Functional neuroanatomy of peripheral inflammatory physiology: A meta-analysis of human neuroimaging studies

Abstract

Communication between the brain and peripheral mediators of systemic inflammation is implicated in numerous psychological, behavioral, and physiological processes. Functional neuroimaging studies have identified brain regions that associate with peripheral inflammation in humans, yet there are open questions about the consistency, specificity, and network characteristics of these findings. The present systematic review provides a meta-analysis to address these questions. Multilevel kernel density analysis of 24 studies (37 statistical maps; 264 coordinates; 457 participants) revealed consistent effects in the amygdala, hippocampus, hypothalamus, striatum, insula, midbrain, and brainstem, as well as prefrontal and temporal cortices. Effects in some regions were specific to particular study designs and tasks. Spatial pattern analysis revealed significant overlap of reported effects with limbic, default mode, ventral attention, and corticostriatal networks, and co-activation analyses revealed functional ensembles encompassing the prefrontal cortex, insula, and midbrain/brainstem. Together, these results characterize brain regions and networks associated with peripheral inflammation in humans, and they provide a functional neuroanatomical reference point for future neuroimaging studies on brain-body interactions.

Keywords: Brain; Functional magnetic resonance imaging; Immunity; Inflammation; Meta-analysis; Multilevel kernel density analysis; Neuroimaging; Positron emission tomography; Stress.

Figure 1.. Flow diagram depicting study selection and screening procedures.

Figure 2.. Consistently reported activations.

Multilevel kernel density analysis identified brain regions consistently reported across included studies. Results are thresholded according to height-based and extent-based methods. Color bar describes the proportion of studies activating in a given voxel.

Figure 3.. Activations specific to study design and task type.

Left: voxel-wise chi-square analysis revealed a cluster encompassing the anterior insula that was specific to study deign. Here, inflammatory manipulation designs reported greater absolute proportions of activations than inflammatory observational designs. Right: voxel-wise chi-square analyses revealed a cluster encompassing the ventral striatum, subgenual anterior cingulate and orbitofrontal cortex that was specific to task type. Here, emotion tasks reported greater absolute proportions of activations than cognitive or resting state tasks. Results presented using p < 0.005 uncorrected threshold. Results of specificity analyses regarding clinical status are described in the supplemental material.

Figure 4.. Spatial similarity of identified brain regions to large-scale brain networks.

Top panel: Associations (mean point-biserial correlation ± standard error) describing the similarity of contrast indicator maps (CIMs) to intrinsic brain networks derived from resting-state fMRI (Yeo et al., 2011). Bottom panel: Associations describing the similarity of CIMs to striatum subdivisions (left) and corticostriatal loops (right) [from (Pauli et al., 2016)]. VS: ventral striatum; Ca: anterior caudate; Pp: posterior putamen; Pa: anterior putamen; Cp: posterior caudate.

Figure 5.. Co-activation of identified brain regions forming functionally connected ensembles.

Regions identified by the MKDA results map were grouped into functionally connected ensembles. Lines between two regions indicates they were significantly co-activated (Kendall’s tau-b) across studies. Thick lines: significant co-activation with correction for multiple comparisons, FDR q < 0.05, and not mediated by another brain region. Thin lines: significant co-activation, p < 0.05 uncorrected.