OUP user menu

The effect of negative emotional context on neural and behavioural responses to oesophageal stimulation

Mary L. Phillips, Lloyd J. Gregory, Sarah Cullen, Steven Cohen, Virginia Ng, Christopher Andrew, Vincent Giampietro, Edward Bullmore, Fernando Zelaya, Edson Amaro, David G. Thompson, Anthony R. Hobson, Steven C. R. Williams, Michael Brammer, Qasim Aziz
DOI: http://dx.doi.org/10.1093/brain/awg065 669-684 First published online: 1 March 2003

This article has a correction. Please see:

Summary

Sensory experience is influenced by emotional context. Although perception of emotion and unpleasant visceral sensation are associated with activation within the insula and dorsal and ventral anterior cingulate gyri (ACG), regions important for attention to and perception of sensory and emotional information, the neural mechanisms underlying the effect of emotional context upon visceral sensation remain unexplored. Using functional MRI, we examined neural responses to phasic, non‐painful oesophageal sensation (OS) in eight healthy subjects (seven male; age range 27–36 years) either during neutral or negative emotional contexts produced, respectively, by presentation of neutral or fearful facial expressions. Activation within right insular and bilateral dorsal ACG was significantly greater (P < 0.01) during OS with fearful than with neutral faces. In a second experiment, we measured anxiety, discomfort and neural responses in eight healthy male subjects (age range 22–41 years) to phasic, non‐painful OS during presentation of faces depicting either low, moderate or high intensities of fear. Significantly greater (P < 0.01) discomfort, anxiety and activation predominantly within the left dorsal ACG and bilateral anterior insulae occurred with high‐intensity compared with low‐intensity expressions. Clusters of voxels were also detected in this region, which exhibited a positive correlation between subjective behaviour and blood oxygenation level‐dependent effect (P < 0.05). We report the first evidence for a modulation of neural responses, and perceived discomfort during, non‐painful visceral stimulation by the intensity of the negative emotional context in which the stimulation occurs, and suggest a mechanism for the effect of negative context on symptoms in functional pain disorders.

  • Keywords: oesophagus; faces; fear; mood; emotion
  • Abbreviations: ACG = anterior cingulate gyri; BA = Brodmann area; BOLD = blood oxygenation level‐dependent; EPI = echoplanar imaging; fMRI = functional MRI; FPQ = fundamental power quotient; GBAM = generic brain activation map; GI = gastrointestinal; OS = oesophageal sensation; PCC = posterior cingulate cortex

Introduction

Negative mood states, such as fear or sadness, are often associated with abnormal sensory perception such as abdominal pain (Chen et al., 1989; Weisenberg et al., 1998). Beaumont (1833) and Pavlov (1910) demonstrated that external sensory events eliciting strong emotional reactions may alter gastrointestinal (GI) function. A close relationship between emotional state and GI function is repeatedly reported in patients with functional GI disorders, including irritable bowel syndrome and non‐cardiac chest pain (Whitehead et al., 1988; Ho et al., 1998).

There is increasing experimental evidence to suggest an interaction between emotional context, cognition and sensory processing. Previous studies have demonstrated that sensory reflexes, including the acoustic startle (Kumari et al., 1996; Kaviani et al., 1999) and eye‐blink (Vrana and Lang, 1990) reflex, are modulated by the emotional context in which they occur. Furthermore, the amplitude of cortical evoked responses to GI stimulation has been shown to be modulated by attentional processes (Hollerbach et al., 1997; Hobson et al., 1998, 2000).

Advances in functional neuroimaging have provided information about the brain neural networks for emotional, cognitive and sensory processing. For instance, it has been demonstrated that human facial expressions (considered to be the primary source for conveying the emotional valence regarding a particular situation) depicting different emotions activate different brain neuronal networks. Fearful facial expressions activate the amygdala, (Breiter et al., 1996; Morris et al., 1996; Phillips et al., 1997), while facial expressions of disgust activate the insular cortex and ventral striatum (Phillips et al., 1997; Sprengelmeyer et al., 1998). Functional neuroimaging studies have also demonstrated activation of the anterior‐mid insula and anterior cingulate gyrus (ACG) during perception of somatic pain (Coghill et al., 1994; Davis et al., 1997; Derbyshire et al., 1997; Derbyshire and Jones, 1998; Oshiro et al., 1998; Casey, 1999) and unpleasant oesophageal sensation (Aziz et al., 1997).

The ventral ACG (including the perigenual and subgenual regions of the ACG, Brodmann areas 32, rostral 24 and 25) is considered to be involved in affective processing and is activated during the assessment of emotional and motivational information and, with the anterior insula, during recognition of negative emotions and during depressed states (Drevets and Raichle, 1998; Phillips et al., 1997, 1998; Mayberg et al., 1999). The dorsal ACG, including Brodmann areas (BAs) 24 and 32, is considered to be the cognitive division, and is activated during cognitively demanding tasks involving modulation of attention and executive functions (Bush et al., 2000). The dorsal and ventral ACG, together with the anterior‐mid insulae, are therefore important components of a neural system mediating attention to and perception of sensory and emotional information. However, to date, no study has examined the extent to which an emotional context modulates the central processing of visceral sensation.

We wished to examine the effect of presentation of a negative emotional context upon the intensity of activation within neural regions important for attention to and perception of sensory and emotional information, dorsal and ventral ACG, and anterior insular cortices, during visceral stimulation which would not be perceived as salient in a non‐emotional context: non‐painful oesophageal stimulation (OS). We predicted that altering the emotional context in which the non‐painful OS occurred from neutral to negative would then be associated with increased attentional and emotional processing demands, and increased activation in these neural regions.

In two functional MRI (fMRI) experiments, we employed fearful and neutral facial expressions from a standardized series (Ekman and Friesen, 1975) to provide negative and neutral emotional contexts, respectively, whilst healthy subjects experienced phasic, non‐painful OS. The facial stimuli have been employed in many previous functional neuroimaging studies examining neural responses to emotional stimuli (e.g. Morris et al., 1996; Phillips et al., 1997). In the first experiment, we examined the effect of a negative emotional compared with a neutral context upon ventral and dorsal ACG, and insular responses to phasic, non‐painful OS. In a second experiment, we examined the effect of altering the intensity of the negative emotional context during non‐painful OS upon the level of reported anxiety and discomfort, and the intensity of activation within these neural regions.

Experiment 1

Material and methods

Subjects

Eight healthy volunteers (seven male; mean age: 32 years; range: 27–36 years; mean number of years of education: 16 years) participated in the study. Subjects reported no history of neurological, gastrointestinal or psychiatric disorder and were on no medications at the time of study. Informed, written consent was obtained after the nature and possible consequences of the study were explained. Approval for the study was obtained from the Ethical Committee (Research) of the Institute of Psychiatry.

Experimental design

A traditional box‐car design comprising alternating 30 s blocks of experimental and control stimuli for a duration of 5 min was employed. Either 10 prototypically fearful or 10 neutral facial expressions displayed by eight different identities (five female and three male) from a standardized series (Ekman and Friesen, 1975) were presented for 2 s each with a 1 s interstimulus interval per 30 s block. The facial stimuli were presented on a screen placed in front of the subjects in the scanner by a back projection technique, and subtended ∼10° of visual angle vertically and 8° horizontally.

The experiment was a simple 2 × 2 design: with oesophageal stimulation either on or off, and emotional context either fearful or neutral. The experiment was divided into four experimental conditions:

(A) Presentation of fearful faces with phasic, non‐painful OS versus presentation of fearful faces alone thereby generating a continuous context of fear throughout the condition.

(B) Presentation of fearful faces versus presentation of neutral faces.

(C) Presentation of neutral faces with phasic, non‐painful OS versus presentation of neutral faces alone thereby generating a continuous context of neutrality throughout the condition.

(D) Presentation of fearful faces with phasic, non‐painful OS versus presentation of neutral faces with phasic, non‐painful OS.

A schematic representation of the experimental design is shown in Fig. 1.

Fig. 1 The design of the four 5 min experimental conditions in the first experiment: each comprised alternating 30 s blocks of: (A) presentation of 10 fearful faces with phasic, non‐painful OS and presentation of 10 fearful faces without OS; (B) presentation of 10 fearful faces and presentation of 10 neutral faces without any OS in either block; (C) presentation of 10 neutral faces with phasic, non‐painful OS and presentation of 10 neutral faces alone; and (D) presentation of 10 fearful faces with phasic, non‐painful OS and presentation of 10 neutral faces with phasic, non‐painful OS. Each facial expression was presented for 2 s. During alternate blocks of conditions A and C, and both blocks of condition D, a single phasic non‐painful balloon distension was delivered to the distal oesophagus 1.5 s into the presentation of each facial expression.

With this design, subjects participated in two experimental conditions in which the emotional context was similar but with additional sensory stimulation in one condition and not in the other, i.e. presentation of alternating blocks of fearful and neutral faces with either no OS or phasic OS throughout the condition (conditions B and D, respectively), and two experimental conditions in which there were similar amounts of sensory stimulation but with different emotional contexts, i.e. periodic, phasic OS with continual presentation of fearful faces or periodic, phasic OS with continual presentation of neutral faces (conditions A and C, respectively). The order of experimental conditions was counterbalanced across subjects to avoid effects of order.

Oesophageal stimulation was performed by phasic distension of a 2 cm long silicone balloon with air. The balloon was mounted 15 mm from the tip of a 4 mm diameter multilumen polyvinyl catheter (Wilson Cook, Letchworth, UK). Prior to scanning, the balloon catheter was passed perorally into the oesophagus, and the balloon positioned 30 cm from the incisors in the distal oesophagus. The balloon was connected to a specially constructed pump (Medical Physics Department, Hope Hospital, Salford, UK), which was capable of rapid phasic distension. The maximum flow rate produced by the pump was 200 ml/s and the rise time to maximum balloon inflation remained constant (165 ms) for any given volume. The balloon volume was controlled using a dial on the front of the pump, which allowed the pressure in the system to be regulated [pressure range 0–25 psi (pound force per square inch)]. Increasing the pressure in the system increased the flow rate in the airlines and, therefore, a greater volume was delivered during the inflation cycle. In vitro, the pump was capable of delivering a maximum balloon volume of 30 ml. The balloon was completely deflated immediately after maximum inflation.

To determine sensory and pain thresholds for each individual, the catheter was connected to the pump and the balloon was phasically inflated, whilst the volume was increased in a stepwise manner in 1 psi increments. A value, in psi, representing 50% of the difference between the sensory and pain threshold was calculated as the volume necessary to produce a clearly perceptible, but non‐painful sensation, in each subject (Hobson et al., 1998, 2000). In alternate 30 s blocks in experimental conditions A and C, and for both 30 s blocks in experimental condition D, a single phasic non‐painful balloon distension was delivered to the distal oesophagus 1.5 s into each 2 s face presentation, so that subjects received 10 OS per 30 s block.

Subjects were requested to view the faces carefully in each experimental condition. At the end of each experimental condition, subjects were asked to identify the emotion depicted by the faces.

Image acquisition and analysis

Gradient echo echoplanar imaging (EPI) data were acquired on a Neuro‐optimized 1.5 T MR system (General Electric, Milwaukee, WI, USA) at the Maudsley Hospital, London, UK. A quadrature birdcage headcoil was used for radio frequency (RF) transmission and reception. A hundred T2*‐weighted images depicting blood oxygenation level‐dependent (BOLD) contrast (Ogawa et al., 1990) were acquired over 5 min (for each task) at each of 14 near‐axial non‐contiguous 7 mm thick planes parallel to the intercommissural (AC–PC) line: TE (echo time) = 40 ms; TR (repetition time) = 3 s; matrix size = 64 × 64; FOV (field of view) = 240 × 240; in‐plane resolution = 7 mm; interslice gap = 0.7 mm.

This EPI dataset provided almost complete brain coverage. In the same scanning session, an inversion recovery EPI dataset was acquired at 43 near‐axial 3 mm thick planes parallel to the AC–PC (anterior–posterior commissure) line: TE = 80 ms; TI (inversion time) = 180 ms; TR = 16 s; in‐plane resolution = 1.5 mm; interslice gap = 0.3 mm; number of signal averages = 8. This higher resolution EPI dataset provided whole brain coverage, and was later used to register the fMRI datasets acquired from each individual into standard stereotactic space.

The statistical inferential procedure used in this paper does not utilize any underlying assumptions about the distribution of the test statistic (the fundamental power quotient, FPQ), but instead calculates it from the data itself by previously described randomization techniques (Ogawa et al., 1990; Bullmore et al., 1996; Brammer et al., 1997). The data were first realigned (Bullmore et al., 1999a) to minimize motion related artefacts. Responses to the experimental paradigms were then detected by time series analysis using a truncated Fourier series consisting of pairs of sine and cosine terms at the alternation frequency of the block periodic (A/B) experimental paradigm and its first two harmonics. Using this approach, the power and phase of the BOLD responses to the block periodic (A/B) experimental paradigm (Bullmore et al., 1996) could be computed. In addition to the sine and cosine terms, the mathematical model for the response incorporated an intercept term (image intensity at time zero) and an estimate of time‐dependent signal drift. If the amplitudes of the sine and cosine components at the stimulus frequency that gave the best fit to the observed data at a given voxel are represented by γ and δ, the power of periodic response to the input function is given by {γ22}. Dividing the power by its standard error yields the standardized power (FPQ) of the response at each voxel. The FPQ is therefore a measurement of the magnitude or strength of the BOLD effect at each intracerebral voxel. Parametric maps representing FPQ observed at each intracerebral voxel were constructed.

In order to sample the distribution of FPQ under the null hypothesis that observed values of FPQ were not determined by experimental design (with minimal assumptions), the time series at each voxel was permuted randomly and FPQ estimated exactly as above but using this permuted time series. This process was repeated 10 times at each voxel and then over all voxels, resulting in 10 permuted parametric maps of FPQ at each plane for each subject. Combining this data yields the distribution of FPQ under the null hypothesis. Voxels activated any desired level of type 1 error can then be determined obtaining the appropriate critical value of FPQ from the null distribution. For example, FPQ values in the observed data lying above the 99th percentile of the null distribution have a probability under the null hypothesis of <0.01. In order to extend inference to the group level, the observed and randomized FPQ maps were transformed into standard space and smoothed by a 2D Gaussian filter with full width half maximum = 11 mm. This filter size was chosen in order to accommodate regional differences in brain anatomy between subjects (Clark et al., 1996).

Generic brain activation maps (GBAMs)

A generic brain activation map (GBAM) was produced for each experimental condition by testing the median observed FPQ (median values were used to minimize outlier effects) at each intracerebral voxel in standard space (Talairach and Tournoux, 1988) against a critical value of the permutation distribution for median FPQ ascertained from the spatially transformed permuted data (Brammer et al., 1997). The generic neural response of all participating subjects during each experimental condition was therefore represented by the corresponding GBAM, with activated regions identified by reference to the Talairach Atlas (Talairach and Tournoux, 1988). The threshold of P < 0.004 for activation at each intracerebral voxel was chosen, since this represented a total number of false positive (Type 1) errors (false positive activated voxels) of 50 per brain volume (∼14 000 voxels), i.e. two false‐positive activated voxels per each of the 25 brain slices comprising the total brain volume following the transformation of the fMRI data into standardized space. This threshold therefore ensured a relatively low number not only of false positive, but also of false negative activated voxels per total brain volume.

In order to examine the effect of the two factors (OS and negative emotional context upon ventral and dorsal ACG and insular responses), average values of the statistic maps were produced for each subject for both conditions in which OS was contrasted with no OS (conditions A and C), and for both conditions in which fearful faces were contrasted with neutral faces (conditions B and D). Two group images were then computed from these averaged maps.

Comparison of experimental conditions

In order to examine the effect of presentation of a negative emotional context upon neural responses to OS and the effect of OS upon neural responses to fearful facial expressions, we estimated the differences in mean FPQ between conditions A and C, and between conditions D and B, respectively, by fitting a repeated measures analysis of variance (ANOVA) model at each voxel of the observed FPQ maps in standard space: FPQi,j = b0 + bjE + ei,j. Here, FPQi,j denotes standardized power in the ith individual under the jth condition, b0 is the overall mean FPQ, b0 + bj is the mean FPQ under the jth condition, E is a dummy variable coding condition, and ei,j is a residual quantity unique to the ith individual. The null hypothesis of zero difference in mean FPQ between conditions was tested by comparing the coefficient bj to critical values of its non‐parametrically ascertained null distribution. To do this, the elements of E were randomly permuted 10 times at each voxel, bj was estimated at each voxel after each permutation, and these estimates were pooled over all intracerebral voxels to sample the permutation distribution of bj (Bullmore et al., 1999b). For a two‐tailed test of size P = 0.05, the critical values were the 100 × p/2th and 100 × (1 – p/2)th percentile values of the permutation distribution. Differences in mean FPQ between conditions were tested for significance only at those voxels, which were generically activated by one or both of the conditions considered independently, thereby substantially reducing the search volume or number of tests conducted.

Results

All subjects tolerated the study well. Oesophageal sensation was perceived as a pulsatile sensation over the sternum. The mean (±SD) value for balloon intensity was 18 psi ± 5.75. After scanning, all subjects were able to identify correctly facial expressions viewed during the experimental conditions as either fearful or neutral.

Generic brain activation maps of neural responses to OS and fearful facial expressions

The mean GBAMs of conditions A and C, and of conditions B and D, demonstrating neural responses to OS compared with no OS, and to fearful compared with neutral faces are shown in Fig. 2A and Fig. 2C, respectively. The comparison of GBAMs for conditions A and C, demonstrating neural regions activated to a significantly greater extent when fearful than when neutral faces were presented during OS, is shown in Fig. 2B. The comparison of GBAMs for conditions B and D, demonstrating neural regions activated to a significantly greater extent when fearful faces were presented without OS than when OS occurred during presentation of fearful faces, is shown in Fig. 2D.

Fig. 2 Generic brain activations are shown for the eight subjects participating in the first experiment representing: (A) the mean neural response to conditions A and C; (B) the mean neural response to conditions B and D; (C) the comparison of neural responses during conditions A and C, demonstrating regions activated to a significantly greater extent when fearful than when neutral faces were presented during non‐painful OS; and (D) the comparison of neural responses during conditions B and D, demonstrating regions activated to a significantly greater extent when fearful faces were presented without OS than when non‐painful OS occurred during presentation of fearful faces. Brian slices are shown at 4 mm and 37/42 mm above the transcallosal plane. The right side of the brain is shown on the left side of the picture for each brain slice, and vice versa. In A, major regions of activation are demonstrated in the insula and, in A and B, within a dorsal region of the anterior cingulate gyrus. In C, major regions of activation are demonstrated predominantly within bilateral occipitotemporal regions (BA 18 and 31) and, in C and D, within the hippocampus.

Conditions A and C: overall effect of OS versus no OS

Activation was demonstrated within bilateral dorsal and ventral ACG, and bilateral mid‐insulae, in addition to the right hippocampus, bilateral cerebellum, occipitotemporal cortical regions, and dorsolateral and ventral prefrontal cortices (Table 1AA and Fig. 2A).

View this table:
Table 1A

Major brain regions activated by non‐painful OS (conditions A and C)

Cerebral regionSidex*y*z*BAFPQSize
CerebellumR7–50–22.1167
L–17–67–72.1225
Occipitotemporal regions:
 PrecuneusR4–564271.9137
 CuneusR4–739171.9101
L–11–7615181.838
 Superior temporal gyrusR36–4–7381.971
L–50–79221.993
 Middle temporal gyrusR53–394211.816
L–43–6026391.873
 Posterior cingulate gyrusR7–4615311.957
 Insula (middle)R40–7–21.878
L–43042.073
 Anterior cingulate gyrus (ventral)R/L039–7322.763
 Anterior cingulate gyrus (dorsal)R/L03331321.819
L–43326321.842
–42637321.611
 HippocampusR36–4–131.748
 Dorsolateral prefrontal cortexL–43720441.833
 Ventromedial prefrontal (orbitofrontal) cortexR1143–13112.119

*Talairach co‐ordinates refer to the voxel with the maximum FPQ in each cluster. All such voxels were identified by a one‐tailed test of the null hypothesis that median FPQ is not determined by experimental design. The probability threshold for activation was P ≤ 0.004. L = left; R = right.

Conditions B and D: presentation of fearful faces versus presentation of neutral faces

Activation was demonstrated within bilateral ventral and left dorsal ACG, and the left hippocampus, in addition to bilateral cerebellum, occipitotemporal cortical regions, bilateral thal amus, left dorsolateral and right ventral prefrontal cortices (Table 1BB and Fig. 2C).

View this table:
Table 1B

Major brain regions activated during presentation of fearful faces (conditions B and D)

Cerebral regionSidex*y*z*BAFPQSize
CerebellumR25–46–131.837
L–36–63–132.4109
ThalamusR4–2641.857
L–4–1791.610
Occipitotemporal regions:
 Posterior cingulate gyrusR7–4642311.744
L–32–1742181.662
 Fusiform gyrusR21–46–7371.837
 PrecuneusL–4–673771.621
 Superior temporal gyrusL–50–3015421.716
 Lingual gyrusR4–604181.614
 Middle temporal gyrusR430–18211.611
 Anterior cingulate gyrus (dorsal)L –42631321.618
 Anterior cingulate gyrus (ventral)R7439321.610
L–7464321.615
 Dorsolateral prefrontal cortexL–40731441.616
 Ventromedial prefrontal cortexR1146–7101.69
 HippocampusL–25–43–21.58

*Footnote as for Table 1A.

Comparison of conditions A and C: the effect of presentation of fearful faces upon neural responses to OS

Significantly greater activation was demonstrated within the dorsal region of the left ACG during condition A than during condition C (x = –3, y = –8, z = 42; number of activated voxels = 31 in this region; Table 1CC). No regions were activated to a significantly greater extent during condition C than condition A (P = 0.01; overall search volume = 3530 voxels; expected number of false positive activated voxels over the whole brain = 35; number of observed activated voxels = 62; Table 1CC and Fig. 2B).

View this table:
Table 1C

Major brain regions activated significantly more by condition A than C, and by B than D

Cerebral regionSidex*y*z*BAFPQSizeComparison
Anterior cingulate gyrus (dorsal)L–3–842242.031A>C
HippocampusR17–33–71.114B>D

*Talairach co‐ordinates refer to the voxel with the maximum FPQ in each cluster. Regions activated significantly more in A compared with C, and B than D are demonstrated (P = 0.01). L = left; R = right.

Comparison of conditions D and B: the effect of OS upon neural responses to fearful facial expressions

Significantly greater activation was demonstrated within the right hippocampus during condition B than during condition D (x = 17, y = –33, z = –7; number of activated voxels = 14; Table 1CC). No regions were activated to a significantly greater extent during condition D than condition B (P = 0.01; overall search volume = 1120 voxels; expected number of false positive activated voxels over the whole brain = 11; number of observed activated voxels = 43; Table 1CC and Fig. 2D).

These findings allowed us to design a second experiment to test the hypothesis that increasing the intensity of negative emotional context in which OS occurred would be associated with increased activation predominantly within dorsal ACG.

Experiment 2

Methods

Subjects

Eight healthy, right‐handed male volunteers (median age: 22 years; age range: 22–41 years; mean number of years in education: 16 years) participated in the study. Subject exclusion criteria and the method for obtaining informed consent were as described in Experiment 1.

Experimental design

A modified box‐car design was employed in this experiment, comprising alternating 20 s ‘active’ blocks of presentation of fearful faces and non‐painful OS versus fearful faces without OS, with each of these 20 s blocks preceded by a 16 s period of silence (see below). In order to modulate the emotional context during this study, prototypical expressions of fear from a standardized series (Ekman and Friesen, 1975) were morphed with prototypically neutral expressions from the same series to create facial expressions depicting two lower intensities of fear in addition to prototypical or high intensity fear: moderate intensity fear (50% fear and 50% neutral) and mild intensity fear (25% fear and 75% neutral; Young et al., 2002). In each of three conditions, therefore, either ten 100% fearful, ten 50% intensity fearful, or ten 25% intensity fearful facial expressions were displayed by eight different identities (five female and three male), each presented for 2 s, in each of the 20 s blocks. During the 20 s active blocks, facial stimuli were presented and distal oesophageal non‐painful stimulation was performed as described in Experiment 1.

During the 16 s periods preceding the 20 s active blocks, subjects were asked to rate on a visual analogue scale (ranging from 1–8) the following:

(i) Oesophageal discomfort experienced during the preceding 20 s block (1 = slight, 8 = severe).

(ii) Anxiety experienced in the preceding 20 s block (1 = little and 8 = much).

(iii) The intensity of fear in the facial expressions displayed in the previous 20 s block (1 = very mild; 8 = very intense).

These three questions were presented on the screen in front of subjects, and the selection was made using an analogue button box underneath the subject’s right hand. Two buttons were employed to move a cursor in each direction of the scale in order to prevent too much movement.

Subjects therefore participated in three experimental conditions, each having a duration of 6 min and 30 s:

(1) Presentation of facial expressions of high intensity (100%) fear with phasic, OS versus presentation of the fearful faces alone.

(2) Presentation of facial expressions of fear of moderate (50%) intensity, with phasic, OS versus presentation of the fearful faces alone.

(3) Presentation of facial expressions of fear of mild (25%) intensity, with phasic, OS versus presentation of the fearful faces alone.

A schematic representation of the experimental design is shown in Fig. 3.

Fig. 3 The design of the three 6.5 min experimental conditions in the second experiment. Each comprised alternating 36 s blocks: 16 s of silence, followed by a 20 s active block in which subjects viewed 10 facial expressions, each for 2 s. In all but the first 16 s silent period, subjective ratings (SR) of anxiety, perceived discomfort and intensity of fear in the facial expressions viewed in the preceding 20 s block were obtained. In the three conditions, subjects viewed either (1) facial expressions of high intensity (100%) fear with phasic, non‐painful OS versus presentation of the fearful faces alone; (2) facial expressions of fear of moderate (50%) intensity, with phasic, non‐painful OS versus presentation of the fearful faces alone; or (3) facial expressions of fear of mild (25%) intensity, with phasic, non‐painful OS versus presentation of the fearful faces alone. Phasic non‐painful balloon distension was delivered to the distal oesophagus in the same manner as in the first experiment during alternate blocks in all three conditions.

With this design, subjects participated in three experimental conditions in which the degree of sensory stimulation was the same over all conditions, but the intensity of the emotional was increased from mild (condition 3) to moderate (condition 2) to high intensity of fear (condition 1).The order of experimental conditions was counterbalanced across subjects to avoid effects of order

Acquisition

Parameters for fMRI data acquisition were as described for Experiment 1. During the active 20 s blocks, 10 whole brain image volumes were collected with TR = 2 s. During the 16 s of silence preceding the 20 s active blocks, brain water magnetization was maintained in equilibrium (steady state) by delivering spatially selective RF pulses at the same rate as that with which the images were collected (every 2 s). To minimize the background acoustic noise, these RF pulses were delivered whilst the frequency encoding gradient (read gradient) was turned off. This was performed because the ‘read gradient’ is the one that is primarily responsible for the scanner noise, as it is played almost at full amplitude during acquisition. The first 4 s of each 16 + 20 s period was a period of total silence during which the scanner operation was interrupted. Sixteen near‐axial images were collected for each brain volume (thickness = 7 mm, inter‐slice gap = 0.7 mm, in‐plane resolution = 3.75 mm). A hundred brain volumes were collected in total, and the acquisition time of the entire paradigm was 6 min 30 s.

Analysis

Subjective ratings

A Friedman test and post hoc Wilcoxon signed ranks tests were performed in order to determine the effect of experimental condition upon subjective ratings of perceived anxiety, discomfort and the intensity of fear in the facial expressions displayed in each of the ten 20 s blocks for each of the three experimental conditions.

fMRI data analyses

In order to determine the overall effect of OS and fearful context, we obtained individual GBAMs for each of the three conditions (as described for Experiment 1) and then computed average values of the three statistic maps for each subject. A group image from these averaged maps was used to create a mean GBAM to demonstrate neural responses to OS compared with no OS during presentation of fearful facial expressions for all three conditions. We then examined the relationship between subjective ratings and generic power of response (mean FPQ) at each voxel by calculating the Pearson product moment correlation coefficients between these parameters. Observed correlation coefficients were tested against the null distribution obtained by randomizing subjective ratings 10 times at each voxel and recalculating the above correlation coefficients. Combining these data across all voxels yields an estimate of the distribution of the correlation coefficient under the null hypothesis. Using critical values of the correlation coefficient calculated from this distribution, voxels exhibiting correlation coefficients with a cluster‐wise probability under the null hypothesis of <0.05 were identified and reported here (Bullmore et al., 1999b).

To determine the effect of increasing the intensity of fear upon generic neural responses to presentation of fearful faces and OS, we fitted an analysis of covariance model at each intracerebral voxel of the individual standardized power maps after their co‐registration in standard (Talairach) space, as described in Experiment 1, for the following contrasts: condition 1 versus condition 2; condition 1 versus condition 3; and condition 2 versus condition 3.

Results

Subjective ratings of anxiety and discomfort

During presentation of expressions of mild, moderate and high intensity fear, mean ratings (and standard deviations) across all subjects were: anxiety 1.5 (±0.78), 1.7 (±0.9) and 2.5 (±1.2), respectively; perceived discomfort 2.6 (±1.8), 2.7 (±2.0) and 3.9 (±2.0), respectively; and intensity of fear displayed in the facial expressions 1.7 (±1.0), 2.9 (±1.4) and 5.3 (±2.3), respectively. There was a significant effect overall of experimental condition upon subjective ratings of anxiety (P < 0.0001), discomfort (P < 0.028) and the intensity of fear displayed in the preceding facial expressions (P < 0.0001), with:

(i) anxiety being significantly higher during presentation of high compared with either moderate (P < 0.0001) or mild (P < 0.0001) intensity fearful facial expressions;

(ii) perceived discomfort being significantly higher during presentation of high compared with either moderate (P = 0.002) or mild (P < 0.036) intensity fearful facial expressions; and

(iii) ratings of fear intensity significantly higher for high compared with either moderate (P < 0.0001) or mild (P < 0.0001) intensity fearful facial expressions.

Generic brain activation maps of neural responses to OS occurring during presentation of facial expressions of mild, moderate and high intensities of fear

The mean GBAM for all three conditions, demonstrating neural responses to OS compared with no OS during presentation of fearful facial expressions is shown in Fig. 4A. Correlational analyses between mean FPQ and subjective ratings are described below. The comparisons of GBAMs for conditions 1 and 3, 1 and 2, and 2 and 3 (demonstrating neural regions activated to a significantly greater extent in response to high compared with mild‐intensity, high compared with moderate intensity, and moderate compared with mild intensity fearful expressions, respectively) are shown in Fig. 4B–D.

Fig. 4 Generic brain activations are shown for the eight subjects participating in the second experiment representing: (A) the mean neural response to all three conditions; (B) the comparison of neural responses during conditions 1 and 3; (C) during conditions 1 and 2; and (D) during conditions 2 and 3, demonstrating, respectively, regions activated to a significantly greater extent in response to high compared with mild intensity, high compared with moderate intensity, and moderate compared with mild intensity fearful expressions. Brain slices are shown at 31 and 37 mm above the transcallosal plane in A, and at 37 and 42 mm above the transcallosal plane in BD. The right side of the brain is shown on the left side of the picture for each brain slice, and vice versa. In AC, major regions of activation are shown in the dorsal anterior cingulate gyrus. This region was not activated to a significantly greater extent in the comparison of conditions 2 and 3 (D).

Overall effect of non‐painful OS versus no OS in a fearful context

Activation was demonstrated within the right dorsal ACG, and bilateral insulae, in addition to bilateral cerebellum and occipitotemporal cortical regions (Table 2AA and Fig. 4A).

View this table:
Table 2A

Major brain regions activated by non‐painful OS in a fearful context (conditions 1, 2 and 3)

Cerebral regionSidex*y*z*BAFPQSize
CerebellumR21–63–182.0402
L–25–56–291.832
Occipitotemporal regions:
 Superior temporal gyrusR47–74222.0103
L–53–49222.1120
 Middle temporal gyrusR53–13–7211.969
L–53–17–7211.985
 Lingual gyrusL–4–764181.870
 PrecuneusL–7–6942181.847
 CuneusR7–7331191.733
L–4–739171.730
 Posterior cingulate gyrusR4–509301.614
L–17–6015311.626
 Fusiform gyrusL–50–13–24201.726
 Inferior temporal gyrusL–43–17–29201.825
 InsulaR40–792.190
L–43–4–22.176
 Anterior cingulate gyrus (dorsal)R41031241.511

*Footnote as for Table 1A.

Correlational analyses

Mean FPQ correlated positively (P < 0.05) with subjective anxiety ratings within three clusters within right dorsal cingulate gyrus (x = 8, y = –14, z = 39, number of voxels = 59) and bilateral posterior cingulate gyri (x = –10, y = –29, z = 36; x = 17, y = –33, z = 20; number of voxels = 20 and 18, respectively), with subjective discomfort ratings in two clusters within right lingual gyrus (x = 19, y = –59, z = –1, number of voxels = 96) and right posterior cingulate gyrus (x = 13, y = –40, z = 42; number of activated voxels = 14), and with subjective ratings of fear intensity in right lingual gyrus (x = 26, y = –76, z = –1).

Comparison of GBAMs for the three conditions

Major regions activated to a significantly greater extent during presentation of facial expressions of high compared with mild intensity fear were demonstrated within the dorsal region of the left ACG and bilateral anterior insulae, in addition to bilateral dorsolateral prefrontal cortices. Few regions activated to a significantly greater extent in response to expressions of mild compared with high intensity fear. These included the left cerebellum and right posterior but not anterior insula (P = 0.01; overall search volume = 2222 voxels; expected number of false positive activated voxels = 22; number of observed activated voxels = 9; Table 2BB and Fig. 4B).

View this table:
Table 2B

Major brain regions activated by OS with facial expressions of high compared with mild intensity fear

Cerebral regionSidex*y*z*BAFPQSize
High>mild:
 Anterior cingulate gyrus (dorsal)L–32242321.412
 Dorsolateral prefrontal cortexR292515451.46
L–322215451.24
 Anterior insulaR292591.56
L–292291.26
Mild>high:
 CerebellumL–23–44–351.47
 Posterior insulaR38–8151.25

*Talairach co‐ordinates refer to the voxel with the maximum FPQ in each cluster. P < 0.01 for comparison of generic activation in the two conditions. L = left; R = right.

Major regions activated to a significantly greater extent during presentation of expressions of high compared with moderate intensity fear were demonstrated within the left dorsal and right ventral ACG and right anterior insula, in addition to bilateral dorsolateral and ventromedial prefrontal cortices. Regions activated to a significantly greater extent in response to facial expressions of moderate compared with high intensity fear included bilateral cerebellum, left hippocampus and left posterior but not anterior insula (P = 0.01; overall search volume = 4145 voxels; expected number of false positive activated voxels = 41; number of observed activated voxels = 219; Table 2CC and Fig. 4C).

View this table:
Table 2C

Major brain regions activated by OS with facial expressions of high compared with moderate intensity fear

Cerebral regionSidex*y*z*BAFPQSize
High>moderate:
 Dorsal prefrontal cortexR292815451.610
 Anterior cingulate gyrus (ventral)R336–2241.59
 Anterior cingulate gyrus (dorsal)L–32242321.37
 Anterior insulaR292591.49
 Ventromedial prefrontal cortexR/L036–13111.26
Moderate>high:
 CerebellumR6–36–181.118
L–3–58–131.17
 Posterior insulaL–43–3–71.114
 HippocampusL–29–17–71.15

*Footnote as for Table 2B.

Major regions activated to a significantly greater extent during presentation of expressions of moderate compared with mild intensity fear were demonstrated bilaterally within posterior but not anterior insula, middle temporal gyrus, cerebellum and the left hippocampus. Only the left cerebellum was activated to a significantly greater extent in response to expressions of mild compared with moderate intensity fear (P = 0.01; overall search volume = 3985 voxels; expected number of false positive activated voxels = 39; number of observed activated voxels = 199; Table 2DD and Fig. 4D).

View this table:
Table 2D

Major brain regions activated by OS with facial expressions of moderate compared with mild intensity fear

Cerebral regionSidex*y*z*BAFPQSize
Moderate>mild:
 Posterior insulaR40–3–71.07
L–43–11–21.019
 Middle temporal gyrusR38–6–24211.011
L–46–11–13211.05
 CerebellumR12–39–181.09
L–26–47–351.411
 HippocampusL–29–17–71.19
Mild>moderate:
 CerebellumL–26–47–351.411

*Footnote as for Table 2B.

Discussion

We report the first demonstration of a significant effect of a negative emotional context upon subjective and neural responses to phasic, periodic non‐painful OS. In the first experiment, we demonstrated that whilst the two factors, OS and presentation of fearful facial expressions were each associated with activation within dorsal and ventral ACG, there was a significant interaction between these factors. Specifically, there was a significant effect of negative emotional context upon neural responses during OS, with the dorsal region of the left ACG activated to a significantly greater extent during OS in a negative emotional context (condition A) than during OS in a neutral context (condition C). In contrast, there was no significant effect of the addition of OS upon neural responses to a negative emotional context (fearful facial expressions) i.e. the contrast of conditions D and B).

In the second experiment, we investigated the effect of altering the intensity of fearful facial expressions upon neural responses to phasic, periodic non‐painful OS. We demonstrated that, whilst OS occurring in a negative emotional context per se was associated with activation within the dorsal ACG and bilateral insulae, significantly greater activation was demonstrated within the dorsal ACG and the anterior insula during a high intensity than during either a mild or moderate intensity negative context. However, there was no significant difference in activation within these regions during the moderate and mild intensity negative contexts. In this experiment, we were also able to demonstrate increased subjective ratings of anxiety and perceived discomfort with increased intensity of the negative emotional context, and a positive correlation between subjective ratings of anxiety and the intensity of dorsal cingulate gyral activation.

We also observed a positive correlation between both anxiety and discomfort ratings with activation of the posterior cingulate cortex (PCC). Activation of PCC has been reported in several previous studies of somatic pain; its role, however, remains uncertain. It is clear that it is more reproducible activated when using a phasic rather than tonic stimulus (Derbyshire, 2000). Gelnar and colleagues reported activation of PCC in response to thermal pain and suggested that this region, lying between the mid‐cingulate motor area and caudal visuospatial region, is a somatosensory area that receives direct nociceptive projections from the spinothalamic tract (Gelnar et al., 1999).

The consistent finding from both experiments was a significant increase in activation within dorsal ACG with increasing intensity of the negative emotional context in which constant intensity, non‐salient (non‐painful) visceral stimulation occurred, whereas the findings from the first experiment indicate no significant effect of the addition of OS upon neural responses to a constant intensity negative emotional context (condition D). Together, these findings suggest a modulatory effect of emotional context upon the dorsal ACG response to OS rather than a simple additive effect of emotional context and OS upon the response in this region.

Whilst the role of the dorsal region of the ACG in the modulation of attention and executive functions has been highlighted previously (Bush et al., 2000), there is also increasing evidence for the role of this region in the processing of emotionally salient information. Cingulotomy, involving lesions to this region, has been demonstrated to be a successful treatment for patients with major affective and anxiety disorders (Ballantine et al., 1967; Cosgrove and Rauch, 1995) and chronic pain disorders (Devinsky et al., 1995). Increased regional cerebral blood flow to rostral and dorsal regions of the ACG has been reported during attention to subjective emotional states and experiences (Lane et al., 1997, 1998; Gusnard et al., 2001). These findings indicate that the dorsal anterior cingulate gyrus is important for the direction of attention to internal emotional and sensory states.

In the second experiment, the anterior insula was activated to a significantly greater extent during presentation of high compared with either moderate or mild intensity fearful facial expressions. Functional brain imaging studies have also highlighted the importance of the anterior insula in mediating negative emotions (Phillips et al., 1997, 1998; Drevets and Raichle, 1998; Mayberg et al., 1999) and, with the dorsal ACG, in the response to painful and non painful visceral (Silverman et al., 1997; Aziz et al., 2000; Mertz et al., 2000) and painful somatic sensation (Coghill et al., 1994; Vogt et al., 1996; Davis et al., 1997; Derbyshire et al., 1997; Derbyshire and Jones, 1998; Casey, 1999).

The insular cortex is also an important area for coordinating visceral sensory and motor information and is involved in autonomic regulation (Augustine, 1996). It has a crude viscerotopic representation for the gastrointestinal and cardiovascular afferents (Cechetto and Saper, 1987, 1990). Functional mapping studies of the insula cortex using intracortical electrodes in patients with temporal lobe epilepsy have shown that the posterior/mid insula region is part of a somesthetic network involved in processing painful and non‐painful somatic sensation (Ostrowsky et al., 2000). In contrast, stimulation of the anterior insula in the same subjects elicited viscerosensory and visceromotor responses indicating that the anterior insula—with its dense projections to the piriform cortex, orbitofrontal cortex, hippocampus and amygdala—is part of a visceral network (Ostrowsky et al., 2000). Activation of the insular cortex is observed routinely in functional brain imaging studies of visceral sensation, with the anterior insula being activated more intensely in response to noxious visceral stimulation in comparison to non‐noxious stimulation (Aziz et al., 1997, 2000; Aziz and Thompson, 1998). Activation of the anterior insular cortex is also noted in studies of somatic pain especially when accompanied by a strong emotional response (Hsieh et al., 1995; Rainville et al., 2001) and lesions of the anterior insula have been shown to reduce the affective response to pain (Berthier et al., 1988). In addition, animal studies have shown afferents from the anterior insula project to ventral and dorsal regions of the anterior cingulate cortex (Augustine, 1996). These data support our findings that anterior insula and dorsal cingulate play an important role in processing and modulating visceral sensation and that these regions may belong to a more extensive cortical network involved in integrating emotional and visceral information.

Our findings from both experiments indicate that presentation of a negative emotional context during experience of otherwise non‐painful visceral stimulation is associated with activation within neural regions important for performance of attentional tasks, and for attention to internal emotional and sensory states. This activation increases with greater intensity of the negative context. We have further demonstrated in the second experiment that subjective experiences of anxiety and discomfort during non‐painful OS also increase with greater intensity of the negative context. Taken together, these findings indicate that when non‐painful visceral stimulation, otherwise associated with low attentional and emotional processing demands, occurs in an increasingly negative context, attentional and emotional processing demands increase, with corresponding increases in subjective anxiety and perceived discomfort, and activation within neural regions important for attentional and emotional processing.

In the first experiment, there was no significant increase in activation within the dorsal ACG and anterior insula when continual phasic, non‐painful OS occurred during presentation of alternating blocks of fearful and neutral faces (condition D) than when presentation of the faces occurred in the absence of OS (condition B). One possibility is that attenuation of response may have occurred within regions important for visceral sensory perception during condition D, so that significant increases in activation with the dorsal ACG and anterior insula were not demonstrated in this condition compared with condition B. Indeed, in condition D, subjects experienced phasic, non‐painful OS on 100 occasions throughout the 5 min period. Previous studies have indicated that attenuation of the cortical evoked response to oesophageal stimulation occurs during continuous stimulation runs consisting of greater than 50 stimuli at a time. (Hobson et al., 1998). Alternatively, continual rather than episodic presentation of fearful faces throughout the experimental condition may be required to increase attentional and sensory processing demands during otherwise non‐salient visceral stimulation.

We demonstrated a significant effect of experimental condition upon activation within ventral ACG in the second experiment during OS occurring in high versus moderate intensity fearful expressions, but not during OS occurring in high versus mild intensity fearful expressions. In contrast, we did not demonstrate a significant effect of experimental condition upon activation within ventral ACG in the first experiment. Ventral (perigenual and subgenual) regions of the ACG have been associated with the experience of negative emotional states (Drevets and Raichle, 1998; Mayberg et al., 1999) and emotion perception (Bush et al., 2000). In the second experiment, we were able to demonstrate that increasing the intensity of fear displayed by facial expressions viewed by subjects during OS resulted in significant increases in the experience of oesophageal discomfort and anxiety. However, viewing faces depicting negative emotional expressions has not been considered to be an adequate method of inducing emotional states in previous functional neuroimaging studies. Instead, other methods, for example, recollection of autobiographical memories, have been employed (Mayberg et al., 1999). It is possible, therefore, that a more persistent change in emotional state resulting from the use of mood induction paradigms may further modulate the intensity of subjective anxiety and would lead to the induction of negative mood and more robust activation within ventral ACG.

Other regions activated in response to periodic, OS and to presentation of fearful faces per se in the first experiment included dorsolateral and ventral prefrontal cortices, regions associated, respectively, with performance of cognitively‐demanding and decision‐making tasks (e.g. Bechara et al., 1998). In both experiments, activation within occipitotemporal cortical regions was demonstrated in response to periodic, OS occurring in fearful contexts, and, in the first experiment, in response to presentation of fearful faces. In the second experiment, we also observed a positive correlation between subjective rating of fear intensity and right lingual gyrus activation. In previous studies, greater activation within occipitotemporal cortex has been demonstrated in response to emotionally salient than to neutral visual stimuli (e.g. Lang et al., 1998; Taylor et al., 2000). Our findings therefore suggest that the combination of non‐painful OS and presentation of fearful faces was perceived as more salient than fearful faces alone, whilst fearful faces per se were perceived as more salient than neutral faces. There was, however, no consistent increase in activation within either prefrontal or occipitotemporal cortical regions in response to increasing the intensity of the negative emotional context in which periodic, OS occurred in the second experiment.

In the first experiment, periodic, OS and presentation of fearful faces per se were both associated with activation within the hippocampus. These findings are consistent with those of previous studies in which activation within the hippocampus has been reported in response to negative emotional visual and auditory stimuli (e.g. Phillips et al., 1998) and with the postulated role of this region as a comparator to match or compare novel, including emotionally‐salient, stimuli (e.g. fearful faces or OS) with a stored template of previously‐processed, familiar stimuli (e.g. neutral faces or no OS; Gray, 1982). The hippocampus was also activated to a significantly greater extent in response to presentations of fearful contrasted with neutral faces (condition B) alone than to presentation of these facial expressions with continuous OS (condition D). This finding is suggestive of an attenuated neural response to novelty during condition D, as discussed previously.

In this study, we were interested in examining the effect of presentation of a negative emotional context upon neural responses to visceral stimulation. Another possibility for future studies is to examine the effect of increasing arousal per se upon neural responses to visceral stimulation by presentation of a novel rather than specifically negative context upon responses to visceral stimulation, using facial expressions depicting other negative and positive emotions.

Conclusions

To date, there has been little investigation of the relationship between the type of emotional context in which sensory stimulation occurs and the extent of activation in brain regions such as the dorsal ACG and anterior insula that mediate attention, sensory perception and emotions. We report the first evidence for a modulatory effect of emotional context upon the extent of activation within these regions and subjective reports of anxiety and discomfort during phasic, non‐painful visceral stimulation. The findings of this study may provide some insight into the potential importance of emotional context in the management and treatment of functional gastrointestinal disorders such as irritable bowel syndrome and non‐cardiac chest pain.

Acknowledgements

Q.A., S.C. and funding for this project were supported by the Medical Research Council (UK). A.H. is funded by the Lord Dowding Fund for Humane Research.

References

View Abstract