Tasks activating the default mode network map multiple functional systems

Recent developments in network neuroscience suggest reconsidering what we thought we knew about the default mode network (DMN). Although this network has always been seen as unitary and associated with the resting state, a new deconstructive line of research is pointing out that the DMN could be divided into multiple subsystems supporting different functions. By now, it is well known that the DMN is not only deactivated by tasks, but also involved in affective, mnestic, and social paradigms, among others. Nonetheless, it is starting to become clear that the array of activities in which it is involved, might also be extended to more extrinsic functions. The present meta-analytic study is meant to push this boundary a bit further. The BrainMap database was searched for all experimental paradigms activating the DMN, and their activation likelihood estimation maps were then computed. An additional map of task-induced deactivations was also created. A multidimensional scaling indicated that such maps could be arranged along an anatomo-psychological gradient, which goes from midline core activations, associated with the most internal functions, to that of lateral cortices, involved in more external tasks. Further multivariate investigations suggested that such extrinsic mode is especially related to reward, semantic, and emotional functions. However, an important finding was that the various activation maps were often different from the canonical representation of the resting-state DMN, sometimes overlapping with it only in some peripheral nodes, and including external regions such as the insula. Altogether, our findings suggest that the intrinsic–extrinsic opposition may be better understood in the form of a continuous scale, rather than a dichotomy. Supplementary Information The online version contains supplementary material available at 10.1007/s00429-022-02467-0.

: Surface mapping of the three DMN masks used, their union, their intersection, and the result of selecting the voxels shared by at least two masks. Figure S2: PRISMA flow chart of the selection of studies. Figure S3:Surface mapping of the Activation Likelihood Estimation maps obtained with the Fail-safe procedure with 6% of file-drawer effect. Figure S4: Surface mapping of the Activation Likelihood Estimation maps obtained with the Fail-safe procedure with 60% of file-drawer effect. Figure S5: Volume mapping of the overlaps between the ALE maps and the three y = 14,episodic recall,self reflection = 6; Figure S7:Two-dimensional scaling graph of the main analysis. The volume maps are centered on the MDS coordinates. Figure S8: Two-dimensional and three-dimensional scaling graphs calculated excluding the condition Self-Reflection from the analysis.

Testing the MDS interpretation
We interpreted the MDS results in terms of a lateral-medial axis and a dorsal-ventral axis. This interpretation was partly justified by the fact that our distances were calculated from the Pearson correlations between the maps, and thus they are informed only by their spatial differences. However, this does not necessarily imply that these differences should be arranged along the major brain axes. Also, although not readily apparent to us, it might be that the third MDS axis could be explained by rostro-caudal activations. We tested these hypotheses with the following procedure. First, we estimated the laterality, dorsality and rostrality of our ALE maps. An index of laterality L (not to be confused with the lateralization index LI [Desmond et al., 1995] that is often incorrectly called laterality index) was calculated as: where v i stands for the ALE value of a given voxel i, and x i is the x coordinate of voxel i, taken in absolute value. Note that this formula is extremely similar to the calculation of the Talairach x coordinate of the image's weighted center of mass. In practice, the value of each voxel is weighted by its x coordinate, and their average (or, simplifying, their sum) is normalized by the average (or sum) of the unweighted map.
Taking the x coordinate in absolute value has the effect of weighting the lateral voxel more than the central ones. L could be seen as the distance (in mm) of the weighted center of mass from the midline, if the left hemisphere was flipped over the right one. Similarly, the indices of dorsality and rostrality D and R were computed as: where z i and y i are the z and y coordinates of voxel i, respectively. These are simply the z and y Talairach coordinates of the weighted center of mass of the map.
Then, we correlated the coordinate assumed by an ALE map on a given 3D MDS axis with the index that corresponded to our hypotheses. We hypothesized that the MDS x coordinates correlated (negatively) with L, and that the MDS y coordinates correlated (positively) with D. We also explored the possibility that the MDS z axis could be correlated with R even if we were not able to see this relation in our results.
As hypothesized, the first MDS axis was correlated with the laterality of the maps with r = -0.6 (p = 0.045, one-tailed t-test). Some observations did not fit our interpretation perfectly. In particular the Theory of Mind map, despite its obvious medial activations (and its longstanding association with the canonical DMN) scored more lateral than expected because of its strong activations in the angular gyrus. Furthermore, Reward was found in the MDS axis closer to the lateral maps despite showing large medial activations, possibly due to its marked insular involvement and its medial activations in non-DMN, salience areas (supplementary motor area). Thus, the placement of the maps on the MDS x axis was mostly consistent with our hypothesis of an anatomical lateral-medial axis, and even its most obvious outlier fit an internal-external psychological description. In fact, Theory of Mind is considered an internal function, while the Salience Network activations of Reward suggest a more external mode of cognition.
Our second hypothesis was that the MDS y axis could be explained by a dorsal-ventral arrangement of activations. Indeed, Pearson's correlation between the MDS y axis and the dorsality index D was r = 0.93 (p < 0.001). Lastly, we tested if the third MDS axis was correlated with rostrality R. As expected, the two variables were not correlated, with r = 0.06 (p = 0.44). See Table S3 and Figure S6 for further details.

Testing the heteromodality of the principal gradients
To evaluate if the third principal gradient could be considered a map of heteromodal areas, we replicated the meta-analysis made by Margulies et al. (2016), using the Behavioral Analysis toolbox (Lancaster et al., 2012).
The given PG, converted into Talairach space, was divided in 20 ROIs including 5% of the map values distribution (e.g., from 0 to the 5 th per-centile, from the 5 th to the 10 th percentile, etc.), and each ROI was fed to the Behavioral Analysis plugin for Mango. The Behavioral Analysis is akin to the Paradigm Analysis (see Methods), but works on behavioral domains instead of experimental paradigms. The cognitive profile of the ROI is represented by the z-scores obtained by such analysis for the 60 behavioral categories, and it is illustrated in Fig. S14. Figure S14: Behavioral Analyses z-scores of the ROIs containing the map distribution quintiles (QU) for principal gradient 1 and 3 by Margulies et al. (2016).