In the present study we sought to investigate the impact of the COMT val158met single nucleotide polymorphism on cerebral pain processing as measured by fMRI.
The overall BOLD activation pattern in response to painful laser stimulation in our study is consistent with numerous previous imaging studies showing activation of a distributed network of cortical and subcortical structures including the core regions of cerebral pain processing (often referred to as the 'pain matrix') such as the bilateral parietal operculum containing the secondary somato-sensory cortex (S2), bilateral insula and bilateral anterior cingulate cortex (ACC)/mid-cingulate cortex (MCC) [5, 6, 17].
Subjects homozygous for the met158 allele exhibited a higher BOLD response to laser stimulation than carriers of the val158 allele (homozygous and heterozygous val158 carriers combined) in key areas of the 'pain matrix' - i.e. in S2, the insula the amygdalae and foremost in the ACC. In the voxel-by-voxel analysis restricted to the ACC we found the most significant activation difference between genotype groups in the posterior portion of the ACC/the MCC.
The ACC has been shown to be a key structure of cortical pain processing that is involved in the cognitive/emotional evaluation of pain as well as in antinociception [5, 6, 18–21]. Furthermore, Büchel et al.  showed that activation in the posterior ACC/MCC is correlated with stimulus intensity and pain ratings. Therefore our data are consistent with previous studies that found an association of the COMT met158 allele with higher pain ratings [10, 13] and with altered μ-opioid receptor density and opioid responses to pain in several brain areas including the dorsal ACC . Our results add to the current literature in that they suggest that COMT-dependent ACC/MCC activity plays a prominent role in mediating sensitivity to pain, possibly due to a reduction of opioid-mediated inhibitory control. However, it needs to be pointed out that negative findings with regard to the effect of this SNP on pain-related phenotypes have also been reported [14, 15]. This discrepancy may be explained by sample heterogeneity, differences in the modality used to elicit pain and by differences in the behavioral readout. Furthermore it has been suggested that haplotypes that also include other SNPs in COMT may exert a stronger effect on pain-related phenotypes [22, 23] than the COMT val158met polymorphism alone. We observed an effect of COMT genotype on brain activation in the absence of any behavioral effects (subjective pain ratings). This discrepancy may be explained by different effect sizes of imaging endophenotypes and behavioral phenotypes with larger genotype effects on brain activation than on overt behavior [16, 24].
The COMT val158met polymorphism may affect pain processing in several ways. The met158 allele codes for an enzyme variant that is less stable at body temperature leading to higher dopamine levels in the brain, mainly in - but not restricted to - the (pre)frontal cortex. Dopamine levels may modulate the activity of the endogenous opioid system indirectly by regulating the neuronal content of enkephalins [10, 25, 26]. This could affect activation measures in brain areas involved in processing and modulation of painful stimuli and ultimately behavioral pain-related phenotypes . Alternatively/additionally, COMT may affect pain processing by decreasing the metabolism of epinephrine which has also been shown to modulate pain processing. For instance, Khasar et al.  showed that β2-adrenergic stimulation induces hyperalgesia in the rat. Furthermore it has been shown that a COMT inhibition-induced increase in pain sensitivity is blocked by beta-adrenergic antagonists . Yet another alternative may be that the COMT val158met SNP affects brain activity in response to painful stimulation in a rather unspecific way: a substantial amount of genetic imaging papers found an impact of this SNP on brain activity in various cognitive domains such as attention, working memory and affective regulation . The genotype effect on pain processing that we found in the present study may therefore be 'downstream' of a more direct influence on attention or affect regulation. For instance, Smolka et al.  showed that compared to carriers of the val allele subjects homozygous for the met allele exhibit an increased BOLD response to unpleasant pictures in a partially overlapping network of structures including the amygdala. Therefore homozygous met carriers may be more reactive to a variety of negative stimuli including pain. However, activation in the dorsal ACC/MCC - the site of the most significant genotype effect in our study and a key structure of the pain matrix - was not affected by COMT genotype in the study by Smolka et al. This region of the ACC has been implicated in several aspects of pain processing such as encoding pain- and stimulus intensity . Furthermore the dorsal ACC has been shown to be one of the sites of pain-induced opioid release . In a subsequent paper Zubieta et al.  showed that the COMT val158met polymorphism affects the μ-opioid response to painful stimulation in the dorsal ACC. Last, it has been shown that the ACC/MCC is the main generator of the N2 and P2 laser-evoked potentials in studies using EEG source localization , intracranial recordings  or more recently EEG-informed fMRI . All this may suggest that the observed effect of COMT genotype on fMRI BOLD activation is not entirely unspecific. However, it has recently been shown that laser-evoked potentials themselves are not nociceptive-specific .
The genetic basis of pain perception and processing has mostly been studied in clinical settings such as chronic pain states, post-surgical pain or experimental settings in which tonic pain models were applied. Correspondingly, an effect of COMT genotype on pain-related phenotypes has for instance been shown in cancer patients with regard to analgesic requirements [12, 35] and in experiments in which models of sustained pain such as the intramuscular injection of hypertonic saline ; tonic heat pain  or the temporal summation of thermal pain were studied . By contrast laser stimulation of the skin as applied in our study induces phasic pain which is quite different from tonic pain models or clinical pain states. An effect of COMT genotype on phasic pain has not been reported so far. Therefore our findings require further replication and cannot readily be transferred to clinical pain states and their genetic basis.
Our study has several limitations. First, our sample comprised only 9 subjects with COMT val/val genotype, a number that we considered to low to constitute a separate group in our genetic fMRI analyses. Thus, we combined homozygous and heterozygous val158 carriers in one group which leaves the question of additive vs. dominant/recessive gene effects unanswered. Second, due to technical reasons we did not obtain continuous online pain ratings, but only post-hoc ratings which are considered less accurate. This could explain the lack of a relationship between pain ratings on one hand and COMT genotype and fMRI BOLD activation on the other hand. Then again genotype effects are well known to require comparatively large samples to be detected when behavioral measures and probably even more so subjective measures like pain ratings are used which is one of the main reasons why intermediate phenotypes like brain activation as measured by fMRI are studied . Third, our paradigm did not include a control condition with innoxious stimuli. While laser stimulation of the skin provides Aδ and C fibre-mediated nociceptive-specific input to the brain, cerebral processing of that input has been shown not to be pain-specific . Therefore we cannot rule out the possibility that the observed differences between COMT genotype groups in the BOLD activation pattern reflect a more general genotype effect on brain activity especially as COMT affects several aspects of human behavior including, attention working memory and emotional regulation . Lastly, we investigated a rather heterogeneous sample, which is reflected by the high impact of age on BOLD activation measures in our study. Heterogeneity of the sample may increase the background variance in the data diminishing the effect of the genetic variant that is under investigation . On the other hand, the genotype effect survived age correction and we found an impact of the COMT158val/met on ACC/MCC activation in a sample that was largely selected from a population-based sample even after age correction. The latter may in fact be considered a plus with respect to the generalizability of the results of a genetic imaging study.