An inflammatory pathway links atherosclerotic cardiovascular disease risk to neural activity evoked by the cognitive regulation of emotion

Abstract

Background

Cognitive reappraisal is a form of emotion regulation that alters emotional responding by changing the meaning of emotional stimuli. Reappraisal engages regions of the prefrontal cortex that support multiple functions, including visceral control functions implicated in regulating the immune system. Immune activity plays a role in the preclinical pathophysiology of atherosclerotic cardiovascular disease (CVD), an inflammatory condition that is highly comorbid with affective disorders characterized by problems with emotion regulation. Here, we tested whether prefrontal engagement by reappraisal would be associated with atherosclerotic CVD risk and whether this association would be mediated by inflammatory activity.

Methods

Community volunteers (n=157; aged 30–54; 80 women) without DSM-IV Axis-1 psychiatric diagnoses or cardiovascular or immune disorders performed a functional neuroimaging task involving the reappraisal of negative emotional stimuli. Carotid artery intima-media thickness and inter-adventitial diameter were measured by ultrasonography and used as markers of preclinical atherosclerosis. Also measured were circulating levels of interleukin-6 (IL-6), an inflammatory cytokine linked to CVD risk and prefrontal neural activity.

Results

Greater reappraisal-related engagement of the dorsal anterior cingulate cortex (dACC) was associated with greater preclinical atherosclerosis and IL-6. Moreover, IL-6 mediated the association of dACC engagement with preclinical atherosclerosis. These results were independent of age, sex, race, smoking status, and other known CVD risk factors.

Conclusions

The cognitive regulation of emotion may relate to CVD risk through a pathway involving the functional interplay between the anterior cingulate region of the prefrontal cortex and inflammatory activity.

Cardiovascular disease (CVD) is the primary cause of premature death in developed nations (1). Angina, myocardial infarction, and other clinical outcomes of CVD result from atherosclerosis, a progressive inflammatory condition that contributes to ‘clinically-silent’ or preclinical changes in arterial morphology (2). These preclinical changes unfold over a long period of time, and they appear decades before obstructive or unstable atherosclerotic plaques that precipitate late-stage CVD outcomes (3). The severity and progression of preclinical atherosclerosis and consequent risk for clinical CVD outcomes, however, differ between individuals as a function of biological and behavioral risk factors.

Epidemiological evidence suggests that a specific source of biobehavioral risk for atherosclerotic CVD relates to the regulation of negative affect (4). Hence, individuals with mood and anxiety disorders that are characterized by problems with the regulation of negative affect are at disproportionate risk for developing CVD (5–8). Moreover, otherwise healthy individuals who exhibit dysregulated negative affect are at risk for showing an accelerated progression of preclinical atherosclerosis and developing clinical CVD (5, 9–11). Despite such epidemiological evidence, however, the pathways linking CVD risk to the regulation of affect in general and to the regulation of negative affect in particular are unclear (4).

Affect or emotion regulation can be defined as “…how we try to influence which emotions we have, when we have them, and how we experience and express these emotions” (page 275 in Gross (12)). One adaptive form of emotion regulation is cognitive reappraisal, which entails the restructuring of the meaning of an emotional event in a way that changes emotional responding (13). Individuals differ in their use of reappraisal to regulate emotional responding in daily life (14), as well as in their capacity to use reappraisal effectively in evocative experimental protocols (15). It is well established that reappraisal relates inversely across individuals to indicators of negative affect (12, 14, 16). Further, individual differences in reappraisal have been implicated specifically in vulnerability to affective disorders that confer CVD risk (4). Thus, individual differences in the regulation of negative affect by reappraisal may plausibly relate to CVD risk, possibly via pathways related to atherogenesis.

To elaborate, reappraisal appears to engage regions of the prefrontal cortex, encompassing areas of the anterior cingulate, medial, dorsolateral, and ventrolateral prefrontal cortices (17–22). In extension, some of these same prefrontal regions are components of so-called visceral control circuits that coordinate neuroendocrine and autonomic outflow with cognitive and affective processes via efferent and afferent neuroanatomical projections. More precisely, subgenual, perigenual, and dorsal areas of the anterior cingulate cortex (ACC), as well as the ventromedial prefrontal cortex (vmPFC), exhibit direct and indirect visceromotor projections to subcortical cell groups that govern the release of neurohormones and neurotransmitters of the hypothalamic-pituitary-adrenal axis and sympathetic and parasympathetic limbs of the autonomic nervous system (23–32). Prefrontal visceral control is also enabled by homeostatic and afferent feedback from the periphery, as relayed by direct and indirect projections from viscerosensory regions (e.g., area postrema and solitary tract nucleus) that ultimately target networked prefrontal areas (33–35), as well as another frontal region implicated in affect regulation and peripheral physiological regulation: the anterior insula (19, 36–40). Presumably supported by this anatomical organization, parameters of neuroendocrine and autonomic physiology have been associated with functional variation in prefrontal, cingulate, and insular regions, as evoked in the context of reappraisal and other cognitive and affective behavioral paradigms (22, 41–43). Likewise, the functional engagement of prefrontal, cingulate, and insular regions has been associated with markers of systemic inflammation, which is putatively mediated via intermediate neuroendocrine-immune and autonomic-immune communication pathways (44, 45, 46; see Discussion).

Notwithstanding evidence suggesting (i) an association between emotion regulation and CVD risk and (ii) prefrontal, cingulate, and insular involvement in emotion regulation by reappraisal, as well as in visceral control functions important for inflammatory regulation, it is unknown whether neural engagement within prefrontal, cingulate, and insular regions by reappraisal associates with atherosclerotic CVD risk or whether such an association might be mediated by an immune marker of inflammatory activity. Accordingly, we used functional magnetic resonance imaging (fMRI) to test whether prefrontal, cingulate, or insular neural activity measured during the reappraisal of negative affective stimuli would covary with an indicator of CVD risk assessed by arterial measurements of preclinical atherosclerosis. We next tested whether reappraisal-related neural activity would covary with circulating interleukin (IL)-6, an inflammatory cytokine linked to dysregulated affect (47), preclinical atherosclerosis (48), and CVD outcomes (49, 50). Finally, we tested whether IL-6 would mediate any associations between reappraisal-related neural activity and preclinical atherosclerosis.

MATERIALS AND METHODS

Participants

Participants were 95 men and 88 women (aged 30–54 years; 162 Caucasian; 20 African American; 2 multiracial/other ethnicities) from the Adult Health and Behavior project, phase-II (AHAB-II). AHAB-II is an epidemiological registry of biobehavioral correlates of CVD risk among community-dwelling adults. Supplement I provides recruitment, eligibility, and other details about AHAB-II and the present sample. All participants provided informed consent. The University of Pittsburgh Institutional Review Board granted study approval. Table 1 provides sample characteristics.

Table 1.

Summary of participant characteristics, N = 157 (77 men, 80 women)

Characteristic	Mean or N	SD
Age (years)	42.7	7.3
Educational attainment
HS/GED/Technical Training/Some College	29
Associate Degree	11
Bachelors	60
Masters Degree or higher	57
Body mass index (kg/m2)	27.5	5.4
Waist circumference (inches)	36.4	5.8
High density lipoproteins (mg/dl)	54.8	14.5
Glucose (mmol/L)	98.9	10.8
Total cholesterol (mg/L)	201.0	36.9
Triglycerides (mg/dl)	114.1	75.1
Seated Resting SBP (mmHg)	114.8	11.6
Seated Resting DBP (mmHg)	72.9	8.2
Continuous Metabolic Risk Score (z-units)	1.3	3.8
Interleukin-6 (pg/mL)	1.2	1.1
Carotid artery intima media thickness (mm)	0.6	0.1
Carotid artery adventitial diameter (mm)	6.8	0.6
Preclinical Atherosclerotic Risk Score (z-units)	0.0	0.9
Smoking Status
Never	95
Former	33
Current	29

Note. HS = High school; GED = General equivalency degree (or diploma); Continuous Metabolic Risk Score = standardized average of metabolic syndrome components (see Methods and Materials); Preclinical Atherosclerotic Risk Score = standardized average of carotid artery intima media thickness and adventitial diameter.

Protocol and measures

IL-6

Participants underwent medical and psychiatric interviews and phlebotomy to assess blood-derived variables. Participants also provided measurements of blood pressure, waist circumference, height and weight. Circulating IL-6 was determined using methods detailed in Supplement I. Natural-log transformation was applied to correct the skew of IL-6 values.

Preclinical atherosclerosis

Two indicators of carotid artery morphology were measured to assess preclinical atherosclerosis: intima-media thickness (IMT) and inter-adventitial diameter (IAD). IMT and IAD predict CVD outcomes (51, 52), especially when considered together (53). IMT and IAD also increase during atherogenesis (54, 55) and associate with CVD risk factors (56–59). Raw IMT and IAD values were correlated (Spearman’s rho = 0.503, p < 0.001). Natural-log transformation was applied to correct the skew of both measures. To compute a continuous measure of preclinical atherosclerosis, log-transformed IMT and IAD values were z-scored and averaged. Henceforth, this measure is referred to as a preclinical atherosclerotic risk (PAR) score. (Details in Supplement I)

Covariates

The following measures were selected as a priori covariates because they may plausibly confound study findings: age, gender, race, educational attainment, smoking status (non-smoker, former smoker, and current smoker), as well as six additional factors that (i) comprise the metabolic syndrome and (ii) when combined associate strongly with CVD risk (60, 61). These were systolic blood pressure, waist circumference, and fasting glucose, high-density lipoprotein (HDL) levels, total cholesterol, and triglycerides. From these factors, a cardio-metabolic risk (CMR) score was computed by z-scoring each factor and averaging the z-scores (HDL signs reversed). Thus, six covariates were used: age, sex, race, educational attainment, smoking status, and the CMR score. (Details in Supplement I)

Reappraisal task

The reappraisal task was administered in an event-related fMRI paradigm developed by Ochsner and colleagues (62, 63). Participants were told that they would view negative and neutral images on a display after one of two cues. The first was the “Look” cue, for which they were instructed to maintain their attention on the image and allow their emotional reaction to occur as it naturally would. Following the nomenclature of Gyurak and colleagues (64), these are referred to as reactivity trials. The other cue was the “Decrease” cue, for which they were instructed to change the way they thought about the image to feel less negative. These are referred to as regulation trials (64). This task permits contrasts between blood-oxygenation level-dependent (BOLD) signal changes evoked by (i) the negative valence of the stimuli (Look Negative vs. Look Neutral) and (ii) reappraising negative stimuli (Regulate Negative vs. Look Negative). (Details in Supplement I)

MRI

MRI data were collected on a 3T Trio TIM scanner (Siemens, Erlangen, Germany) using a 12–channel phased-array head coil. Acquired MRI data were preprocessed and analyzed using methods detailed in Supplement I. After preprocessing, BOLD responses from regions-of-interest (ROIs) were extracted for mediation testing. In primary whole-brain and ROI analyses, we employed a corrected false-positive detection rate (FDR) threshold of 0.05 and cluster threshold of k ≥ 10 contiguous voxels. (Details further in Supplement I)

ROI analyses

We selected these ACC areas for a priori hypothesis testing of reappraisal effects (i.e., by the Regulate Negative vs. Look Negative contrast): subgenual ACC (sgACC), perigenual ACC (pgACC), and dorsal ACC (dACC). Also selected were the vmPFC, dorsolateral prefrontal cortex (DLPFC), ventrolateral prefrontal cortex (vlPFC), and anterior insula (AI). Ancillary analyses tested these ROIs and the amygdala to explore the effects of reactivity-related activity (i.e., revealed by the Look Negative > Look Neutral contrast) in relation to IL-6 and PAR scores. (Details in Supplement I)

Extraction of ROI responses for mediation testing

We tested whether ROI BOLD responses were associated with IL-6 and PAR scores and whether IL-6 mediated any associations between ROI responses and PAR scores using an a priori BOLD response extraction strategy. BOLD contrast parameter estimates were thus extracted for voxels defining clusters within the ROIs that exhibited condition-specific effects at FDR ≤ 0.05 and k ≥ 10 using MarsBaR (http://marsbar.sourceforge.net). Extracted parameter estimates were then averaged within each ROI and imported into SPSS (v.20, IBM Corp., Armonk, NY) for statistically independent mediation testing.

Mediation testing

Path analyses tested whether extracted ROI BOLD responses (X) were associated with PAR scores (Y), and whether any observed associations were mediated by IL-6 (M). Associations of ROI responses with IL-6 were tested as the effect of X on M (X→M), corresponding to the a Path. Associations of IL-6 with PAR scores controlling for ROI responses were tested as the effect of M on Y (M→Y), corresponding to the b Path. Associations reflecting the total effects of ROI responses on PAR scores without controlling for IL-6 were tested as the unadjusted effects of X on Y (X→Y) or c Path. Associations reflecting the direct effects of ROI responses on PAR scores while controlling for IL-6 were tested as the adjusted effects of X on Y (X→Y) or c′ Path. Indirect path effects reflecting the association of ROI responses and PAR scores—as mediated by IL-6 (X→M→Y)—were tested as the product of Paths a and b (a*b) (65). Statistical testing of mediation was done by nonparametric bootstrapping (5000 iterations), with 95% confidence intervals (CIs) for indirect (a*b) mediation effects generated by the bias-corrected method. Mediation modeling was executed in the SPSS macro, PROCESS (65, 66). All models included the a priori covariates described above. (Details in Supplement I)

RESULTS

Task effects on negative affect

Negative images on reactivity trials evoked more negative affect than neutral images, as reflected by self-reports on a 1-to-5 scale (M = 3.5 ± SE = 0.06 vs. M = 1.13 ± 0.02, respectively, paired t[156] = 41.01, p < 0.001). Negative images on regulation trials (M = 2.78 ± 0.05) also evoked less negative affect than negative images on reactivity trials (paired t[156] = 12.86, p < 0.001). These self-report results verified expected task effects on negative affect.

Task effects on BOLD responses

Look Negative > Look Neutral contrast

Negative images on reactivity trials evoked increased BOLD activity compared to neutral images in all ROIs, including sgACC, pgACC, dACC, vmPFC, vlPFC, DLPFC, AI, and amygdala, FDR ≤ 0.05 and k ≥ 10 voxels (Table S1, Figure S2). No decreases in BOLD activity were observed in these ROIs for the Look Negative < Look Neutral contrast.

Regulate Negative vs. Look Negative contrast

Negative images on regulation trials evoked increased BOLD activity compared with negative images on reactivity trials in the pgACC, dACC, vlPFC, DLPFC, and AI at FDR ≤ 0.05 and k ≥ 10 voxels (Table S2, Figure S3). Negative images on regulation trials also evoked decreased BOLD activity on reactivity trials in the sgACC and vmPFC at FDR ≤ 0.05 and k ≥ 10 voxels (Table S2, Figure S3). No amygdala BOLD signal changes were detected with FDR thresholding; however, a more lenient threshold (p < 0.05_uncorrected, k ≥ 10 voxels) revealed decreased amygdala activity on regulation trials (Figure S4), apparently consistent with cumulative evidence (20). For exploratory purposes, BOLD activity changes were extracted from the amygdala, but interpreted with caution (see Discussion). For completeness of reporting, whole-brain effects are in Tables S3 and Figures S5–S6.

Association of IL-6 and preclinical atherosclerosis

IL-6 positively correlated with PAR scores (r = 0.34, p < 0.001), which persisted in a partial correlation adjusting for age, sex, race, education, smoking status, and CMR scores, pr = 0.24, p = 0.003. These results support testing IL-6 as a mediator between ROI responses and preclinical atherosclerosis.

Mediation testing of reappraisal-related ROI responses, IL-6, and preclinical atherosclerosis

Reappraisal-related BOLD responses in the pgACC, dACC, vmPFC, vlPFC, and AI, but not sgACC, DLPFC, or amygdala, were positively associated with IL-6, a Path coefficients > 0.27, ps < 0.05 (Table 2). IL-6 was also positively associated with PAR scores in all mediation models, b Path coefficients > 0.38, ps < 0.05. However, only reappraisal-related BOLD responses in the dACC and pgACC showed total and direct associations with PAR scores, with c and c′ Path coefficients > 0.48, ps < 0.05, respectively. Finally, the 95% CI of the indirect (a*b) effect of reappraisal-related pgACC responses on PAR scores—as mediated by IL-6—included 0 (−0.007 to 0.44). By contrast, the same indirect-effect CI of reappraisal-related dACC responses on PAR scores through IL-6 did not include 0 (0.0032 to 0.43)—indicating mediation (Figure 1).

Table 2.

Summary of mediation models predicting preclinical atherosclerosis with IL-6 as the mediator.

Regions of Interest	Total effect c Point est. (SE)	Direct effect c′ Point est. (SE)	Path a Point est. (SE)	Path b Point est. (SE)	Indirect effect a* b
					Point est. (SE)	95% CI
dACC	.613* (.243)	.484* (.244)	.333** (.128)	.387* (.153)	.129* (.108)	[.003, .429]
sgACC	.391 (.220)	.351 (.216)	.094 (.117)	.434*** (.151)	.041 (.052)	[−.033, .185]
pgACC	.615** (.236)	.499* (.237)	.301* (.125)	.389* (.152)	.117 (.109)	[−.007, .435]
vmPFC	.462 (.275)	.314 (.275)	.355* (.143)	.415** (.154)	.147 (.123)	[−.003, .475]
amygdala	.486 (.262)	.380 (.259)	.254 (.138)	.417 (.152)**	.106 (.098)	[−.008, .385]
DLPFC	.518 (.282)	.424 (.278)	.222 (.149)	.423** (.152)	.094 (.116)	[−.028, .415]
vlPFC	.344 (.246)	.219 (.245)	.294* (.128)	.425** (.154)	.125 (.105)	[−.003, .419]
AI	.345 (.246)	.228 (.245)	.274* (.129)	.426** (.153)	.117 (.096)	[−.007, .383]

Note.

^*p<.05.

^**p<.01.

^***p<.005

Point est. (SE) = unstandardized point estimate with the corresponding standard errors in parentheses.

95% CI = Bias-corrected 95% confidence interval.

dACC, dorsal anterior cingulate cortex (ACC); sgACC, subgenual ACC; pgACC, perigenual ACC; vmPFC, ventromedial prefrontal cortex (PFC); DLPFC, dorsolateral PFC; vlPFC, ventrolateral PFC; AI, anterior insula (see Figure S1 for details regarding regions of interest).

Figure 1.

(A) Path model summarizing the association of reappraisal-related activity in the dorsal anterior cingulate cortex (dACC) and a preclinical atherosclerosis risk (PAR) score (see methods), as mediated by IL-6. In panel A, the purple shaded portion of a dACC region of interest (ROI) mask (illustrated in blue) corresponds to the cluster of voxels showing significant activation in the Regulate Negative > Look Negative contrast at a FDR threshold of 0.05 and cluster extend threshold of 10 voxels. The values correspond to each coefficient for the indirect (a, b), direct (c′), and total (c) paths. In the mediation model, the following variables were included as covariates: age, sex, race, educational attainment, smoking status, and a cardio-metabolic risk score (see methods). (B) Distribution of 5000 bootstrap samples of the indirect (a*b) effect for the mediation results shown in A. The grey shaded area of the distribution encompasses a*b indirect effects falling within a 95% confidence interval. (C) Log-transformed IL-6 values shown as a function of reappraisal-related dACC activity. (D) PAR scores shown as a function of reappraisal-related dACC activity. (E) PAR scores shown as a function of log-transformed IL-6. The scatter plots shown in C–E are unadjusted for any covariates for illustration purposes.

Exploratory testing of reactivity-related ROI responses, IL-6, and preclinical atherosclerosis

As noted, the Look Negative > Look Neutral contrast revealed increased BOLD activity in all ROIs, raising the question of whether ‘reactivity-related’ activity would associate with IL-6 and PAR scores in mediation models. Only reactivity-related pgACC responses were associated with IL-6. And only reactivity-related sgACC responses exhibited a direct association with PAR scores. Because no reactivity-related mediation (a*b) paths reached significance (Table S4), our main mediation findings appear specific to reappraisal-related activity, particularly for the dACC. Additional mediation modeling supported this interpretation. Hence, when extracted dACC reactivity-related activity was added as another covariate to age, sex, race, education, smoking status, and CMR scores in a mediation model, the 95% CI of the indirect (a*b) effect of reappraisal-related dACC activity on PAR scores via IL-6 did not include 0 (0.003 to 0.38). These findings appear to suggest unique explanatory effects of reappraisal-related dACC activity, apart from dACC reactivity-related activity.

DISCUSSION

This study provides three new lines of evidence that neural activity evoked by reappraisal may relate to atherosclerotic CVD risk. First, elevated reappraisal-related activity in the pgACC, dACC, vmPFC, vlPFC, and AI covaried with elevated IL-6, an inflammatory cytokine involved in atherogenesis and CVD risk (50). Second, elevated reappraisal-related activity, particularly in the pgACC and dACC, covaried with greater preclinical atherosclerosis, as reflected by two arterial measures that predict CVD outcomes: IMT and IAD (53). Third, IL-6 mediated the association between reappraisal-related dACC activity and preclinical atherosclerosis. We thus speculate that functional variation within regions of the prefrontal cortex, particularly the dACC, and associated variation in systemic inflammation may represent components of a neurobiological pathway linking emotion regulation by reappraisal to CVD risk.

Emotion regulation is widely implicated in CVD risk (4, 5, 11). However, while emotion regulation is studied extensively in the context of affective disorders that confer CVD risk (12, 67, 68), the etiological or preclinical pathways linking emotion regulation to CVD are uncertain and understudied (4). One speculation is that reappraisal and other adaptive emotion regulation strategies may support the control of affective responses and the management of disadvantageous health behaviors that confer CVD risk (4, 69); however, only one study to our knowledge has tested whether reappraisal in particular relates to a known marker of CVD risk (70). In that study, individuals who self-reported infrequently using reappraisal exhibited elevated C-reactive protein, an inflammatory marker that predicts future CVD and whose production is stimulated by IL-6 (50, 71). Interestingly, a post-hoc observation from our study was that higher IL-6 covaried with greater self-reported negative affect on regulation trials, even after accounting for negative affect on reactivity trials, pr = 0.20, p = 0.01. This appears consistent with prior findings (70) and may further suggest an association between inflammation and poorer in vivo reappraisal performance. Though, we also note that after accounting for reactivity-related affect ratings, regulation ratings did not significantly correlate with PAR scores or reappraisal-related activity in the only ROI cluster localized to the dACC (shown in Figure 1) to associate with IL-6 and PAR scores in a mediational framework, prs = −0.06 to −0.03, ps > 0.50. These null findings agree with another post-hoc finding of a moderate correlation between affect ratings and reappraisal-related BOLD activity, which was localized to a cluster in the dACC that did not spatially overlap with the cluster engaged by reappraisal and associated with IL-6 and PAR scores (Figure S7). Moreover, adding affect ratings and changes in these ratings as covariates to models testing the reappraisal-related dACC→IL-6→PAR pathway did not alter the indirect mediation effect’s significance (a*b 95% CIs = 0.005 to 0.45). Collectively, these patterns of association agree with work showing that elevated IL-6 and inflammation are associated with self-reported negative affect (47) and potential alterations in reappraisal (e.g., 70), but they also suggest that the neural correlates of self-reported affect per se may be different from those of biological and vascular markers of CVD risk.

The associations between IL-6 and reappraisal-related neural activity (a Paths, Table 1) appear to agree with existing neuroimaging findings. For example, typhoid vaccination, which induces an inflammatory state, increases dACC activity during an executive function task (72). These findings were extended recently to show that elevated dACC activity induced by Salmonella-typhi mediated the effects of inflammation on cardiovascular activity (73). Additionally, inflammatory states induced by endotoxin increase dACC and AI metabolic activity at rest (74), as well as dACC and vlPFC activity during an affective processing task (75). In parallel to findings on induced inflammatory states and neural activity, it is noteworthy that elevated dACC and AI activity evoked by social exclusion predict greater increases in levels of the soluble receptor for the inflammatory cytokine, tumor necrosis factor-α (76). Elevated vmPFC activity in bereaved women also predicts elevated levels of the inflammatory cytokine, interleukin-1β, and soluble tumor necrosis factor receptor-II (77). Finally, elevated pgACC and vmPFC activity is associated with greater increases in behaviorally-evoked circulating natural killer cells (78, 79). Taken together with our findings, it thus appears that across diverse behavioral contexts, markers of acute and chronic inflammation associate with functional activity and evoked changes within several prefrontal, cingulate, and insular regions.

Anatomically, such associations may be mediated by efferent (visceromotor) and afferent (viscerosensory) brain-immune and immune-brain communication pathways (44–46). Indeed, efferent projections from prefrontal, cingulate, and insular regions could influence subcortical cell groups that govern neuroendocrine and autonomic traffic to immune cells in a manner that could modulate the production of cytokines and other mediators of systemic inflammation (44, 45). In parallel, immune-related afferent input from peripheral sources (e.g., systemic circulation) can be relayed from viscerosensory brainstem regions and other cell groups to influence prefrontal, cingulate, and insular targets (33, 45). Accordingly, reappraisal and other functional processes instantiated in prefrontal, cingulate, and insular regions may plausibly affect or be affected by inflammatory status via top-down (efferent) or bottom-up (afferent) mechanisms. With respect our observations, however, the precise functional and directional interpretations of the neural activity changes evoked by reappraisal and associated with inflammation, as reflected by IL-6, are unclear because of our cross-sectional study design.

One interpretation is that inflammatory-related neural changes may reflect the altered engagement of circuits important for executive functions that presumably support the effortful and deliberate cognitive regulation of emotion by reappraisal (19). Supporting this interpretation, systemic inflammation reflected by elevated IL-6 predicts poorer performance on tests of executive functions (80). The specific impact of inflammation on brain regions possibly supporting such functions, as well as visceral control functions, may in part explain the more consistent patterns of association between IL-6 and BOLD activity changes for regulation trials (Table 2) compared with reactivity trials (Table S4), which presumably did not entail a high degree of cognitive demand. Another compatible interpretation is that elevated inflammation, as reflected by IL-6, may have up-regulated negative affect or arousal processes (47), which consequently increased the cognitive demands required for the regulation of emotion at the neural level (e.g., there may have been more negative affect to regulate). Finally, it may be that the inflammatory-related neural changes observed here correspond to a stable individual difference characteristic reflecting the enhanced processing of salient stimuli during deliberate or effortful emotion regulation attempts that when expressed repeatedly over time has chronic and downstream (efferent) autonomic or neuroendocrine effects on immune regulation. In this regard, it is notable that the mediation findings linking dACC activity specifically to preclinical atherosclerosis (which reflects a long-term inflammatory process) via IL-6 appeared specific to activity evoked by emotion regulation (reappraisal) and not emotional reactivity. Yet, reactivity trials may have still engaged spontaneous or implicit emotion regulation processes at the neural level, as compared with deliberate or effortful reappraisal processes engaged by regulation trials (64). If so, then this may extend the individual difference interpretation offered above to suggest differential patterns of association between the neural correlates of implicit (spontaneous, automatic) and explicit (deliberate, effortful) emotion regulation processes and CVD risk factors. To test these possibilities, future work could examine more fine-grained temporal dynamics of BOLD responses within and across prefrontal and networked subcortical regions that support emotion regulation and visceral control functions, which may better differentiate implicit vs. explicit emotion regulation neural processes (21, 81, 82).

We reported previously that amygdala activity and pgACC activity evoked by viewing explicitly threatening facial expressions were associated with preclinical atherosclerosis (83). Here, we observed associations between reappraisal-related pgACC activity and IL-6, as well as pgACC activity and PAR scores (Table 2), with trend-level evidence that IL-6 mediated the association of pgACC activity and PAR scores (a*b path 90% CI = 0.011—0.39). We also observed associations between reactivity-related pgACC activity and IL-6 and reactivity-related sgACC activity and PAR scores. But, we observed no associations between amygdala responses on regulation or reactivity trials and IL-6 or PAR scores. Differences in experimental paradigms may explain these discordant findings. Prior work has shown that amygdala responses differ between the viewing of complex visual scenes (used here) and facial expressions (84). Prior work has also shown that an inflammatory challenge (endotoxin administration) increases amygdala reactivity to explicitly and socially-threatening images, but not other images (85). Thus, associations between amygdala activity and inflammatory pathways relevant to CVD risk may be influenced by dimensions of threat or social factors in ways that have yet to be determined.

Four final issues bear on the interpretation of our findings. First, we excluded individuals with medical and psychiatric disorders, limiting generalizability. Second, we studied a single emotion regulation strategy: reappraisal. We targeted reappraisal because it is a widely studied and adaptive emotion regulation strategy that engages brain systems plausibly involved in visceral and immune functions associated with CVD risk. Moreover, reappraisal is a core component of cognitive-behavioral therapies now being used to reduce CVD risk, where mechanistic understanding is lacking. Third, our findings and measures are cross-sectional, precluding causal inference. Finally, the exact functional role of the dACC in the dACC→IL-6→PAR reappraisal mediation pathway is unclear. The dACC supports multiple functions important for effortful and goal directed behaviors, with growing evidence further implicating the dACC in emotional appraisal and regulation functions that are integrated with visceral control processes (22, 37, 86, 87). Thus, the elevated dACC reappraisal activity observed in association with IL-6 and preclinical atherosclerosis could reflect an alteration in one or more of the integrative cognitive, emotional, and visceral functions of the dACC.

To close, the regulation of negative affect by reappraisal may relate to atherosclerotic CVD risk via a pathway encompassing the interplay between prefrontal activity, particularly dACC activity, and systemic inflammation. An open question is whether interventions that involve cognitive reappraisal could reduce CVD risk in part by affecting the functionality of the dACC and other regions that may be jointly involved in emotion and inflammatory regulation.