The decreased relative activation in NCF of the migraine group suggests a dysfunction in this brainstem descending pain modulatory system. NCF sends dense neural projections to the rostroventral medulla 
, which directly modulates dorsal horn nociceptive transmission neurons in the spinal cord and in the analogous medullary dorsal horn 
. NCF receives reciprocal input from lamina I dorsal horn neurons 
, which may drive its activation and thus complete a possible negative-feedback loop 
. Inputs to the NCF from higher brain structures related to modulatory processing may also contribute to NCF output 
. In the healthy group, the NCF activation we observed with both 41°C and noxious heat could be conceivably driven by thermal and nociceptive inputs from lamina I in the trigeminal nucleus (face)/dorsal horn (hand) 
, as well as wide dynamic range neurons in deeper lamina such as IV and V 
In addition to the aforementioned structural connectivity data, functional investigations of NCF are suggestive of a role in descending pain modulation. Animal and human studies have suggested that the NCF can inhibit and facilitate nociception through cholinergic and glutamatergic mechanisms, which may be triggered by noxious stimulation, central sensitization, as well as the expectation of pain. In animal studies, electrical stimulation of NCF produces opioid-mediated analgesia through excitatory cholinergic projections to the nucleus raphe magnus (NRM) 
. Additionally, electrical stimulation or microinjection of morphine into NCF in the rat can also activate glutamatergic projections to the NRM, leading to modulation of pain-related behaviors 
. In addition to the NCF-NRM pathway, recent evidence suggests that morphine microinjection into NCF may activate a compensatory descending modulatory pathway when the NRM is lesioned 
. Human imaging studies have reported that activations in NCF and the rostroventral medulla are correlated during repeated noxious stimulation 
, and also that NCF may activate during punctate mechanical hyperalgesia, suggesting that it is involved in a alterations in descending pain modulation 
. NCF is also activated during the expectation of pain, indicating that awareness of impending pain can trigger a preparatory modulatory process in NCF 
We interpret our results as follows: NCF hypofunction is a characteristic of migraine sufferers that is detectable during their interictal phase, and reflects a dysfunction of descending modulation. The dysfunction we observed may be a result of damage secondary to a history of repeated migraine attacks 
. However, the mechanism of such damage is not known although a neurovascular etiology has been suggested. Though the hypofunction we observed intuitively suggests a decrease in inhibition, the NCF is an integrative structure that contains both ‘off’ and ‘on’ cells 
, and a decrease in the overall level of activation would not rule out the presence of enhanced facilitation. The hyperexcitability of nociceptive circuitry downstream of the NCF may contribute to central sensitization at the onset of migraine. NCF dysfunction could thus lead to progressive changes in the spinal trigeminal nucleus (localized allodynia), and/or the thalamus and spinal cord (generalized allodynia). However, note that while NCF hypofunction was found in migraineurs during their interictal phase, the NCF dysfunction may change or even be supplanted by abnormalities in other brain regions during the different phases of an actual migraine attack.
The absence of a difference between migraine and control subject NCF activation for noxious stimuli to the face in migraine patients () is unexpected, in light of the impaired descending modulation model. We speculate that the NCF dysfunction manifests only when insufficient afferent drive is present to fully trigger the descending modulatory system. In other words, increased afferent drive may be required to overcome the NCF dysfunction to activate descending inhibition. In migraine patients, noxious heat applied to the face may more effectively activate descending modulatory circuitry relative to the control group, whereas non-noxious heat did not seem to activate descending inhibition to the same extent. That this speculative intensity-dependent relationship appears to be specific to the face could be attributed to the involvement of potentially sensitized trigeminal afferents in migraine patients. At higher intensity levels than used in this study, increased afferent drive may also recruit activation of other modulatory centers (e.g., PAG). The existence of such an intensity-dependent activation of the descending modulatory system is mere conjecture at this point.
Why would a non-noxious stimulus (41°C) activate a brainstem modulatory region like the NCF? Several possible explanations may account for this: (1) 41°C is known to activate nociceptors, and what we observe may reflect an enhancement of their sensitivity; (2) since NCF sends descending projections to second-order wide dynamic range neurons 
, which encode both non-noxious and noxious stimulus intensities, perhaps NCF also affects innocuous stimulus processing at sub-perceptual levels in the migraine interictal phase; (3) low-threshold C-fiber afferents may convey other information that differentially activate or involve descending modulatory systems 
, including the NCF; (4) corticobulbar input in migraine patients may differentially modulate NCF activity 
regardless of stimulus intensity via a sensory-cognitive mechanism; (5) the interictal migraine brain may be hyper-excitable in regards to general sensory processing 
, which may or may not include the involvement of NCF. These explanations expand the possible role of descending modulatory processing in migraine.
How is it possible to see differences in modulatory circuitry without seeing a difference in pain sensitivity between migraine and healthy subjects using comparable stimuli? While we do not know the specific answer to this question, there are a number of possible explanations. NCF can be triggered during the anticipation of pain, before a stimulus is even applied 
: this suggests that the perception of physical stimuli is not necessary for NCF activation, and that cognitive processes may influence NCF activation in healthy subjects perhaps more effectively than in migraine patients. In addition, structural changes detected in this region of the brainstem 
, which includes white matter changes, may alter the coupling of the blood-oxygen-level- dependent response and neural activity 
. The latter interpretation leaves open the possibility that NCF in interictal migraine patients may function perfectly well, but that the fMRI signal itself is dissociated with neural activity in this area. Finally, patient medications may have dampened the fMRI signal, given that healthy controls were free of medications. However, this possibility is perhaps less likely given that: (1) eight out of twelve patients were not taking pre-emptive medications for their migraine, (2) patients discontinued their medications for one dosage cycle prior to imaging, (3) the significant changes observed were specifically localized to NCF, and were not global as might be expected for a drug, and (4) the heterogeneity of the medications taken by the patients reduces the likelihood of a mass action of any one pharmacological mechanism influencing the fMRI signal. An important caveat is that intermittent use of abortive migraine medications may have unknown long term effects on nociceptive processing.
We expected to observe changes in the PAG given its involvement in descending modulation during migraine attacks 
. The relationship between the NCF and PAG in descending modulation is incompletely understood, but they share cytoarchitectural homology 
and are extensively connected 
. However, only in one condition (Hand Thr+1°C) did we observe hypofunction in the dorsolateral PAG (data not shown). We interpret this negative result as an indication that PAG may only be recruited in descending modulation with higher levels of pain, such as that experienced during migraine.
We propose that NCF hypofunction in migraine patients contributes to central sensitization during attacks through partial loss of inhibition and/or enhanced facilitation of ascending nociceptive pathways. A prevailing theory to explain the occurrence of migraine attacks is that hyperexcitability develops along the trigeminovascular pathway 
. Disruption of a descending modulatory system in migraine patients could cause such hyperexcitability, and has been previously hypothesized to be an underlying cause for migraine pathology 
. While this state of putative disinhibition/facilitation did not appear to impact the perception of thermal stimuli during the interictal phase, we hypothesize that it may facilitate central sensitization at the onset of migraine. In our migraine patients, we found that NCF activation was disrupted not only for stimuli applied to the face, but also the hand. This relationship suggests that for migraineurs with extended allodynia, the disinhibition/facilitation of the encoding of noxious heat that accompanies NCF dysfunction is not specific to the head, but may be generalized to other parts of the body.
Currently, drugs that are developed to treat migraine have focused on alleviating symptoms during a migraine attack. However, migraine patients have shown abnormalities in cortical sensory processing even between migraine attacks. Using innocuous and noxious thermal stimuli as test stimuli, we have identified specific brain structures that are hypofunctional in migraine patients when they are not having an attack. This model may be a useful surrogate in evaluating pre-emptive or disease modifying therapies for migraine patients. Considering that recurrent episodic migraines appear to transform into severe daily headache 
, this brainstem dysfunction could be used to evaluate treatments to prevent this transformation 
. Future studies should evaluate the potential of current and future migraine preventative agents to correct this dysfunction, thereby improving descending inhibition and reducing headache frequency.