In this study, we investigated functional connectivity of the PAG during genuine and sham EA. We found that during both genuine and sham EA, the PAG was significantly connected with brain areas surrounding the PAG, including the midbrain tegmentum, substantia nigra, raphe nucleus, hypothalamus, striatum, globus pallidum, left insula, thalamus, hippocampus, brain stem, and cerebellum as well as distant regions such as ACC, MPFC, MCC, PCC and precuneus. However, genuine EA - relative to sham EA - showed significantly stronger connectivity between the PAG and left PCC, and significantly weaker connectivity between the PAG and rAI.
Our result that, during both genuine and sham EA, the PAG is functionally connected with surrounding regions and some distant regions, including the RVM and ACC, is similar to our previous study investigating the intrinsic functional connectivity of PAG during resting state using data from a different cohort . Thus, neither genuine nor sham EA appeared to significantly disrupt the connectivity between PAG and the network of brain regions seen during resting state.
During genuine EA, we found that the PAG had a significantly stronger connectivity with the PCC. Through an examination of the existing literature, we could not find reliable evidence for a direct (monosynaptic) connection between the two structures; however, previous studies suggest a functional linkage between the PCC and brain stem . A di-synaptic connection is the most parsimonious solution for which direct evidence exists. Thus, we speculate that this functional linkage may be conveyed via the thalamus and ACC, two regions that have direct connections with both PCC and PAG.
Previous studies have implicated the PCC's involvement in responses to treatment in chronic pain patients [79, 80]. In a study by Niddam and colleagues , patients with myofascial pain syndrome were given painful stimulations during fMRI, and in between scanning sessions the same area was treated with low-intensity electrostimulation. When comparing responders and non-responders, a treatment effect was observed for responders in the dorsal midbrain, PCC, and the caudate .
In an early PET-study from 1995 , a normalization of attenuated PCC activity after non-opioidergic treatment was seen in patients with chronic neuropathic pain. Patients with localized peripheral neuropathic pain were treated with a regional nerve block using lidocaine, resulting in significant analgesic effects. Interestingly, the neural correlate to the pain alleviated state was an increase of cerebral blood flow in the ACC and PCC. Hseieh et al. suggest that the increased neural response in the PCC could reflect the altered subjective perception of pain relief rather than the afferent blockade. The PAG is one of the key regions in the descending pain inhibitory circuitry, enabling regulation of afferent pain signals. The strong connectivity between the PAG and the PCC in response to active treatment furthers the idea that the PCC plays an important role in pain treatment.
The PCC is a key region in the default mode network (DMN): a set of specific brain structures with intrinsic fluctuations that constitute a baseline of attention and wakefulness in the human brain [81–83]. In humans, the PCC has the highest level of resting cortical glucose metabolism , and is involved in processing intentions related to the self, self-awareness and conscious experience, which are key functions attributed to the DMN [78, 85]. DMN activity has been shown to decrease in relation to task-evoked activity, and demonstrates an inverse relationship with the cognitive work load of the task [81–83]. Other studies show that activity in the DMN decreases in response to repeated painful stimuli , and chronic pain patients seem to exhibit permanently altered DMN activity. In a study by Baliki and colleagues , chronic low back pain patients displayed reduced activity in several key DMN regions compared with healthy subjects. We believe the results from this study indicate that EA may modulate the functional connectivity of the DMN, as evidenced by the increased connectivity of PCC with PAG during EA stimulation.
In this study we also found that during sham acupuncture, the PAG has stronger connectivity with the rAI compared with genuine acupuncture. The anterior insula is a key region in the pain matrix  and is involved in integration and interoception of pain [88–90] and pain modulation processes such as placebo analgesia [29, 60]. In a more recent study, investigators found that the pre-stimulus functional connectivity between the PAG and the anterior insula can predict subsequent pain perception . Thus, we speculate that EA stimulation may reduce brain responses to calibrated pain stimuli by interfering with the functional connectivity between the PAG and insula. Further research is needed to test this hypothesis.
An interaction between the acupuncture treatment modes and expectancy levels was observed in right superior parietal lobule. Studies have suggested that this region is involved in attention  and somatosensory perception modulation . We speculate that our results may indicate that functional connectivity in genuine and sham EA are differentially modulated by expectancy levels, however further research is needed to fully understand the sources of observed functional connectivity in this study.
In this study, we did not find any significant functional connectivity changes between the high and low expectancy groups. We believed that this may be attributed to several reasons. 1) The PAG seed we chose for this study was identified in our previous analysis of the fMRI signal changes evoked by calibrated heat pain  as being selectively involved in mediating acupuncture treatment effects (genuine EA vs sham EA) and not expectancy effects (high expectancy vs low expectancy). 2) Although previous studies suggested that PAG is involved in expectancy evoked placebo analgesia [94–96] or attention modulation of pain , the involvement of PAG in these studies is observed during the pain application process; in contrast, this study measures functional connectivity changes during acupuncture treatment. Thus, our results are not necessarily in conflict with findings from previous studies. 3) The relatively small sample size may also prevent us from finding significant functional connectivity between the high and low expectancy conditions. Further study is needed to elucidate the influence of expectation of analgesia on the functional connectivity of PAG during the treatment phase.