernel.The general linear models used reference vectors for the task conditions convolved with the hemodynamic response function. Estimated motion and linear, quadratic, and cubic trends were also included as nuisance variables. The resulting model generated scaled beta coefficient maps representing the mean percent BOLD signal change for each condition of the task relative to baseline.Functional connectivity analyses were conducted using gPPI with anatomical-defined seeds in the left and right amygdala. Seed regions were chosen using Neurosynth , a meta-analytic tool for selecting fMRI activation coordinates from a database of studies , where we examined studies using tasks of emotional faces to elicit amygdala activation. Specifically, we created a 5-mm sphere around the coordinates X = −26, Y = 6, Z = −14 and resampled to the template space of our fMRI scans. Visual inspection of these seeds ensured that they were anatomically constrained to the amygdala using the Talairach Tournoux atlas. For each participant, the average time series of the BOLD signal was extracted for each seed, and trends were removed. A one parameter gamma model was used to estimate the hemodynamic response function of the task and a deconvolution of the seed’s time series with this function was calculated using the 3dTfitter AFNI program yielding scaled coefficients. PPI regressor interaction terms for each condition of the Hariri task were computed by multiplying the mean time series of the de-convolved seed with the condition vector of interest, and then convolved with a gamma basis function using the AFNI program Waver.Separate voxel-wise GLMs were conducted for each bilateral amygdala seed. Each GLM contained the PPI regressors, the physiological regressor , the psychological regressors , and nuisance variables .
Each seed’s connectivity with other brain areas was examined separately. Whole-brain between-group differences in functional connectivity were estimated using mixed-effects ANOVAs with the AFNI 3dANOVA3 program where the within-subject factor was alcohol/placebo condition and the between-subjects’ factor was Low/High LR. We applied a cluster correction to guard against identifying false positive areas for potential functional connectivity differences. Specifically, we estimated the noise in our fMRI volumes using 3dFWHMx in AFNI and used that information to apply a cluster-size threshold of α = 0.01 for a given voxel-wise threshold of α = 0.01 to protect our α at a 0.01 level using the 3dClustSim AFNI program. Using this approach,cannabis drying the minimum volume for a significant cluster in our analyses was set at 448 µl or seven contiguous significant voxels.Connectivity analyses were carried out examining LR groups during the alcohol and placebo conditions using left and right amygdala seeds. Clusters showing significant LR, alcohol, or LR-by-alcohol interaction effects in functional connectivity differences in response to fearful, angry, and happy faces are described below and detailed in Table 2. Specifically, Table 2 lists brain regions, associated Brodmann area , peak activation, Talairach coordinates, and volume for each significant cluster. Figure 2 illustrates the pattern of functional connectivity main effects for exemplar regions, while Figure 3 illustrates the pattern of functional connectivity interaction effects for exemplar regions. For all significant clusters we also ran ANCOVA models covarying for sex and did not find any significant effect of sex in our results.During angry faces, using the left amygdala seed, the differences between low and high responders illustrate that low LR participants had lower functional connectivity in several cortical regions compared with high LR participants in both placebo and alcohol conditions. This was observed in the left medial frontal gyrus, bilateral ventral anterior cingulate, bilateral posterior cingulate, and left supramarginal gyrus. Using the right amygdala seed, the main effects of LR during processing of angry faces were identical to left amygdala such that low LR participants had diminished functional connectivity as compared to high LR participants in both placebo and alcohol conditions in the bilateral posterior cingulate. In contrast, during happy faces, significant effects were observed only with the right amygdala seed where, opposite to the LR main effects pattern observed for angry faces, low LR participants had greater functional connectivity compared with high LR participants in the right dorsal anterior cingulate gyrus and right caudate.During angry faces, within the right precuneus, lower functional connectivity was observed with the left amygdala following alcohol compared with placebo in low LR individuals, but increased connectivity following alcohol was found in high LR individuals. Interactions during processing of happy faces were characterized as a decrease in functional connectivity with right amygdala following alcohol as compared to placebo in low LR individuals, but increased connectivity following alcohol in high LR individuals in left middle frontal gyrus regions; a pattern similar to the interaction observed during angry faces.The main goal of these analyses was to expand upon the findings of Paulus et al. by evaluating LR group differences in functional connectivity. This is the first demonstration of differences in connectivity between low and high LR individuals, results that might help explain why drinkers with low LRs might require greater cognitive effort to perform optimally on some tasks. Consistent with our hypotheses, in the current analyses the most prominent patterns found were attenuation of amygdala connectivity with cortical regions in low LR participants, both during placebo and in response to alcohol, relative to high LR participants while viewing angry faces. Additionally, comparing alcohol and placebo sessions, low LR individuals exhibited decreased functional connectivity in response to alcohol, whereas high LR individuals showed increased functional connectivity while viewing both angry and happy faces. The present findings add to the regional brain changes originally reported by Paulus et al. . While those authors found no difference in amygdala regional activation patterns across LR groups, we found several differences in functional connectivity with the amygdala that varied depending on the valence of facial affect being decoded. Thus, BOLD signal changes in the cortex in response to an emotional processing task in low versus high LR individuals might reflect aberrant functional connectivity in corticoamygdalar circuits. These functional connectivity differences were observed in relatively highly functioning individuals from a nonclinical sample, some of whom carried an enhanced risk for alcohol problems through a low LR but none of whom had yet developed an AUD. Along with a growing body of evidence that the low LR phenotype is characterized by CNS differences relative to their high LR peers , these emotional processing data suggest that low LR individuals may have an altered neurobiological process in which they recognize some negative or positive social cues through decoding facial affect. It is also possible that given the pattern of brain regions in the frontal lobe, anterior cingulate, and insular cortex that showed altered amygdala functional connectivity, as well as regional BOLD signal changes in the insula, low LR individuals may have an impaired ability to recognize intoxication, or the rewarding aspects of alcohol, in social situations. Functional connections between the amygdala and cortical regions play important roles in processing human emotions. For example, connections between the PFC and amygdala facilitate the cortex’s top-down regulation in decoding emotional stimuli , allowing appraisals based on affective information to moderate goal-directed behaviors .