Our study applied er-fMRI to evaluate the brain correlates of ANS response to acupuncture stimulation at different acupoints. This approach allowed us to investigate the brain circuitries underlying different ANS outflows, for different acupoints, as well as different response patterns, such as HR– and HR+. Acupuncture stimuli produced activation in S2, insula, and mid-cingulate cortex, and deactivation in default mode network (DMN) areas. Differences between acupoints in brain response were noted for anterior insula activation, which was greater following SP9 compared with ST36 stimulation, while ST36 produced more robust DMN deactivation than both SP9 and sham acupuncture at SH1. On average, HR deceleration (HR–) and SCR were noted following both real and sham acupuncture, though magnitude of response was greater following real acupuncture and intersubject magnitude of response correlated with evoked sensation intensity. Acupuncture events with strong SCR also produced greater anterior insula activation than without SCR. Moreover, acupuncture at SP9, which produced greater SCR, also produced stronger sharp pain sensation, and greater anterior insula activation. Conversely, acupuncture-induced HR– was associated with greater DMN deactivation. Between-event correlation demonstrated that this association was strongest for ST36, which also produced more robust HR–. In fact, DMN deactivation was significantly more pronounced across acupuncture stimuli producing HR–, versus those events characterized by acceleration (HR+). Thus, differences in brain response to different acupoints may relate to differential autonomic outflows and may result from different sensations elicited by stimuli at these different bodily locations.
Previous studies of ANS response to acupuncture have found that acupuncture stimulation can induce both increased [Haker et al., 2000
; Knardahl et al., 1998
; Yao, 1993
], and decreased [Abad-Alegria et al., 2001
; Backer et al., 2002
; Imai and Kitakoji, 2003
; Nishijo et al., 1997
; Sugiyama et al., 1995
; Wang et al., 2002
] heart rate, but have rarely [Backer et al., 2002
] explicitly placed their results into the context of known psychophysiological reflexes, such as the orienting response (OR) and startle/defense response (SR/DR), which have been differentiated by decelerative versus accelerative HR response, respectively [Graham and Clifton, 1966
; Ohman et al., 2007
; Turpin 1986
]. In our study, events with HR– response patterns may represent OR, while HR+ patterns may represent SR/DR. For SR/DR, HR increase onset is typically greater than 2 s (consistent with our data), and peak response can be as long as 30 s [Turpin, 1986
], much longer than OR. The OR is a physiological response to any novel, moderate intensity stimulus and is thought to arise from a mismatch between the stimulus and the subject's expectation for that stimulus [Sokolov and Cacioppo, 1997
]. It is characterized by stimulus-associated SCR, pupillary dilation, and HR deceleration [Cook and Turpin, 1997
; Graham and Clifton, 1966
; Stelmack and Siddle, 1982
]. The OR is accompanied by reduced somatic activity and hypothesized to be associated with enhanced sensitivity to sensory input—in effect priming the subject for future sensory evaluation [Sokolov, 1963
]. In our study, acupuncture at both acupoints (ST36 and SP9) and noninsertive sham acupuncture at a nonacupoint (SH1) all produced HR deceleration—a hallmark of the OR. In fact, stimuli of any sensory modality (auditory, visual, somatosensory, etc.) can induce an OR. Also, some of the brain correlates of phasic ANS response to acupuncture are similar to those described by studies evaluating pain and other sensory stimuli [Piche et al., 2010
]. However, as acupuncture directs orienting towards an internal target (the needle stimulus is localized inside the body), the acupuncture-induced OR may have greater salience and may thus differ in behavioral/therapeutic consequence and duration of physiological response from auditory or visual-induced OR—a hypothesis that should be tested in future studies.
In our study, a HR+ HR– difference map as well as an intrasubject correlation analysis found a close linkage between HR deceleration and DMN deactivation. While research evaluating the brain correlates of ANS outflows consistent with OR and other psychophysiological reflexes has been limited, er-fMRI studies suggest that visual OR (evidenced by SCR, not HR deceleration as in our study) was related to activation in the hippocampus, anterior cingulate and ventromedial prefrontal cortices [Williams et al., 2000
]—the latter a key node of the DMN. Also, auditory OR, again evidenced by SCR, was related to activation in the amygdala and ventromedial and inferior prefrontal cortices [Williams et al., 2007
]. Some of the differences between our results and those of Williams et al. may be due to differences in the ANS metric used to characterize OR (i.e., SCR versus HR), and/or stimulus modality. In fact, Williams et al. analysis approach was more similar to our analysis contrasting brain response for high versus low SCR events, which for us also demonstrated SCR-associated activity in ACC, but, in contrast to Williams et al., also noted activity in anterior insula, and pSMA. In fact, insula and cingulate activity are commonly related to HR acceleration and SCR response to different stimuli [Critchley et al., 2003
; Gray et al., 2009
], supporting our SCR results. In our study, the discordance in brain activity associated with high SCR versus high HR– (i.e., robust DMN deactivation) is also supported by the lack of cross-correlation between SCR and HR response to acupuncture events (see below). In fact, taken together, these results demonstrate that using only SCR to define an OR event is potentially problematic, as SR/DR will also produce SCR, while only HR can reliably differentiate these two patterns [Turpin, 1986
]. Additionally, marginal differences between our results and those of previous studies may also be due to differences in stimulus modality, supporting our contention that acupuncture-induced OR involves different affective, cognitive, and sensorimotor integration brain centers compared with visual and auditory-induced OR.
In contrast to the predominant HR– response, some acupuncture stimulus events produced more robust HR acceleration (HR+) and thus are more characteristic of SR/DR [Turpin, 1986
]. While it is not known why some acupuncture stimuli produced HR+, while others produced HR–, we suspect that, while context was generally consistent across stimuli, variable affective states stemming from variable sensation quality and/or intensity may have played a role. In fact, we found that SP9 stimulation produced more sharp pain intensity, less HR deceleration, and greater SCR compared with ST36. Greater sharp pain in response to acupuncture was correlated with greater SCR, while a trend was found for greater deqi sensation (MASS Index) correlating with greater HR deceleration. This result is broadly consistent with previous pain studies, which have demonstrated cross-correlation between greater pain perception and greater SCR and insula activation [Piche et al., 2010
]. While we could not correlate sensation with SCR or fMRI response on an event-by-event basis, future studies should collect sensation data following individual stimulus events for more detailed correlation analyses. In addition, SR/DR is likely an aversive response (e.g., “fight-or-flight”), to be avoided in a clinical setting, as it may preclude the establishment of beneficial tonic ANS responses, such as the “relaxation response,” which has been associated with characteristic modulation in ANS outflow to various end-organs [Peng et al., 2004
]. However, direct correlation of acupuncture-induced phasic ANS response with tonic ANS outflow and, ultimately, clinical outcomes, remains to be investigated.
Provocative differences were noted between real and sham acupuncture, and between real acupuncture at two different acupoints, SP9 and ST36. While the phasic ANS response to both real and noninsertive sham acupuncture resulted in predominant HR decrease and SCR increase, differences in degree existed between different acupoints—e.g., ST36 stimulation produced greater HR decrease, while SP9 stimulation produced greater SCR. Previous studies have also demonstrated phasic SCR [Kang et al., 2011
], and longer-latency, post-stimulus (tonic) skin conductance decrease [Hsu et al., 2006
]. Our results suggest that the magnitude of deqi
sensation and the phasic ANS response, which can differ for different acupoint locations (), may relate to the efficiency of different acupoints in modulating different neural circuitries. In fact, DMN deactivation significantly correlated with HR deceleration to a greater degree in real (SP9, ST36) versus sham acupoint locations. However, this relationship may have been driven by the greater sensation evoked by verum versus sham acupoint stimulation. DMN regions such as vmPFC have been linked to acupuncture modulation of HR in recent studies [Beissner et al., 2012
], which used a different, non-erfMRI analysis approach. Additionally, increased correlation between activation in iPS and HR deceleration was noted for SP9, compared with ST36. This region has been implicated in higher order, visuo-spatial and visuo-motor integration [Uno et al., 2000
]. Ultimately, we propose a more conservative form of “acupoint specificity”—i.e., one of modulatory degree rather than kind. Thus different locations for stimulation may be more efficient to produce physiological effect, and may also relate to the choice of different acupoints to treat different conditions in the clinic. For example, both insertive and noninsertive acupuncture needle stimulation at different locations has been shown to produce analgesia for chronic pain, with real acupuncture at specific points producing slightly better analgesia (a consistent, if not always a statistically significant finding across clinical trials) [Cummings, 2009
; Hopton and MacPherson, 2010
]. While many clinical trials of acupuncture have used nonpenetrating needling as a placebo control, placebos in efficacy studies should be physiologically inert. Our results suggest that fMRI and ANS response to sham acupuncture stimulation (even at nonacupoints) is robust and, while lower in amplitude, is grossly similar to real acupuncture; hence not physiologically inert.
We found very little cross-correlation between the different ANS outflow measures—HR and SCR. Organ-specific outflow in the parasympathetic system is generally accepted [Porges, 2007
], and similar organ-specific organization in the sympathetic system, while not as readily acknowledged, has recently been demonstrated as well [Janig, 2006
; Morrison, 2001
]. Thus, lack of cross-correlation suggests that acupuncture can independently modulate ANS outflow to different end-organs. This lack of cross-correlation was further supported by different event-related ANS/fMRI relationships for HR and SCR. Manual acupuncture, a relatively gross stimulus that cannot readily control evoked sensation, was very effective for producing significant dynamic range in ANS outflow, particularly HR deceleration. In fact, intrasubject (i.e., between-event) variability dominated for HR deceleration (ICC = 0.12), while intersubject variability dominated for SCR (ICC = 0.68). We believe that this difference in ICC was related to greater consistency in results for difference map versus correlational approaches noted for the HR–/fMRI analyses, compared with SCR/fMRI analyses. Thus, our results suggest that future ANS/fMRI studies using similar correlational methods may choose to focus on HR response, if only one metric can be incorporated into the study design.
Our study represents the first application of er-fMRI experimental design to acupuncture neuroimaging. Previous fMRI studies of brain response to acupuncture stimulation have mostly used blocked designs, and found activation in sensorimotor (e.g., SI, S2, SMA
) and stimulus salience (e.g., insula) related areas, and deactivation in default-mode network (DMN) areas and the amygdala [Dhond et al., 2007
; Hui et al., 2005
; Napadow et al., 2005
]. Thus, similarity exists for brain response to acupuncture using block and er-fMRI designs, with the latter having the added benefit of concurrent ANS response estimation, differentiation of multiple stimulus locations within the same run, and (potentially) greater ecological validity. While a direct comparison should be performed, our anecdotal observations suggest that er-fMRI designs for acupuncture stimuli are more robust than block designs, particularly for long duration stimulus blocks, which are likely to involve varying sensation profiles over the needle stimulation period [Napadow et al., 2009
Several limitations should be noted. First, while several different analysis models (i.e., correlation, difference maps) were used in our study, it is important to note that for all models, we can only determine a correlative, not causal, relationship, and not every variable can be controlled. For instance, stimuli that produce more somatosensation and somatosensory cortical response could also be more startling and hence more apt to produce HR+ rather than HR– response (as we found). Thus, we do not claim that somatosensory regions are, per se, control regions of HR response to acupuncture (unlikely), only that somatosensory activation is also related to ANS outflow. In addition, our imaging protocol was optimized for whole-brain data acquisition. While we report very few results in brainstem nuclei, many premotor ANS nuclei lie in the brainstem (e.g., medulla). Brainstem focused neuroimaging approaches, which for example improve co-registration for group analyses [Napadow et al., 2006
], have been used to investigate acupuncture stimuli [Napadow et al., 2009
], and future studies should adopt er-fMRI designs to similar brainstem-focused neuroimaging methods. In addition, our approach was focused on more short-term phasic ANS and brain response to acupuncture and we needed to balance shorter ISIs to maximize events with longer ISIs to allow for adequate evolution of ANS response. While for mean ANS response, not all responses returned completely to baseline for the shortest ISI tested (8 s), and this is a limitation to the study, peak response was within our analysis window. Moreover, to probe if incomplete recovery affected subsequent SCR response, we correlated SCR response for event i
, with the change in baseline used for event i
-1. The correlation was r
= –0.003—i.e., a potentially shifting baseline due to incomplete recovery did not affect subsequent SCR response amplitude. Finally, acupuncture has also been noted to produce longer-term sustained, post-stimulus effects [Dhond et al., 2008
], and future studies should attempt to resolve the interaction between short- and long-term physiological response patterns.
In conclusion, our approach allowed us to link ANS and fMRI response in order to infer potential control brain circuitry underlying different autonomic responses to acupuncture stimuli at different acupoints. Our results suggest that different sub-regions of the brain circuitry responding to acupuncture appear to be associated with different ANS outflow responses to needle stimuli and may result from different sensations elicited by stimuli at different bodily locations. This study applied er-fMRI to link brain and ANS response to acupuncture, and future studies should explore how this central autonomic network response to acupuncture influences clinical outcomes.