Assessing the role of the amygdala in fear of pain: Neural activation under threat of shock

Abstract

Introduction:

The DSM-5 explicitly states that the neural system model of specific phobia is centered on the amygdala. However, this hypothesis is predominantly supported by human studies on animal phobia, whereas visual cuing of other specific phobias, such as dental fear, do not consistently show amygdala activation. Considering that fear of anticipated pain is one of the best predictors of dental phobia, the current study investigated neural and autonomic activity of pain anticipation in individuals varying in the degree of fear of dental pain.

Method:

Functional brain activity (fMRI) was measured in women (n=31) selected to vary in the degree of self-reported fear of dental pain when under the threat of shock, in which one color signaled the possibility of receiving a painful electric shock and another color signaled safety.

Results:

Enhanced functional activity during threat, compared to safety, was found in regions including anterior insula and anterior/mid cingulate cortex. Importantly, threat reactivity in the anterior insula increased as reported fear of pain increased and further predicted skin conductance changes during pain anticipation.

Limitations:

The sample was comprised of women.

Conclusions:

Individual differences in fear of pain vary with activation in the anterior insula, rather than with the amygdala, indicating that fear is not uniquely associated with amygdala activation. Whereas coping techniques such as emotion regulation have been found to vary with activation in a frontal-amygdala circuit when confronted with visual cues, precision psychiatry may need to target specific brain circuits to diagnose and treat different types of specific phobia.

1. Introduction

Among specific phobias, dental fear is common, with prevalence rates varying between 13% and 19% (Oosterink et al., 2009). Specific phobias are characterized by excessive fear of phobic stimulation and are typically accompanied by persistent avoidance behavior, which, in the case of dental fear, is clearly a health risk. DSM-5 states that “Current neural system models for specific phobia emphasize the amygdala and related structures ...” (page 199; American Psychiatric Association, 2013). Neuroimaging evidence for a major role of the amygdala activation in specific phobia primarily comes from studies investigating animal fear, in which individuals reporting spider or snake fear show enhanced blood-oxygen-level-dependent (BOLD) activity in the amygdala when viewing phobogenic, compared to non-feared, scenes (Dilger et al., 2003; Sabatinelli et al., 2005; Straube et al., 2004, 2005, 2006; Veltman et al., 2004; Wendt, et al., 2008).

On the other hand, when visual cues have been used to investigate dental phobia, amygdala engagement for phobogenic visual images is not reliably found (Caseras et al., 2010; Hermann et al., 2007; Hilbert et al., 2014; Lueken et al., 2011, 2014; Schienle et al., 2013; for amygdala involvement, see Feldker et al., 2017). Considering that fear of anticipated pain is one of the best predictors of dental phobia, the current study investigated neural correlates and autonomic activity of pain anticipation in a sample of individuals varying in the degree of fear of dental pain. To the extent that fear of pain is mediated by differences in a functional network involving the amygdala, we expected heightened amygdala activation when high fear participants anticipated receiving painful stimulation.

In the laboratory, fear of pain has been investigated by threatening individuals with the possibility of receiving a painful shock, with one cue associated with potential shock exposure, while another cue signals safety. In healthy participants, threat of shock reliably elicits defensive reactivity, evidenced by potentiation of the startle reflex during threat, compared, to safety (e.g., Bradley, Moulder, Lang, 2005; Bublatzky et al., 2013; Costa, Bradley, Lang, 2015; Dunning, DelDonno, Hajcak. 2013; Grillon et al., 1991, 1993) as well as enhanced skin conductance responses (Bradley et al., 2017; Costa et al., 2015; Kopacz & Smith, 1971), and heightened event-related positive potentials over centroparietal EEG sensors (Böcker et al., 2004; Bublatzky and Schupp, 2013). Importantly, participants reporting high dental fear show enhanced psychophysiological defensive reactivity when under threat of shock - a potentially painful stimulus that is not specifically related to dental or facial pain (Bradley, Silakowski, & Lang, 2008).

The DSM-5 suggests that amygdala involvement is a central node in fear processing, based, in part, on data from threat of shock studies. Early neuroimaging studies reported enhanced functional activity in the amygdala when under threat of shock, compared to safety (Alvarez et al., 2011; Phelps et al., 2001). A number of recent studies investigating pain anticipation in healthy participants during threat of shock, however, reported increased activation in the anterior insula, as well as in mid and anterior cingulate cortex, supramarginal gyrus, and parietal associative cortex (Alvarez et al., 2015; Baldestron et al., 2017; Kirlic et al., 2019; Mechias, Etkin, Kalish, 2010; Reicherts, Wiemer, Gerdes, Schulz, Pauli, & Wiser, 2017) rather than in the amygdala. In the current study, the neural circuit underlying fear of pain anticipation was further studied as it varied in individuals reporting high or low fear of pain. To the extent that different neural regions underlie fear expression in specific aversive contexts, individual differences in fear of pain might covary with functional activity in the consistently replicated pain anticipation network, including the anterior insula. Because threat of shock prompts increased skin conductance changes that covary with activation of the anterior insula (Kirlik et al., 2019) or the amygdala (Phelps et al., 2001), skin conductance was continuously measured during threat and safe periods.

Understanding the neural networks underlying affective disorders (such as fear) represents a major advancement in precision psychiatry (Fernandes, Williams, Steiner, Leboyer, Carvalho, & Berk, 2017), because it allows targeting specific brain regions when assessing or treating the underlying dysfunctional emotional response. For instance, it has been proposed that frontal-limbic circuitry that includes the amygdala mediates dysfunctional emotion regulation (Morawetz, Bode, Baudewig, Heekeren, 2017; Ochsner et al., 2002, 2004; Porta-Casteras, et al., 2020; Urry et al., 2006), with treatment focusing on this circuit to regulate and reduce fear response (Hartley & Phelps, 2010). If fear of pain does not reliably engage the amygdala, on the other hand, different therapeutic techniques may be required.

2. Method

Thirty-four female students (M age = 18.8, SD = 0.8) from the University of Florida participated for course credit or $20. The University of Florida Institutional Review Board approved the study, and informed consent was obtained before entering the scanner. Data from two participants were discarded for excessive motion and from one participant for scanner malfunction, resulting in a final sample of 31 participants.

Considering gender differences in fear response strategies (Schienle et al., 2013; Shansky. 2019) and previous studies on similar populations that used a sample sizes approaching 30 (e.g., Schienle et al., 2013). a planned sample size of approximately 30–35 women was targeted. To include participants with varying degrees of fear of dental pain. three items from the Dental Fear Survey (DFS; Kleinknecht. Klepac. and Alexander. 1973; items 1. 14. and 20) were administered in a prescreening survey to students taking an introductory psychology course. To allow a complete coverage of the fear continuum. women that varied in reported fear. from low to high (score above the 95th percentile at the three items used during the screening). were contacted by phone and invited to participate.

On the day of the fMRI session. all participants completed the entire DFS (which consists of 20 items). In addition. the Fear of Dental Pain questionnaire. which probes specific pain scenarios involved in dentistry (FDP; van Wijk & Hoogstraten. 2003). and the Fear of Pain Questionnaire-III (FPQ-III; McNeil and Reinwater. 1998) which assesses fear of pain more broadly were completed. The State-Trait Anxiety Inventory (STAI; Spielberger et al., 1983). and the Beck Depression Inventory (BDI-II; Beck et al., 1996) were administered to assess individual differences in temperament and mood.

2.1. Materials and Design

Stimuli consisted of 16 red and blue squares equivalent in hue and color saturation. Half (8) of the cues signaled that an electric shock could be delivered through an electrode attached to the leg. whereas half (8) signaled safety. Cues were presented for 12 s followed by a 12 s inter-trial interval consisting of a black screen. Stimulus presentation was controlled using a PC-compatible computer running E-Prime (Psychology Software Tools. Inc., Sharpsburg. PA). In addition to the critical 16 anticipatory trials. an additional set of trials presented standard (i.e., not phobogenic) emotional and neutral scenes that were analyzed separately (Sambuco. Costa. Lang. & Bradley. 2020).

A single mild shock (2.6 mA, 250 ms duration), was delivered through a bar electrode to the right ankle, was generated using a Powerlab stimulator attached to a ML180 Stimulus Isolator (AD Instruments, Inc., Colorado Springs, CO), following completion of all 32 anticipatory trials.

2.2. Procedure

Prior to entering the scanner, participants were instructed that hen ne color (red or blue, counterbalanced across participants) was displayed, an electric shock could be delivered through a stimulating bar electrode attached to their right ankle, and that no shock was possible during presentations of the other color.

In the scanner, the shock electrode was attached to the participant’s right ankle and the shock contingencies were repeated. Participants were not informed about the strength of the electric shock, and only a single mild shock was administered after all the experimental trials, before leaving the bore of the scanner.

After leaving the scanner, each participant rated the unpleasantness of shock anticipation and exposure on a 7-point scale ranging from unpleasant (1) to pleasant (7). Confirming the administration of the mild shock, participants rated the single shock only mildly unpleasant (M=4; SD = 1.7).

2.3. Data Acquisition and Analysis

2.3.1. functional MRI

A Tl-weighted anatomical volume was acquired using a Siemens 3T Allegra MR scanner. A total of 288 functional volumes (50 coronal slices, 2.5 mm thick, .5 mm gap) were collected using a T2* weighted echo planar imaging sequence (3 s TR, 35 ms TE, 160 mm FOV, 64 × 64 acquisition matrix). Offline, the functional data were slice-time adjusted, motion corrected, spatially smoothed (5.0 mm FWHM Gaussian kernel), and converted to percent BOLD signal change (for each voxel based on the mean across the entire time series) using the Analysis of Functional Neuroimages software (AFNI, Cox, 1996).

Hemodynamic responses were deconvolved using a multiple linear regression model with a cubic spline response function (24-s) that coded threat and safe trials and motion parameters (6) as variables. A whole-brain ANOVA was conducted on the peak BOLD activity (6-s after cues onset) with threat or safe cue as the independent variable. The alpha-level for voxelwise analysis was determined by using simulations (3dClustSim in AFNI) that model spatial fMRI noise as a mixture of Gaussian plus mono-exponential distributions. Based on a voxel-level uncorrected p-value of 0.001, simulations indicated a minimum cluster extent of 19 voxels (2.5 mm³) for a cluster-level corrected alpha of 0.05.

Based on the results of group-level analysis and the relationship between BOLD activity and reported fear of dental pain, additional follow-up analyses were performed to further test the relationship between regional functional activity and reported fear of pain. BOLD activity (threat minus safe) in the functional clusters of anterior insula, mid cingulate cortex, and supramarginal gyrus were linearly combined with a principal component analysis (PCA), where the resulting single component (eigenvalue > 1 and 71% of the variance explained) eigenvalues used to predict individual differences in reported fear of pain.

Additionally, a region of interest (ROI) analysis on the bilateral amygdala was performed to test the difference in BOLD activity during threat compared to safety, with amygdala identified in one of the structural atlases provided in AFNI (TT_N27).

2.3.2. Skin conductance

Skin conductance was recorded during scanning using two large Ag/Ag-CL electrodes attached to the left hypothenar eminence of the left palm. A low, constant voltage AC excitation (22 mV at 75 Hz) was supplied to one of n electrodes using a Powerlab ML116 –GSR amplifier. Skin conductance levels were sampled continuously at 1000 Hz using a laptop computer running Chart software (v 5.5.1, ADInstruments, Inc., Colorado Springs, CO). Activity was filtered offline in MATLAB (Mathworks, Inc., Natick, MA) using a 3 Hz low pass filter, then averaged into half-second bins, baseline deviated from mean activity 1 s prior to cue onset. Based on the resulting waveforms, skin conductance change was averaged from 4 to 12 s following cue onset and log transformed. Trials in which the skin conductance change was greater than 2 standard deviation from the mean, were excluded to reduce the effect of outliers (.07 trials removed). Because skin conductance changes were not normally distributed (Shapiro-Wilk test: W = .75, p < 0.001), a non-parametric test (Spearman p) was used to investigate the relationship between skin conductance and BOLD activity.

3. Results

Questionnaires measuring fear of dental pain, fear of pain, and dental fear were each normally distributed (Shapiro-Wilk test: W = .97, .95, .97 p = 0.45, 0.57, and 0.21, respectively). Fear of dental pain was significantly related both to reported dental fear (r = .74, p < 0.001) and to fear of general pain (r = .66, p < 0.001).

3.1. BOLD Activity

Figure 1 illustrates regions in which BOLD activity was significantly enhanced during threat compared to safety, and prominently included the bilateral anterior insula, as well as the supramarginal gyrus, anterior cingulate cortex, mid cingulate cortex, medial frontal gyrus, thalamus, dorsal caudate, orbitofrontal cortex, and precuneus (see also Table 1).

Figure 1.

Regions activated during threat of shock. Enhanced functional activity during threat compared to safety was found in the anterior insula, medial frontal gyrus, anterior and mid cingulate cortex, thalamus, caudate, and supramarginal gyrus (A). Across the regions involved in threat processing, the BQLD signal change time course shows a similar pattern of neural activity across all the regions, characterized by heightened changes in the context of a cue signaling possible electric shock, compared to safety, with a larger difference between threat and safety found at 6-s after the cue onset (B). Note: Brain slices are displayed in the neurological orientation (left hemisphere is on the left; right hemisphere is on the right).

Table 1.

Pain anticipation network. Regions showing significantly enhanced BOLD activity during threat compared to safety.

Region		BOLD Δ% Mean (SD)		Cluster Size	Peak t	+LPI coordinates in Talairach space
		Threat	Safe			X	Y	Z
Anterior Insula	R	.31 (.22)	.05 (.20)	203	7.45	−34	−16	11
	L	.30 (.28)	.06 (.17)	51	5.14	29	−26	−1
Dorsal Anterior Cingulate Cortex	M	.19 (.17)	−.06 (.14)	150	5.33	−9	−41	−21
Mid Cingulate Cortex	M	.20 (.26)	−.02 (.19)	20	4.62	−1	1	−39
Medial Frontal Gyr	L	.19 (.16)	−.06 (.19)	88	5.65	−14	−44	−44
Caudate	R	.20 (.22)	−.06 (.16)	74	4.90	−14	−1	−14
	L	.18 (.19)	−.03 (.14)	23	5.56	6	−6	−9
Thalamus	R	.19 (.16)	−.05 (.14)	69	7.63	−9	14	−6
Supramarginal Gyr	R	.30 (.21)	.10 (.16)	24	5.50	−64	41	−21
	L	.19 (.24)	−.04 (.21)	26	4.94	51	54	−9
Precuneus	M	.13 (.26)	−.10 (.21)	29	4.86	1	56	−41
Orbitofrontal	M	.16 (.37)	−.25 (.31)	19	4.22	−6	−54	9

Note. R = right, L = left, M = medial; Cluster size expressed as number of significant voxel for the comparison threat vs. safe; Location of the peak t-statistic is reported in +L +P +I coorainates in Taiairach space.

Enhanced functional activity in the anterior insula during threat, compared to safety, was consistent across the sample, with 90.3% of the participants showing a positive increase in functional activity in the anterior insula during cues signaling shock threat, compared to safety. More importantly, as illustrated in Figure 2, activation of the insula during threat, compared to safety, was significantly related to fear of dental pain (r = .55, p = 0.001), but not dental fear (r = .17, p = 0.35), with the difference in functional activation during threat and safe periods increasing as fear of pain increased. This effect was driven by an increase in insula activity during threat (r = .27), and a decrease during safety (r = −.40) as reported fear of dental pain increases.

Figure 2.

Threat reactivity in the anterior insula. Enhanced BQLD activity during threat compared to safety in the anterior insula (bilateral functional cluster resulting from whole brain analysis showed underneath the X axis) is significantly related with reported fear of dental pain.

In addition to anterior insula activation, functional activity during threat, compared to safety, increased significantly with heightened fear of dental pain in both the mid-cingulate cortex (r = .52, p = 0.003) and the supramarginal gyrus (r = .50, p = 0.004). When activation in these three regions was linearly combined in a principal component analysis, the first principal component from this analysis correlated with fear of dental pain, r = .62, p < 0.001, and reported fear of pain, r = .46, p = 0.009. Enhanced threat reactivity in this network was also related to higher rated aversiveness of anticipated shock exposure (r = .41, p = 0.02).

No voxels in the amygdala were significant in the whole-brain analysis of threat compared to safety. A follow-up ROI analysis was performed to re-assess whether there is any evidence of amygdala activation during threat of shock, resulting in a small but significant increase in functional activity in the amygdala during threat (M = .12, SD = .20), compared to safety (M = .002, SD = .11), F(1,30) = 9.84, p = 0.004, η_p² = .25. However, there were no significant relationships between magnitude of amygdala activation and reported fear of dental pain (r = .08, p = 0.66) or fear of general pain (r = .19, p = 0.30).

3.2. Skin conductance.

Skin conductance changes were greater during threat (M = .06 log μS), compared to safety (M = .02 log μS), F(1,30) = 8.37, p = 0.007, η_p² = .22. As illustrated in Figure 3, skin conductance reactivity during threat, compared to safety, significantly increased with threat activation in the anterior insula (Spearman ρ = .50, p = 0.004), the mid cingulate cortex (Spearman ρ = .49, p = 0.005) and the supramarginal gyrus (Spearman p = .40, p = 0.027). The amygdala was not related to differences in skin conductance threat reactivity (Spearman ρ = .21, p = 0.27).

Figure 3.

Scatter plots show the correlation (Spearman’s rho) between skin conduct. (scores ranked in ascending order) with BQLD threat reactivity in the anterior ins

Skin conductance during threat, compared to safety, was also significantly related to fear of dental pain (Spearman ρ = .50, p = 0.004), with skin conductance changes to threat, compared to safety, increases as reported fear of dental p .increased.

3.3. Shock ratings

Consistent with the mild shock that was delivered following the anticipatory phase, neither fear of pain nor dental fear was significantly related to the rated unpleasantness of the shock (respectively, r = −.08, p = 0.66, and r = .17, p = 0.36). Neither anxiety nor depression was related to reported questionnaire scores or to aversiveness ratings during threat of shock (all Fs < 1).

4. Discussion

Anticipating a painful electric shock prompted enhanced functional activity in a network of regions including the anterior insula, mid cingulate cortex, and supramarginal gyrus, as well as in middle frontal gyrus, thalamus, precuneus, caudate, and anterior cingulate cortex. This network consists of regions reported in previous studies measuring neural activity during anticipation of painful stimulation (e.g., Alvarez et al., 2015; Kirlic et al., 2019 Mechias, Etkin, Kalish, 2010; Meyer, Padmala, Pessoa, 2019; Palermo et al., 2015). More importantly, activation in the anterior insula during threat of shock significantly covaried with individual differences in reported fear of dental pain, with anterior insula enhancement when anticipating painful stimulation increasing as reported fear of pain increased. The magnitude of threat reactivity in the anterior insula when anticipating exposure to a potentially painful event not only covaried with individual differences in the fear of experiencing dental pain, but uniquely, prompted heightened skin conductance activity when anticipating pain.

Whereas a whole brain analysis found no evidence of amygdala activation during threat of shock, compared to safety, a follow-up ROI analysis suggested a small, but significant increase threat effect in the amygdala. However, functional activity in this region did not covary with individual differences in fear of dental pain, fear of general pain, or with skin conductance response during threat compared to safety. A small, threat related difference in amygdala activation is consistent with the results of a meta-analysis which did not find reliable activation of the amygdala during threat, compared to safety, in studies investigating instructed fear (Mechias, Etkin, & Kalisch, 2009), as well as in a more recent meta-analysis of fear processing, which reported reliable activation of both the anterior insula and mid-cingulate cortex, but not the amygdala (Fullana et al., 2016). Amygdala activation may be less reliable during threat of shock because the threat cue signals only the possibility of shock exposure, compared to actual exposure to pain in acute (Kano et al., 2013) or chronic pain contexts (DaSilva & Seminowicz, 2019).

Because the homology of the human neocortex, including anterior insula and anterior/mid cingulate cortex, is absent or unclear in species such as rats (Carlén, 2017; Vogt et al., 2013, 2014), it appears that fear learning in humans may involve different specialized circuits, depending on the feared object. When the feared situation is anticipated pain, the anterior insula and mid cingulate cortex - but not the amygdala - are not only the most reliably activated regions, but also covary with individual differences in fear of pain as well as autonomic reactivity. In humans, the amygdala’s role in fear processing may be specific to situations in which the feared stimulus is cued by a sensory (e.g., visual) representation, such as in animal phobia, consistent with data finding reliable amygdala activation when processing visual cues (Brown et al., 2011; Lang & Bradley, 2010; Lang & Davis, 2006; Sambuco, Bradley, Herring, Lang, 2020; Satpute et al., 2015).

The anterior insula (together with the mid cingulate cortex) may represent a key functional region that integrates information from other anatomical regions involved in the threat of shock. For example, threat activation in the supramarginal gyrus may be reflect constant monitoring the location of lower-limbs, including the ankles to which the shock electrode was attached in the current study (e.g., Goosens et al., 2019; Iandolo et al., 2018), consistent with a hypothesis of heightened vigilance toward the sensory region in which painful stimulation was expected (Baas, Milstein, Donlevy, & Grillon. 2006; Grillon & Davis, 1997). Additionally, the increased activation in the thalamus (medio-dorsal nucleus) found in the context of cues signaling shock threat is also found when viewing aversive scenes (Frank & Sabatinelli, 2014; Sambuco et al., 2020), suggesting broad involvement in sensory processing of cues signaling aversiveness. Moreover, activation of the dorsal portion of the caudate has been reported in learning contexts involving uncertainty (Delgado, Li, Schiller & Phelps, 2008; Balleine, Delgado, & Hikosaka, 2007), which was present in multiple ways in the current study, as participants were given no information regarding the intensity, timing, or probability of possible exposure to shock.

All these sources of information might be integrated in the anterior insula, which is proposed to be a multisesory integrative hub (Chen, Michels, Supekar, Kochalka, Ryali, & Menon, 2015; Namkung, Kim, Sawa, 2017; Satpute et al., 2015) that is primarily involved in interoceptive representation and autonomic activation (Craig, 2009; Critchley et al., 2004). Together with additional evidence showing that functional activity in the anterior insula covaries with subjective expectations in painful contexts (Wiech et al., 2010), the current data suggest that fear of pain may lead to overestimation of pain intensity, providing a marker sensitive to these individual differences.

5. Limitations

Because women are generally more sensitive to pain, as well as disproportionately represented in clinical contexts involving acute or chronic pain, the current study focused on this sensitive population. Of interest is a recent study (Horn, Alappattu, Gay, Bishop, 2014), assessing pain sensitivity to thermal stimulation, in which gender differences were found that were hypothesized to be due to differences in fear of pain. Considering the accumulating evidences on gender differences in fear response strategies (Li & Graham, 2020; Schienle et al., 2013; Shansky, 2019), the current study provides the foundation to further asses potential gender differences in threat reactivity in this sensitive population.

6. Conclusions

When anticipating exposure to potentially painful stimulation, enhanced functional activity is found in a specific subset of regions in the pain matrix -- anterior insula, mid cingulate cortex, and supramarginal gyrus -- that covary significantly with individual differences in reported fear of pain. Functional activity in the amygdala, on the other hand, was not consistently involved in threat processing and did not predict individual differences in reported fear or pain or autonomic reactivity to threat cues, suggesting that treatments such as emotion regulation, which rely on amygdala activation (Morawetz, et al., 2017; Ochsner et al., 2002, 2004; Porta-Casteràs, et al., 2020; Urry et al., 2006) may be unsuccessful. More broadly, the current data challenge the assumption of a single, amygdala-centered, fear circuit that is activated in all aversive contexts. Instead, these data suggests that high fear of anticipated pain involves a brain circuit - also involved in pain perception - that includes anterior insula and cingulate cortex. Taken together, the data suggest the need to (1) identify neural markers underlying different types of phobia and (2) better characterize the role of amygdala in fear processing in humans. Ultimately, understanding the neural circuits engaged during fear processing will lead to appropriate therapeutic treatment for fear-related disorders that acknowledge potentially different neural substrates in specific fear contexts.

Highlights.

• According to DSM-5, the neural system model of specific phobia is centered on the amygdala

• Fear of dental pain was modeled here in a context of pain anticipation

• In pain anticipation, fear of dental pain covaried with anterior insula activity, but not amygdala

• Different neural substrates mediate fear response in different types of specific phobia