A key role of the basal ganglia in pain and analgesia - insights gained through human functional imaging

Early studies of pain using fMRI indicated activation patterns in the BG including the putamen and globus pallidus. These early observations have now been replicated in a number of functional imaging studies of experimental pain in humans, examples of which are shown in Tables 1 and 2. As noted in the tables, a considerable similarity exists in the activation patterns from multiple stimuli, including thermal (heat and cold), mechanical (in a hyperalgesic capsaicin model), painful electrical and visceral pain. Subsequent studies showed what seemed to be specific regional activation in other BG structures including the nucleus accumbens [45] and putamen [46, 47]. However, most studies have evaluated or reported activation patterns and observed striatal involvement in a general context.

Tables 1 and 2 show positive activation is present in the caudate nucleus across multiple studies and across multiple modalities. In addition to those mentioned in the tables, activation in the caudate is present in "prickle" sensation evoked by cold stimuli [48]. In our own studies, cold has previously been shown to activate basal ganglia including the caudate [49]. Activation in the caudate nucleus to noxious stimulation has been suggested to be part of a pain modulatory system [50, 51]. In support of this, electrical stimulation of the caudate in non-human primates diminished pain reactivity [52]. The authors suggested that the results indicate that the effect of caudate stimulation is to reduce the affective components of pain elicited by noxious electrocutaneous stimuli.

A few studies have evaluated specific activation in the putamen and nucleus accumbens. As noted in Tables 1 and 2, the putamen is commonly activated across most acute pain imaging studies. Pain activated the putamen bilaterally and a somatotopic organization for hand- and foot-related responses was only present in the contralateral putamen [53]. In healthy women, pain-evoked putaminal activation occurred during their follicular phase [54].

Activation of globus pallidus following painful stimulation has been shown in healthy subjects (Tables 1 and 2). Unlike the caudate and putamen, fewer studies have specifically reported activation in this BG structure. However, like the caudate and putamen, the globus pallidus has neurons that respond to noxious stimulation [9].

The nucleus accumbens (NAc), a brain substrate known to be involved in reward-aversion processing, has been shown to respond with opposite BOLD signal valence to rewarding or aversive stimuli [45, 55]. Using a pain onset (aversive) and pain offset (rewarding) prolonged stimulus, a negative signal change with pain onset and a positive signal change with pain offset was observed in the NAc contralateral to the stimulus. The study supports the idea that the NAc fMRI signal may provide a useful marker for the effects of pain and analgesia in healthy volunteers. A parallel study on NAc activation in response to the direct effects of the analgesic (pain offset) morphine, showed the same increased (rewarding) BOLD signal in the structure [56]. Direct afferents from the trigeminal nucleus to the nucleus accumbens have been demonstrated in the rat; these have contralateral projections from lamina I but bilateral with contralateral predominance from lamina V [57]. These anatomical studies seem to agree with other studies showing contralateral activation in the NAc [45, 55]. Subsequent studies indicated two components of activation within the NAc - an anterior, superior, and lateral component and another component that was posterior, inferior, and medial within the structure. The anatomical segregation may correlate with the functional components of the NAc (i.e., shell and core) that have been defined in other species (see [58]). The results support heterogeneity of function within the NAc and have implications for understanding the contribution of NAc function in processing of pain and analgesia [58]. The NAc has a pivotal role in aversive and rewarding (hedonic) processing [59].

Given that the reward system is part of the pain network [45], and that the placebo response is clearly involved in analgesia [60, 61], it is not surprising that functional imaging has been the ideal technology to better understand neural networks involved in placebo in humans [62, 63]. Recent reviews suggest that multiple brain regions (e.g., anterior cingulate cortex, anterior insula, prefrontal cortex and periaqueductal grey) are important in the placebo response [64]. Other studies strongly implicate the NAc and ventral BG in the placebo response [65].

The putamen, a structure that shows somatotopic activation to pain [53], is consistently activated during acute and chronic pain conditions and is affected by analgesic administration. Chronic pain has also been shown to increase putamen volume (see [67]) suggesting, perhaps, that a continuous drive producing increased activation may result in processes that enhance gray matter volume in the structure. This region may receive direct inputs from sensory systems, but also from cortical inputs. Some studies (e.g., [83]) have suggested delayed functional activation in BG circuits (e.g., caudate) following remifentanil, implicating that activation in these regions occurs as a consequence or subsequent to initial activation of thalamo-cortical circuits. Activation seems to be present with painful stimuli but missing with non-painful stimuli [95].

In contrast to increases in putaminal activation for acute pain, BOLD signals are opposite in chronic pain across multiple pain studies. The reason for this observation is not known although the nature of reward processing in acute vs. chronic pain is clearly different [45, 59, 71].

Only a few brain imaging studies have reported activation in the smaller nuclei of the BG even though they may be present. Although there are studies in animals investigating SN mechanisms of pain [96], future brain imaging studies in humans will contribute to our understanding of the role of these nuclei in pain processing. Developments such as accurate automated identification of sub-nuclei within the basal ganglia should be helpful in measures of functional brain changes [97].

The BG are involved in the integration of information between cortical and thalamic regions and in particular the three domains of pain processing - sensory, emotional/cognitive and endogenous/modulatory. Some investigators have suggested that these regions have evolved as a "centralized selection device to resolve conflicts over access to limited motor and cognitive resources" [98]. Phylogenetically older, sub-cortical connections exist between the BG and brainstem regions involved in sensorimotor processing [99], and more recent evidence points to BG involvement through direct connections from sensory inputs not involving cortical loops [100]. Dysfunctional cortico-BG-thalamic loops may contribute to the maintenance of chronic pain and the evolution of altered neural processing that may be a basis for co-morbid behaviors. Whether brain processing of pain is dependent on external neural drive (i.e., peripheral/feed-forward inputs) or internal neural drive (i.e., subcortical or cortical drive/feedback), the BG are central processors that may play a role in integrating these divergent inputs that may modify pain over time.

Cortical stimulation has been used either directly with electrodes placed on the motor cortex [101] or through the use of repetitive transcranial stimulation (rTMS) [102]. Some investigators have postulated that motor cortex neurostimulation produces an analgesic effect by modulation of the affective components of pain and not of the sensory components [103]. In support of this hypothesis, stimulation of the secondary somatosensory cortex (SII), but not motor cortex, results in increased pain thresholds and altered discriminative capacity to pain [104]. A combination of motor cortex stimulation (MCS) and postoperative fMRI showed an inhibiting effect on the primary sensorimotor cortex as well as on the contralateral primary motor and sensitive cortices [105] without changes in BG. Others have argued that MCS may act through activation of perigenual cingulate and orbitofrontal areas to alter the emotional appraisal of pain or cortical-brainstem (e.g., PAG) activation that enhances inhibition of pain [106] or alterations in endogenous opioid systems [107]. Given the cortico-striatal loops, stimulation of a number of cortical regions may thus be involved in pain reduction. Stimulation of dorsolateral prefrontal cortex, motor cortex or sensory cortices may have different effects based on the specificity of cortico-striatal loops. In addition, connectivity between the BG and modulatory regions including the PAG/brainstem or through the NAc [108‐110] may also contribute to the potential mechanism of cortical activation resulting in analgesia.

Both clinical [111] and preclinical [112] studies have suggested that the ACC results in alterations of pain processing. The interesting issue seems to be the dissociation of pain intensity from pain affect (caring about the pain) [113, 114]. The ACC in humans may be involved with linking reward-related information [115] and with alternative actions, since destruction of the ACC results in errors related to planning [116]. The ACC has projections to the caudate nucleus and the NAc [117], and ACC lesions may produce an attenuation of a negative pain affect [118] through a combination of neural networks that include these circuits.

Pain is clearly a complex process that affects multiple brain systems. It debatable if pain is a learned behavior, but the BG may have a role in learning due to its involvement in habit and stimulus-response learning [119]. Such learning may be derived from pain related regions involved in sensory (e.g., pain intensity coding regions such as SI), affective (e.g., cingulate or anterior insula) or cognitive functioning (medial and lateral prefrontal cortices). Similar to the notion of "chunking action repertoires" for motor action, it may be that pain related repertoires include motor related changes.

Aside from CRPS, other conditions including depression and Parkinson's disease have BG pathology that may be important in their clinical presentation of centralized pain [120]. In Parkinson's disease with central pain, electrophysiological measures of pain pathways are normal, but hyperalgesia to repetitive stimuli is present; this is attenuated by L-dopa [121], and altered central processing in these conditions results in generalized pain symptoms. Although the basis for these changes is unknown, it may be the result of altered chemicals in the BG, such as central dopamine, or related to aberrant networks that induce a kindling-like pain [122]. Changes in neural connectivity that produce changes in increased synaptic strength [123] and could include pain facilitatory circuits [124] may also underlie these changes. Although the basis for these changes is unknown, alterations may then lead to other changes such as abnormal gating of sensory-motor function involving parallel changes thalamic regions [125]. Imaging patients with depression in whom non-dermatomal sensory deficits were present shows that hypometabolic patterns (FDG-PET) of activation in the putamen was observed [126].

Regions of the thalamus including the paramedian and anterolateral are known to be involved with central pain (e.g., post stroke pain) [38, 127]. The striatum receives excitatory input from the thalamus. The centromedian (CM) and parafascicular (Pf) thalamic nuclei are important sources of thalamostriatal projections [128] and send connections to the putamen and caudate [128, 129]. Moreover, the topography of these connections corresponds with cortical sensorimotor territory observed following cortical injections [130].

The BG has high levels of endogenous opioids, and high binding of opioid receptors are present within the BG [131]. In a number of chronic pain conditions, receptor-binding studies indicate a decreased opioidergic tone. Many analgesics act at the level of the basal ganglia and may contribute to both analgesic and addictive processes. An understanding of how these functional processes are differentiated by different endogenous chemicals (e.g., opioidergic vs. dopaminergic (see [65])) may contribute to better analgesics since opioidergic tone might be abnormal [77].

Given that pain now has a functional basis in terms of regions activated including those that are classically involved in reward [45], and that chronic pain may be a reward deficit syndrome [132], modulating dopamine may have important possibilities for pain treatment. Parkinson's disease patients have improved pain control when treated with L-dopa [21, 121]. Changes in dopamine are a critical element in Parkinson's disease where abnormal pain processing is present in both conditions. Recent imaging data have reported a direct correlation between striatal dopamine D2/D3 receptors and sensory thresholds as being selective for the modality of pain but not for non-painful stimuli [133]. Striatal dopamine D2/D3 receptors may control a modulatory pathway producing a parallel shift in the stimulus-response function for sensory signals [134]. Others have suggested differential processing of pain within the BG - a more dorsal DA D2 receptor-mediated neurotransmission in the caudate and putamen that correlates with subjective ratings of sensory and affective qualities of the pain, along with a more ventral system involving the NAc, is associated with emotional processing [135]. Such differences may be important in drug effects on pain. The use of antipsychotic medications for pain is not new [136], and the number needed to treat (NNT; this is the number of patients to be treated for the first subject to show a 50% analgesic effect) of 2.6 is very competitive with the best drugs available for chronic pain [137]. However, these drugs have extrapyramidal and sedative effects, and atypical antipsychotics that have fewer side effects may have analgesic properties as assessed in a limited number of studies [138]. Newer drugs that target specific dopamine receptors may prove to be more useful.

Brain imaging affords the possibility to measure changes in brain circuits that may be altered as a result of deep brain stimulation. Some of these have shown specific changes in BG circuitry [139]. In the latter, stimulation of the ventroposterior medial thalamus (VPM) resulted in decreases in activation of the substantia nigra when activation prior to stimulation vs. post stimulation that provided pain relief was measured (pain, no stimulation vs. no pain, no stimulation). Thalamic stimulation may activate thalamo-cortical-BG loops that contribute to the analgesic response [140].