Participants

Imaging paradigm and study design

Participants took part in a 5-minute experiment employing event-related fMRI, and were presented with ten different facial identities (six female, four male) (http://www.paulekman.com) depicting 150% intensity of sadness. We used 150% intensity in order to ensure optimal performance in patients and to avoid possible confounding effects of ambiguity with regards to facial emotion recognition. In addition, a neutral expression was used as a control condition. Images of sad and neutral facial expressions were interspersed with a fixation cross. Each stimulus was displayed for 2 s. In all, 60 images were displayed in a random order; the fixation cross, faces with neutral expression, and faces showing affect were each displayed 20 times. Participants were instructed to respond to sad and neutral faces by pressing the right and left button respectively on an MRI-compatible response box. No response was required when participants viewed the fixation cross. Response time and accuracy data were collected.

Healthy controls were scanned once while patients were scanned twice; at the end of the washout period (baseline) and upon completion of 12 weeks of lamotrigine monotherapy. All participants performed the same task as described above.

Image acquisition and analysis

Gradient echo echoplanar magnetic resonance images were acquired for each participant on each occasion of scanning using a 1.5 T General Electric Neurovascular Signa magnetic resonance system (General Electric, Milwaukee, Wisconsin, USA) fitted with 40 mT/m high speed gradients, with foam padding and a forehead strap to limit head motion. A quadrature birdcage head coil was used for radio frequency transmission and reception. In each of the 36 non-contiguous planes parallel to the inter-commissural (anterior commissure–posterior commissure) plane, T2*-weighted magnetic resonance images depicting blood oxygen level-dependent (BOLD) contrast were acquired with time to echo (TE) 40 ms, a repetition time (TR) 2000 ms, slice thickness 7 mm, slice skip 0.7 mm. During the 5-minute facial affect discrimination task, 150 images were collected. At the same session, a high-resolution gradient echoplanar imaging data-set was acquired for subsequent co-registration.

Behavioural data

The mean accuracy of controls and patients at baseline was 38.3 (s.d.=1.18) and 33.8 (s.d.=5.27) respectively, and following lamotrigine treatment patients' mean accuracy increased to 35.2 (s.d.=4.94). There were no significant differences between patients and controls in accuracy of sad facial affect recognition (t=2.36, d.f.=14, P=0.55). Patients' accuracy did not differ compared with their baseline following lamotrigine treatment (t=7.49, d.f.=7, P=0.13).

The mean response time of controls and patients at baseline was 1.17 s (s.d.=0.19) and 1.26 s (s.d.=0.22) respectively; following lamotrigine treatment, patients' mean response time was 1.23 s (s.d.= 0.23). There were no significant differences between patients and controls in the response time (t=–1.00, d.f.=14, P=0.33). Patients' response times did not change compared with their baseline following lamotrigine treatment (t=0.76, d.f.=7, P=0.47).

Imaging data

Distribution of brain activation in control participants and patients with bipolar disorder at baseline and post lamotrigine treatment Both groups at baseline displayed activation in the middle frontal gyrus (Brodmann area; BA 46), temporal regions and cerebellum. In addition, controls activated the medial frontal gyrus (BA 6) while patients activated parietal regions. After lamotrigine treatment, patients displayed additional activation in the medial frontal gyrus (BA 6), inferior frontal gyrus (BA 44), postcentral gyrus (BA 3), medial occipital gyrus (BA 19) and the thalamus. Detailed coordinates of the regional peaks are shown in Table 1.

Table 1 Regions of brain activation during sad facial affect recognition at baseline and post-lamotrigine treatment.

		Controls				Patients with bipolar disorder at baseline				Patients with bipolar disorder post-lamotrigine treatment
		Left		Right		Left		Right		Left		Right
Region	BA	x, y, za	t	x, y, za	t	x, y, za	t	x, y, za	t	x, y, za	t	x, y, za	t
Superior frontal gyrus	6											4, 6, 64	9.05
Middle frontal gyrus	10											34, 58, -8	13.82
Middle frontal gyrus	46			48, 40, 14	6.01	-44, 24, 24	3.46
	9					-38, 8, 34	14.14	40, 16, 32	5.64
Inferior frontal gyrus	44									-42, 42, 2	5.64
Medial frontal gyrus	6			6, 16, 44	9.28							4, -18, 68	8.75
Superior temporal gyrus	22			68, -44, 16	4.63
Middle temporal gyrus	22	-48, 44, 6	9.79
	20			66, -34, -14	3.11
Inferior Temporal gyrus	37					-50, -80, 2	4.70					54, -72, 0	10.28
Superior parietal lobule	7					-32, -54, 62	8.87
	40							34, -54, 48	3.16
Inferior parietal lobule	40					-54, -40, 34	3.55
Postcentral gyrus	3									-28, -28, 64	9.56
Medial occipital gyrus	19									-50, -80, 8	9.22
Thalamus								20, -24, 0	4.55	-2, -18, 10	5.34	2, -24, 4	5.26
Cerebellum		-42, -50, -28	8.49			-36, -54, -24	4.03	40, -64, -28	4.54

Patient–control comparison at baseline

Compared with controls, greater activation was observed in patients in the parahippocampal gyrus. However, controls showed significantly more activation mostly right-sided, in the superior (BA 6) and inferior (BA 47) frontal gyri, and the precentral (BA 4) and cingulate (BA 23) gyri. Additionally, increased activation relative to patients was also seen in the left middle frontal gyrus (BA 6). Detailed coordinates of the regional peaks are shown in Table 2.

Table 2 Regional peak activations during the sad facial affect recognition task: case– control comparison at baseline.

		Left		Right
Region	BA	x, y, za	t	x, y, za	t
Controls>patients
Superior frontal gyrus	6			12, -4, 68	3.50
Middle frontal gyrus	6	-42, -2, 50	2.50
Inferior frontal gyrus	47			30, 30, -18	3.09
Precentral gyrus	4			28, -24, 72	3.14
Cingulate gyrus	23			6, -36, 28	5.64
Patients>controls
Parahippocampul gyrus		-26, -22, -20	2.62

Differences in brain activation patterns in patients with bipolar disorder between baseline and following 12 weeks of lamotrigine monotherapy

Greater activation was observed at baseline compared with study endpoint in the right precentral gyrus (BA 4) and the left inferior temporal gyrus (BA 20).

Following 12 weeks of lamotrigine monotherapy, increased activation compared with baseline was noted bilaterally in the inferior frontal gyri (BA 44, 45) in the left precentral gyrus (BA 6) and on the right in the medial frontal gyrus (BA 6), the paracentral lobule (BA 4) and the thalamus. Detailed coordinates of the regional peaks are shown in Table 3 and illustrated in the online Fig. DS1. Figure 1 shows the mean signal intensity in the areas of treatment effect at baseline and after lamotrigine treatment, extracted with the MarsBar tool of SPM2.

Fig. 1 Mean signal intensity changes in regions of significant treatment effect at baseline and post lamotrigine treatment. GFd, medial frontal gyrus; GPrC, presentral gyrus; Lpc, paracentral lobule; GFi, inferior frontal gyrus; Th, thalamus.

Table 3 Regional peak activations during the sad facial affect recognition task.

		Left		Right
Region	BA	x, y, za	t	x, y, za	t
Regions of increased activation at baseline relative to post-lamotrigine treatment
Precentral gyrus	4			60, -16, 36	10.35
Inferior temporal gyrus	20	-58, -12, -20	8.74
Regions of increased activation post-lamotrigine treatment relative to baseline
Medial frontal gyrus	6			4, -16, 70	6.14
Precentral gyrus	6	-28, -14, 68	5.64
Paracentral lobule	4			2, -34, 76	4.10
Inferior frontal gyrus	45			42, 32, 0	3.26
Inferior frontal gyrus	44	-42, 40, 0	3.82
Thalamus	N/A			14, -20, 8	4.75

Effect of task

Sad facial affect recognition paradigms similar to that employed here activate a number of brain regions in healthy individuals within the dorsomedial prefrontal cortex, the cingulate and parahippocampal gyri as well as subcortical regions such as the thalamus and the caudate, although variability has been noted across studies. ^{Reference Phan, Wager, Taylor and Liberzon8,Reference Murphy, Nimmo-Smith and Lawrence9} In this study, unmedicated patients with bipolar disorder and healthy controls showed a similar pattern of activation during the facial affect recognition task but differed in the degree of neural response in several brain regions.

Effect of medication

Dorsomedial (BA 6) and ventrolateral (BA 47) prefrontal cortical regions were more activated in controls than in patients as was the right dorsal cingulate gyrus (BA 23). Patients showed greater activation within temporal regions around the left hippocampus/parahippocampal gyrus both compared with controls at baseline, but also compared with their post-lamotrigine pattern of activation. This finding is in partial agreement with that of Lawrence et al. ^{Reference Lawrence, Williams, Surguladze, Giampietro, Brammer, Andrew, Frangou, Ecker and Phillips10} They reported a positive correlation between the degree of depression and activation within the left hippocampus, which accords with the increased neural response in the same region observed in this study. Increased activity in temporal lobe structures is commonly seen during processing of negative, sadness-inducing material in healthy people (e.g. ‘dressing to go to their mother's funeral’) ^{Reference Partiot, Grafman, Sadato, Wachs and Hallett15} and has also been implicated in negative affect disturbances in individuals with depression. ^{Reference Kumari, Mitterschiffthaler, Teasdale, Malhi, Brown, Giampietro, Brammer, Poon, Simmons, Williams, Checkley and Sharma16} We observed a 25% reduction in HRSD scores at follow-up in the current sample which was not statistically significant but may still contribute to the changes in neural responses in the temporal lobe as fMRI is likely to have greater sensitivity to detect this effect.

Our results are in the opposite direction with regard to prefrontal and anterior cingulate regions where, at baseline, we found higher activation in controls and not in patients. This difference was noted for facial expressions representing high intensity of sadness and patients with bipolar disorder showed no behavioural deficits in sad affect recognition. The two studies have similar patient groups with comparable levels of residual depressive features and minimal manic symptoms. Although we employed an explicit as opposed to an implicit task, it is unlikely that this accounts for the disparity: Chen et al ^{Reference Chen, Lennox, Jacob, Calder, Lupson, Bisbrown-Chippendale, Suckling and Bullmore17} found no difference in the pattern of activation in patients with depression and bipolar disorder using the two different types of tasks, although implicit processing was generally associated with greater neural response in both patients and controls. A significant difference with our study is the medication status of patients. Patients with bipolar disorder in the study by Lawrence et al ^{Reference Lawrence, Williams, Surguladze, Giampietro, Brammer, Andrew, Frangou, Ecker and Phillips10} were medicated with a combination of antidepressants, atypical antipsychotics, lithium and anticonvulsants, although only one patient was on lamotrigine. However, it is difficult to argue for it leading to overactivity since similarly medicated patients with bipolar disorder in the study by Chen et al ^{Reference Chen, Lennox, Jacob, Calder, Lupson, Bisbrown-Chippendale, Suckling and Bullmore17} were not different to controls in their neural response to sad facial expressions.

Increased neural responses following lamotrigine were seen in dorsomedial and ventrolateral prefrontal cortical regions that are known to be reciprocally connected to limbic (cingulate gyrus, amygdale–hippocampus complex) and subcortical regions (basal ganglia, thalamus, insula, brainstem) involved in generation and regulation of emotional states. ^{Reference Cardinal, Parkinson, Hall and Everitt18} The results of this study could therefore be consistent with the idea that lamotrigine treatment may lead to a ‘normalisation’ in key prefrontal regions associated with emotional self-regulation akin to what has been observed with successful remission of depression. ^{Reference Mayberg19}

Limitations

However, it is possible that the enhanced cortical activation seen at study endpoint may relate to factors other than medication, such as learning or habituation. Patients' behavioural performance at baseline and study end was very similar and it would be difficult to argue for an effect of learning. Familiarity with the experimental set-up at follow-up may have been associated with less anticipatory anxiety or arousal. However, habituation is routinely associated with decreased rather than increased neural responses. ^{Reference Wright, Fischer, Whalen, McInerney, Shin and Rauch20} Similarly, the level of symptoms between baseline and study endpoint were comparable and it is therefore difficult to argue that the changes in the brain activation patterns observed are consequent to significant clinical change, although symptomatic improvement may have contributed. In fact, we found no significant correlations between lamotrigine-induced changes in brain activation and differences in HRSD score between study entry and endpoint. The power of the two-by-eight data-set in this study is possibly sensitive to spurious findings and therefore the results require replication. Our findings, if replicated, may apply to clinical populations only as the effect of lamotrigine on brain function in healthy individuals was not assessed. Lamotrigine may have a unique impact on disease-related mechanisms and therefore its effect on healthy individuals may not have been necessarily informative with regards to its mechanism of action. Lamotrigine appeared to ‘normalise’ activation within the neural networks involved in facial affect processing by enhancing activation in prefrontal and reducing it in temporal regions. The clinical implications of this observation require further study.