Studies in rodents support a central role for afferent interoceptive fibers traveling with the vagus nerve in the motivational reorientation associated with peripheral inflammation. To our knowledge this is the first study that points to a similar mechanism for sickness behavior in humans. It is known that neurons within discrete brain regions express cytokine receptors, including medial thalamic nuclei, ventromedial hypothalamus, basolateral amygdala, and cerebellar Purkinje cells (37
). Therefore, one possibility is that the central neural responses we observed because of typhoid vaccination represent a direct effect of circulating cytokines on brain or even the local synthesis of cytokines as part of a systemic immune response. However, against such an interpretation is evidence that increases in cytokine levels within neural tissue after peripheral inflammatory challenge typically occur approximately 8 hours after inflammatory challenge (and have not been shown to occur within 3 hours) (38
). Moreover, this central cytokine expression is modulated by earlier vagally mediated mechanisms (39,40
). Instead, we suggest that interoceptive signaling provides the most parsimonious account for the observed activity change within interoceptive (and cognitive) regions.
At the conservative threshold reported here we observed inflammation-dependent activity within the right medial thalamus, encompassing the mediodorsal nucleus. This nucleus, part of the interoceptive lamina I pathway, projects to cingulate and prefrontal cortex. More lateral thalamic activity consistent with activations within the basal and posterior ventromedial nuclei (VMb and VMpo), which project to dorsal mid/posterior insula, was evident at a less stringent threshold (p
< .001 uncorrected). Inflammation-associated activity in both bilateral dorsal-mid and anterior insula areas, however, was also seen at the more stringent threshold. The location of these activations is noteworthy. The VMb, which receives predominantly vagal fibers, and VMpo (predominantly sympathetic) project to dorsal mid/posterior insula in a rostrocaudal topographic manner, with vagal projections extending more rostrally. It is interesting therefore that our activations occur in a more rostral region of interoceptive cortex than those reported for thermal sensation (18
), noxious pain, and itch (non-vagal) but close to the activation reported by Rosenkranz et al.
) in response to antigen-induced inflammatory airways response in asthmatic subjects, mediated via afferents traveling with the vagus.
Direct stimulation and deactivation studies in humans (42
) and animals (43
), however, show that insula activity is associated with efferent autonomic change. To isolate efferent from afferent effects of inflammation, we used inflammation- and placebo-associated changes in blood pressure as co-regressors in our analysis. This confirmed that insula activity remained significant after accommodating changes in efferent autonomic activity. Blood pressure change, however, was highly correlated with dorsal pontine activity in a region encompassing the parabrachial nuclei, with much of the variance in this region after inflammation accounted for by changes in efferent autonomic activity. Interestingly, although playing a role in brainstem homeostatic reflexes including regulation of blood pressure, the parabrachial nuclei do not form part of human lamina I spinothalamocortical interoceptive pathways projecting to dorsal insula (17
The interaction between inflammation and the enhanced cognitive demands of processing incongruent (vs. congruent) stimuli in the Stroop task permitted us to investigate the neural basis of cognitive concomitants of sickness behavior. Notably, we observed a more widespread brain engagement when processing high-demand incongruent trials with recruitment of dorsolateral prefrontal and midcingulate (aMCC/pMCC) cortices being expressed during vaccine-induced inflammation compared with the placebo condition (D). Because participants showed no performance differences due to inflammation, these effects are most likely attributable to a requirement for additional neural resources to maintain equivalent task-performance under inflammation. Activity within both dorsolateral prefrontal and dorsal anterior cingulate cortices is typically enhanced with increasing cognitive demands. Both regions also show an increase in activity during performance of a cognitively demanding visual task in the face of cross-modal auditory distracters (36
). Their concurrent activation therefore suggests that interdependent cognitive/attentional (44
) and somatic (autonomic) (45
) mechanisms are invoked to maintain performance and compensate for compromised fluency in cognitive processes in the face of increased conflict from interoceptive processing demands during inflammatory states.
Analysis of the neural basis of inflammation- and placebo-associated changes in fatigue and confusion also highlighted a key role for insula and cingulate cortices in mediating subjective symptoms after peripheral inflammation. Inflammation-associated but not placebo-associated fatigue was predicted by activity change within mid/posterior insula bilaterally and right anterior cingulate (pACC/aMCC). These findings suggest a degree of specificity of neural mechanisms underlying fatigue associated with peripheral inflammation that is not seen in more general placebo-associated fatigue (which might result from more heterogeneous mechanisms). Previous studies showing insula responses to graded cooling (18
), itch (46
), and intensity of dynamic exercise (47
) support our hypothesis that inflammation-associated fatigue results from a similar insula-based interoceptive mechanism. Interestingly, this cingulate region is also activated by pain (48,49
) and, in particular, pain induced by visceral more than somatosensory stimulation (50
). In this context pACC/aMCC activity shows greater correlation with the reported unpleasantness than the intensity of the stimulus, suggesting a role in emotional rather than somatosensory localization components of pain (51
). This role might also be supported by its strong amygdala connectivity. Its tight correlation with inflammation-associated fatigue across subjects in our present study suggests a potentially broader role for this cingulate sub-region in processing emotional components of unpleasant visceral stimuli.
Of note, there were no significant correlations between reported activations and peripheral IL-6 levels (at the stringent thresholds adopted throughout the current study) and, furthermore, no correlation between IL-6 levels and subjects' subjective symptom ratings. This finding is noteworthy, especially when compared with our previous finding of a correlation among interindividual susceptibility to inflammation-associated motor slowing, peripheral cytokine (IL-6) levels, and substantia nigra activity in this group (26
). This finding suggests that, rather than reflecting a simple index of peripheral cytokine levels, subjective symptoms after peripheral inflammation likely result from a complex interplay between bottom-up processes (associated with peripheral inflammation) and top-down modulation of these processes and resultant subjective symptoms. Furthermore, although IL-6 is a useful index of inflammation and functions in an endocrine manner to coordinate inflammatory responses, it acts as part of an interacting network of pro- and anti-inflammatory cytokines. This complex relationship with other inflammatory cytokines, in particular IL-1β (which also acts on vagal afferents), suggests that a simple relationship between IL-6 levels and neural responses might be the exception rather than the rule.
To conclude, our empiric study provides a general framework for how peripheral states of inflammation (typically related to systemic infection) modulate attentional and motivational brain systems. It complements our previous study that implicated changes in substantia nigra activity in low-level psychomotor responses to inflammation (26
) and our second article under consideration in this journal (23
), suggesting that inflammation modulates mood via an effect on Cg25 connectivity, a region implicated in the pathogenesis of primary depression. We acknowledge that fMRI is primarily a correlative technique and that our experiment cannot prove that inflammation directly causes the pattern of activation seen and, furthermore, that our findings in a group of healthy young men might not extrapolate to other ages, gender, or health status. Also, noteworthy in this regard is the literature showing that chronic low-grade systemic inflammation, with minor increases in circulating inflammatory mediators, is associated with age-related cognitive decline (52
) and dementia (53
). Similarly, the therapeutic use of cytokines (in particular IL-2 and IFN-α) in the treatment of malignancies and chronic viral infections is frequently complicated by cognitive, emotional, and behavioral disturbance (54
). Sickness behaviors themselves can lead to changes in immune competence by modulating host sensitivity to pathogens and influencing the degree of activation of the innate immune system and the production of peripheral cytokines. Bidirectional interactions between immune and psychological states ultimately determine the degree of wellbeing (4
The main implication of the data we present is that they highlight a neurobiological substrate for the central representation of peripheral inflammation in humans and suggest how these representations might interact with the resource requirements of complex cognitive processing. Our findings indicate the neural mechanisms underpinning interindividual susceptibility to psychological symptoms associated with inflammation-induced sickness and have broader clinical implications across medical specialties in which these common symptomatologies might be targeted at the level of the central nervous system.