The human brainstem's red nucleus was upgraded to support goal-directed action

Abstract

The red nucleus, a large brainstem structure, coordinates limb movement for locomotion in quadrupedal animals. In humans, its pattern of anatomical connectivity differs from that of quadrupeds, suggesting a different purpose. Here, we apply our most advanced resting-state functional connectivity based precision functional mapping in highly sampled individuals (n = 5), resting-state functional connectivity in large group-averaged datasets (combined n ~ 45,000), and task based analysis of reward, motor, and action related contrasts from group-averaged datasets (n > 1000) and meta-analyses (n > 14,000 studies) to precisely examine red nucleus function. Notably, red nucleus functional connectivity with motor-effector networks (somatomotor hand, foot, and mouth) is minimal. Instead, connectivity is strongest to the action-mode and salience networks, which are important for action/cognitive control and reward/motivated behavior. Consistent with this, the red nucleus responds to motor planning more than to actual movement, while also responding to rewards. Our results suggest the human red nucleus implements goal-directed behavior by integrating behavioral valence and action plans instead of serving a pure motor-effector function.

Fig. 1. Functional connectivity mapping of the red nucleus.

a Axial (top) and coronal (bottom) display of the right red nucleus (white outline) overlaid on a T2w structural image for subject PFM-Nico. b Resting state functional connectivity (RSFC) seeded from the right red nucleus in an exemplar highly sampled participant with multi-echo independent component analysis (MEICA) denoising (PFM-Nico; 134 min resting-state fMRI). Individual specific functional connectivity map shows strongest 20 percent of cortical vertices. Bar graph quantifies the average connectivity per network. The average connectivity was significantly different from zero for the salience, action-mode (AMN) and dorsal attention (DAN) networks (two-sided t-test against null distribution, *P < 0.05, Bonferroni correction, uncorrected p-value = 0.000999), but was only positive for salience and AMN. c Group-averaged functional connectivity map shows strongest 20 percent of cortical vertices using previously defined split-halves (see ref. ; n = 1964 participants each) from the Adolescent Brain Cognitive Development (ABCD) study. For additional participants see Supplementary Figs. 1–4. AMN action-mode, SCAN somato-cognitive action, le-SMN lower-extremity somatomotor, ue-SMN upper-extremity somatomotor, f-SMN face somatomotor, PMN posterior memory, CAN contextual association, DMN default mode, FPN fronto-parietal, DAN dorsal attention, VAN ventral attention.

Fig. 2. Thalamic connectivity of the red nucleus.

Top 20% of red nucleus connections for the thalamus (MNI space) for (a), PFM-Nico, (b), ABCD study (n = 3,928), (c), HCP (n = 812), and (d), UKB (n = 4000). Four different axial slices of the thalamus are shown (MNI space) overlaid on the subject’s structural image (panel a only). Thresholding is based on the top 20% of connections for the thalamus. The VIM (ventral intermediate) nucleus of the thalamus defined in an individual subject is shown in panel (a). The nucleus outline based on a dilated probabilistic map using the THOMAS atlas shown in panels (b–d). See Supplementary Fig. 6 for additional participants.

Fig. 3. Reward and motor control activity in the red nucleus.

a Red nucleus activation from the gambling task of Human Connectome Project (HCP) reward (left, filled light gray) and punishment (right, unfilled) contrasts (paired t-test; t = 4.87, confidence interval = [0.046 0.11], cohen’s d d = 0.15, df = 1080, p = 1.29 × 10^-6). b Automated meta-analysis average for term ‘reward’ for the red nucleus (left, stripe pattern) the region surrounding the red nucleus (peri-nucleus, middle, circle pattern) and whole brain (right, crosshatch pattern). c Red nucleus activation from the motor task of HCP motor cue (left, filled dark grey) and average movement (right, unfilled) contrasts (paired t-test; t = 4.58, confidence interval = [0.04 0.1], cohen’s d d = 0.14, df=1079, p = 5.22 × 10^-6). d Automated meta-analysis average for term ‘motor control’ for the red nucleus (left, stripe pattern) the region surrounding the red nucleus (peri-nucleus, middle, circle pattern) and whole brain (right, crosshatch pattern). Boxplot displays interquartile range, median (horizontal black line), mean (red circle), and 95% confidence interval (black error bars).

Fig. 4. Action versus motor-effector network assignments of red nucleus voxels.

Red nucleus voxels were assigned to networks using winner-take-all logic. The percentage of red nucleus voxels assigned to action related networks (salience network, AMN, or SCAN) is shown in filled circles and the percent assigned to any of the three motor-effector networks (lower-extremity somatomotor, upper-extremity somatomotor, and face somatomotor) is shown in hollow triangles. The left three columns show group-averaged data; the right five columns showing data from highly sampled individuals. AMN (action-mode), SCAN (somato-cognitive action).

Fig. 5. Functional connectivity subdivisions of the red nucleus.

a Anatomical display of dorsal-medial (hatched) and ventral-lateral (no fill) red nucleus subdivisions in an exemplar participant (PFM-Nico) overlaid on the same participant’s T2w image. b Strongest 20 percent of cortical RSFC for ventral-lateral (left) and dorsal-medial (middle) red nucleus subdivisions. The right most image shows the difference map between these two connectivity maps. Note greater ventral-lateral connectivity in red and greater dorsal-medial in blue. c Average cortical RSFC organized by network for dorsal-medial (hatched) and ventral-lateral (no fill) subdivisions. d Similarity (r) in network connectivity for each red nucleus voxel grouped into dorsal-medial and ventral-lateral divisions. For additional subjects/analyses see Supplementary Figs. 16–19. AMN (action-mode) SCAN (somato-cognitive action), le-SMN (lower-extremity somatomotor), ue-SMN (upper-extremity somatomotor), f-SMN (face somatomotor), PMN (posterior memory), CAN (contextual association), DMN (default mode), FPN (fronto-parietal), DAN (dorsal attention), VAN (ventral attention); Resting state functional connectivity (RSFC).