Title: Regional brain responses associated with thermogenic and psychogenic

34 Sweating events occur in response to mental stress (psychogenic) or with 35 increased body temperature (thermogenic). We previously found that both were 36 linked to activation of common brainstem regions, suggesting that they share the 37 same output pathways: a putative common premotor nucleus was identified in 38 the rostral-lateral medulla. We therefore looked in higher brain regions for the 39 neural basis that differentiates the two types of sweating event. Previous work 40 has identified hemispheric activations linked to psychogenic sweating, but no 41 corresponding data have been reported for thermogenic sweating. Galvanic skin 42 responses were used to measure sweating events in two groups of subjects 43 during either psychogenic sweating (n=11, 35.3±11.8 years) or thermogenic 44 sweating (n=11, 34.4±10.2 years) while regional brain activation was measured 45 by BOLD signals in a 3Tesla MRI scanner. Common regions activated with 46 sweating events in both groups included the anterior and posterior cingulate 47 cortex, insula, premotor cortex, thalamus, lentiform nuclei and cerebellum 48 (Pcorrected<0.05). Psychogenic sweating events were associated with significantly 49 greater activation in the dorsal mid-cingulate cortex, parietal cortex, premotor 50 cortex, occipital cortex and cerebellum. No hemispheric region was found to 51 show statistically significantly greater activation with thermogenic than with 52 psychogenic sweating events. However, a discrete cluster of activation in the 53 anterior hypothalamus/preoptic area was seen only with thermogenic sweating 54 events. These findings suggest that the expected association between sweating 55 events and brain regions implicated in ‘arousal’ may apply selectively to 56 psychogenic sweating; the neural basis for thermogenic sweating events may be 57 subcortical. 58 59 60 Introduction 61 Sweating in humans at rest may occur in response to increased body 62 temperature (thermogenic) or during mental or emotional stress (psychogenic). 63 Although early views considered that unique neural pathways (Chalmers and 64 Keele 1952; Iwase et al. 1997; List and Peet 1938), different neurotransmitters 65 (Nakazato et al. 2004; Noppen et al. 1997; Robertshaw 1977) and even sweat 66 glands from separate skin regions (Darrow 1937; Kuno 1956; Ogawa 1975) were 67 responsible for thermogenic and psychogenic sweating, recent studies indicate 68 that these sweating events are all cholinergically driven phenomena (Machado69 Moreira et al. 2012) that are expressed across the entire body surface (Machado70 Moreira and Taylor 2012a; b). Bursts of the sudomotor nerve activity that drives 71 sweating events are also highly correlated in time and amplitude between nerves 72 innervating different skin regions (Bini et al. 1980a). In line with this evidence, 73 we recently found that common brainstem pathways were activated in 74 association with both thermogenic and psychogenic sweating events in humans 75 (Farrell et al. 2013). The brain mechanisms generating the two categories of 76 sweating event cannot be identical, however. To underline this point, volleys of 77 cutaneous vasoconstrictor activity are co-activated with psychogenic sweating 78 events, but not with thermogenic sweating events (Bini et al. 1980a; b; Delius et 79 al. 1972; Macefield and Wallin 1996). So, if their brainstem pathways are 80 indistinguishable, it makes sense to look for differences in higher brain regions. 81 82 Hemispheric brain regions driving psychogenic sweat events have been studied 83 previously, and appear to be relatively stable, irrespective of the psychogenic 84 stress that gives rise to sudomotor activity. For instance, studies of brain activity 85 associated with skin conductance responses during different mental tasks 86 (gambling and working memory), have demonstrated common neural activation 87 across those tasks (Patterson et al. 2002). Brain regions showing sweating88 related activation are widely distributed and include the anterior and posterior 89 cingulate cortices, ventromedial prefrontal cortex, premotor and motor areas, 90 visual cortex, thalamus and cerebellum (Craig et al. 2000; Fredrikson et al. 1998; 91 James et al. 2013; Nagai et al. 2004; Patterson et al. 2002; Williams et al. 2000). 92 Additionally, event-related EEG dipoles have implicated the inferior frontal 93 gyrus, amygdala and hippocampus in the generation of sweating events during 94 mental arithmetic (Homma et al. 1998). However, the hemispheric regions 95 associated with thermogenic sweating are essentially unknown. 96 97 The mental and emotional stimuli that are typically used to elicit psychogenic 98 sweating also evoke a wide range of autonomic responses that collectively are 99 part of the arousal response (Critchley et al. 2013). Cerebral regions associated 100 with arousal have been extensively studied. In this context, the dorsal mid101 cingulate cortex has been identified by Critchley and others as the common 102 region associated with arousal-related activation of a wide range of sympathetic 103 nerves (Critchley et al. 2013). By contrast, body heating is associated with 104 drowsiness (Gilbert et al. 2004; Qian et al. 2014), which could have implications 105 for regional brain activation during thermal sweating. However, the few studies 106 investigating responses in the hemispheres to whole-body heating have failed to 107 produce a consensus on the brain regions that are activated during periods of 108 either increased thermoafferent or thermoefferent flow (Egan et al. 2005; Fechir 109 et al. 2010; Nunneley et al. 2002). 110 111 This study was designed to compare regional activation during both psychogenic 112 and thermogenic sweating events, focusing on regions above the midbrain. As in 113 our previous study (Farrell et al. 2013), we sought brain regions that were 114 selectively activated in association with sweating events rather than with mean 115 ongoing sweating levels. Our aim was to look for similarities and differences in 116 the cerebral mechanisms driving human sudomotor function under these 117 different forms of external stress. 118 Materials and Methods 119 Participants 120 The study was undertaken according to procedures approved by the Melbourne 121 Health Human Research Ethics Committee (#2008.147). Participants provided 122 written, informed consent before enrolment. A total of 22 people contributed 123 data for the study, of which there were two groups of eleven aged 34.4 (±10.2; 124 Thermogenic Group) and 35.3 (±11.8; Psychogenic Group) years, and including 125 the same proportions of men and women (91% male). Some of the data collected 126 from these participants has been reported previously (Farrell et al. 2014; Farrell 127 et al. 2013). The data presented herein from the Psychogenic Group have not 128 previously been reported, and all outcomes reported from the current analyses 129 of the Thermogenic Group data are also novel. The Thermogenic Group data 130 were previously used in an analysis focused on the preoptic area, and 131 investigated the sweating-related activation and functional connectivity of that 132 region based on contrasts with a normothermic resting-state data set (Farrell et 133 al. 2014) 134 135 Recording of Sweating Events and Skin Temperature 136 Sweating was monitored by recording galvanic skin responses with Ag-AgCl 137 electrodes (TSD203 electrodes, Biopac Systems, CA, USA) fixed on the palmar 138 surfaces of the right index and middle fingers. Cabling (MECMRI-3 MRI Cable, 139 Biopac Systems, CA, USA) and filters (MRIRFIF interference filter set, Biopac 140 Systems, CA, USA) connected the electrodes to a constant voltage amplifier 141 (GSR100C Galvanic Skin Response Amplifier, Biopac Systems, CA, USA). The 142 signal was digitized at 1 khz (Power 1401, Cambridge Electronic Design, 143 England) and recorded to computer (Spike2 (ver.7), Cambridge Electronic 144 Design, England). Gradient artefacts related to magnetic resonance image 145 acquisition were excised from the GSR recording. The electrodermal signal was 146 recorded as an AC signal (0.5 Hz high pass filter) to identify discrete 147 electrodermal events independently of any shift in mean signal level (Kunimoto 148 et al. 1991). Output signals from the scanner control panel were used to trigger 149 recordings of electrodermal data so that it could be matched with synchronously 150 acquired functional brain images. 151 152 Type T (copper – constantan) thermocouples were attached to three different 153 sites on each participant’s trunk (left, right, midline) and recorded via an 154 electronic thermometer (TH-5, Physitemp Instruments, NJ, USA). Signals from 155 the thermocouples were filtered, digitised and recorded to computer using the 156 same procedures and equipment as those described above for the recording of 157 electrodermal events. Finally, those data were averaged to yield an unweighted, 158 mean torso skin temperature. 159 160 Sweating-Related Stimulation 161 Thermogenic Sweating 162 Heating was induced by means of a water-perfused garment (LCG)(Med-Eng 163 BCS4 Body Cooling System, Allen Vanguard, ON, Canada) as previously described 164 (Farrell et al. 2014; Farrell et al. 2013). Temperature-controlled water (40 to 165 50oC) was circulated through a network of small diameter tubes sewn into the 166 garment, which covered the torso, arms and legs, but excluded the head, hands 167 and feet. Layers of insulating fabric were added to minimize heat loss to the 168 environment. Passive heating was sustained until regular sweating events were 169 detected electrodermally, at which time functional brain image acquisition 170 commenced with heating maintained throughout. As previously described 171 (Farrell et al. 2013), the water temperature was then adjusted to sustain a low 172 mean rate of sweating events. 173 174 Psychogenic Sweating 175 Psychogenic sweating was achieved through the imposition of a mentally 176 challenging colour/word Stroop task as previously reported (Farrell et al. 2013). 177 Visual stimuli were projected onto a screen that was visible to participants lying 178 in the magnetic resonance imaging (MRI) scanner via a mirror mounted on the 179 head coil. A sequence of coloured words was presented in a random order. This 180 contained both congruently (e.g., the word “red” written with red text) and 181 incongruently coloured letters (e.g., the word “red” written with green text) 182 These were presented for a 2 min block followed by a 30-s rest interval, and the 183 cycle was repeated three times during functional brain scanning scans. The task 184 was to count the number of nominated events (e.g., count the words “red” 185 written with yellow letters) that occurred in each 2-min block. The tasks 186 (counting congruent or incongruent events) and stimuli (colour words and text 187 colours) used during each blocks were varied within and across scanning runs. 188 Participants used a button box to record their count by indicating their choice of 189 answer from options displayed at the conclusion of the 2-min block. These 190 responses were not recorded. However, the timing of the participants’ actions 191 were taken into account in the analysis of functional brain images to prevent 192 their potential confounding influence (elaborated below). 193 194 Image Acquisition 195 Images were acquired with a Siemens 3Tesla magnetic resonance imaging 196 scanner (Trio system) with a 32-channel head coil at the Murdoch Children’s 197 Research Institute (Melbourne, Australia). High resolution structural images of 198 participants’ brains were collected using a T1 weighted image sequence (192 x 199 0.9mm sagittal slices, 256x256 matrix, in-slice resolution 0.8mm x 0.8mm, 200 TR=1900ms, TE=2.59ms, flip angle=9o). Functional brain images sensitive to 201 blood oxygen level-dependent (BOLD) contrast were also acquired for all 202 participants (TR=1900ms, TE=35ms, flip angle=90o). The field of view of the 203 functional images encompassed the brain hemispheres, cerebellum and 204 brainstem (30 slices of 4mm thickness, in-slice resolution=3.6x3.6mm). Each 205 functional scan lasted for 7 min and 55 s and included 250 sequential brain 206 volumes (1.9 s per brain volume). Two of these functional scans were collected 207 from each participant. 208 209 Analysis 210 Preprocessing 211 Preparation and statistical analysis of functional brainstem images was 212 performed with the Oxford Centre for Functional Magnetic Resonance Imaging of 213 the Brain Software Library (FMRIB, Oxford, UK, FSL version 4.1 214 (http://www.fmrib.ox.ac.uk/fsl/)). Sequential functional brainstem images from 215 single scanning runs were realigned to the middle image of the time-series to 216 correct for any head movement during the scanning run using MCFLIRT 217 (Jenkinson et al. 2002). Images were spatially smoothed using a Gaussian kernel 218 of 5mm full width at half maximum. The time-series of each scanning run was 219 mean-based intensity normalised and high-pass filtered. Functional brain images 220 from each participant were co-registered with a standard brain to facilitate 221 amalgamation of statistical outcomes across participants. Transformations were 222 performed with FMRIB’s Linear Image Registration Tool (FLIRT) (Greve and 223 Fischl 2009; Jenkinson et al. 2002; Jenkinson and Smith 2001). 224 225 Statistical Analysis 226 The first level of functional brain imaging analysis was performed on individual 227 scanning runs using general linear modelling including local autocorrelation 228 correction as instituted in the FMRI Expert Analysis Tool (FEAT, FMRIB’s 229 Improved Linear Model (FILM) (Smith et al. 2004; Woolrich et al. 2009). The 230 regressor of interest for each scanning run analysis was derived from the 231 recording of electrodermal events obtained contemporaneously with the 232 functional brainstem images. The signals from the electrodermal recordings 233 were initially down-sampled to correspond with the acquisition time of 234 functional brainstem images (one observation every 1.9 seconds). Timing 235 adjustments were then made to the electrodermal regressor to take account of 236 delays between activation of brain regions involved in sweating control and the 237 occurrence of electrodermal events. Homma and colleagues (1998) recorded 238 EEG, median sympathetic nerve activity and sweating at the finger tip during 239 performance of mental arithmetic and using dipole source localisation, 240 concluded that activity in cortical regions occurred 5 to 5.5 seconds before 241 sweating events (Homma et al. 1998). Consequently, the electrodermal 242 regressors were translated backward in time by an interval corresponding to the 243 acquisition of three brain volumes (5.7sec). The mental task involved 244 instructions, blocks of visual stimuli and button presses to indicate participants’ 245 responses. Regressors representing the mental task components were included 246 in the modelling of BOLD signal changes in order to take account of these known 247 sources of variance. Specifically, a regressor was used to indicate the onsets, 248 durations and offsets of periods during which participants were viewing images 249 for the Stroop task, and another regressor was included in the analysis to 250 indicate the timing of response cues and resulting button presses. The 251 hemodynamic response measured with BOLD contrast occurs at a delay of 4 to 6 252 seconds after neural activation and so the electrodermal and task regressors 253 were convolved with a model of the hemodynamic response function (Gamma 254 function) and included temporal derivatives of the regressor to account for 255 variations in the timing of events (Henson and Friston 2006). 256 257 Additional regressors were included in the general linear modelling of BOLD 258 signal changes during scanning runs to take account of non-neural related 259 variability. Physiological processes contribute to BOLD signal noise including 260 respiratory effects on local magnetic field properties and changes in the blood 261 and cerebrospinal fluid associated with the cardiac cycle. This physiological 262 noise does not usually confound fMRI analysis because it varies independently of 263 many stimuli and tasks employed in functional brain imaging paradigms. 264 However, sweating events are likely to correlate with other physiological 265 processes, especially in the case of mental stress (Fechir et al. 2008). In order to 266 reduce any confounding effects of physiological noise, regressors from three 267 regions of interest were used in the model to account for variance associated 268 with the cardiac and respiratory cycles. The regions of interest were identified in 269 the white matter, ventricles and circulation for each scanning run, according to 270 procedures previously described (Farrell et al. 2012), and signals from those 271 regions were extracted and included as the physiological noise regressors. The 272 influence of head movement on BOLD signal intensity was also taken into 273 account by including the six motion parameters (three translations and three 274 rotations) into the modelling of signal changes. 275 276 A parameter estimate was calculated for the fit of each regressor to the observed 277 BOLD signal for each voxel in the space of functional brain images resulting in a 278 series of statistical parametric maps for each scanning run. Parameter estimates 279 for the fits of regressors of interest (electrodermal responses) were carried 280 forward to higher levels of analysis that firstly averaged responses across 281 scanning runs for individual participants using fixed effects, and then 282 subsequently calculated average responses among group participants 283 (Thermogenic Group, Psychogenic Group) and between groups (Thermogenic 284 greater than or lesser than Psychogenic) using mixed effects (FMRIB’s Local 285 Analysis of Mixed Effects (FLAME) (Beckmann et al. 2003). Regions of activation 286 were considered statistically significant when the constituent voxels had values 287 exceeding z=2.3, and a cluster-corrected threshold of P<0.05 to take account of 288 the spatial smoothness of the images and the effects of multiple comparisons on 289 inferences of significance (Worsley et al. 1992). 290 291 Brainstem hemodynamic responses associated with sweating were 292 characterised with region of interest (ROI) analyses. This was done to confirm 293 that the timing of events were compatible with the expected response profiles, to 294 compare outcomes with previous reports from the study cohort, and to 295 characterise the nature of any between-group differences. ROIs were selected 296 from the midbrain and medulla according to previously described procedures 297 (Farrell et al. 2013). Clusters of between-group activation were also defined as 298 ROI. BOLD signals extracted from ROI were compiled in one of two ways. The 299 first method was used to assess the temporal profile of sweating-related signal 300 change. BOLD signals were averaged across the voxels within ROI for each time 301 point in individual scanning runs after motion correction, high-pass filtering and 302 spatial smoothing. Sections of time corresponding with sweating events were 303 identified for scanning runs by calculating the mean and standard deviation of 304 electrodermal signals and choosing peaks with values greater than the sum of 305 the mean and one standard deviation. Time points 28.5 s before and after each 306 peak were extracted from ROI time-series and averaged. The average BOLD 307 signals during sweating events of individual scanning runs were expressed as a 308 percentage of the average of the first three time points (28.5 to 24.7 s before the 309 peak). Percentage signal changes of all scanning runs were averaged to produce 310 the grand mean of BOLD signal changes during sweating events. The second 311 method was used to visualise the relative size of sweating-related signal changes 312 between the groups. The Featquery tool was used to estimate percentage signal 313 changes across sweating events for each scanning run and the outcomes were 314 averaged across groups. 315 316 Results 317 Skin Temperature and Sweating Event Frequency 318 Passive heating for thermogenic sweating was associated with an increased skin 319 temperature (36.2±S.D. 1.3oC) compared with the non-heated state during 320 psychogenic sweating (33.2±1.2oC, t(20)=5.7, P<0.001). Sweating events 321 occurred with similar frequency during functional brain scans for the two 322 experimental procedures (Thermogenic = 7.7±1.7 per scan, Psychogenic = 323 7.5±S.D. 2.0 per scan, t(20)=0.2, n.s.). Sweating activity patterns were similar to 324 those reported previously (Farrell et al. 2013). 325 326 Sweating Event-Related Activation 327 Sweating events were associated with activation in widely distributed regions of 328 the brain for both thermogenic and psychogenic stimuli (Fig. 1, Tables 1&2). 329 Extensive cingulate activations were principally in the dorsal mid-cingulate 330 cortex with psychogenic sweating events, whereas thermogenic sweating was 331 notable for clusters of event-related activation in the posterior and pregenual 332 cingulate cortices. Mesial activation was also seen in the precuneus for both 333 stimulus types, while only thermogenic sweating was associated with activation 334 in the anterior hypothalamus/preoptic area (AH/PO). Sweating-related 335 activation was seen in the insula with both stimuli, being predominantly anterior 336 in both hemispheres with psychogenic sweating events, and mid (right) and 337 posterior (left) for thermogenic sweating events. Right prefrontal cortex 338 activation was noted with both types of sweating event (Middle Frontal Gyrus, 339 BA10). Other regions activated with both types of sweating event were the 340 bilateral thalami, lentiform nuclei and cerebellum. 341 342 Thermogenic and psychogenic sweating event-related activation was seen in the 343 midbrain, pons and medulla. In agreement with previous findings (Farrell et al. 344 2013), the location of midbrain activations was similar for the two types of 345 sweating and was mainly seen in the dorsal part of the region (Fig. 2), . Overlaps 346 of sweating activations for thermogenic and psychogenic stimuli were also 347 apparent in the rostral medulla, being located symmetrically lateral (Fig. 2). This 348 again agrees with previous findings (Farrell et al. 2013). 349 350 Differential Activation by Psychogenic and Thermogenic Sweating Events 351 Contrasts between the two types of sweating event-related activation revealed 352 sites of significantly greater activation with psychogenic than with thermogenic 353 sweating events, but not the reverse (Fig. 3, Table 3). Regions showing increased 354 psychogenic versus thermogenic sweating activation included the dorsal mid355 cingulate cortex, premotor regions, parietal associative cortex, occipital cortex 356 and cerebellum. Examinations of time series of BOLD signal changes and mean 357 levels of sweating activations suggested that these regions were exclusively 358 activated with psychogenic sweating events, with little or no activation related to 359 thermogenic sweating events (Fig. 3). 360 361 Several brain regions showed substantial clusters of activation that were 362 prominent only with thermogenic sweating, but statistical comparison fell short 363 of proving that this was greater than with psychogenic sweating events. Such 364 regions included the lentiform nuclei (P= 0.1), amygdalae (P= 0.1) and AH/PO 365 (P= 0.09). 366 367 368 369 370 Discussion 371 This study has provided the first direct description of the network of forebrain 372 regions associated with thermogenic sweating events. It also confirmed the 373 findings of previous studies on the cerebral activations associated with 374 psychogenic sweating events (Craig et al. 2000; Fredrikson et al. 1998; Nagai et 375 al. 2004; Patterson et al. 2002; Williams et al. 2000) and skin sympathetic nerve 376 activity (SSNA), which includes both sweating and vasomotor events (James et al. 377 2013). Besides confirming our previous finding of common brainstem regions 378 activated with both thermogenic and psychogenic sweating events (Farrell et al. 379 2013), we now find such common regions in the cerebral hemispheres. Those 380 regions include parts of the cingulate, insular and premotor cortices, thalamus, 381 lentiform nuclei and cerebellum. But the current study also revealed important 382 differences. Psychogenic sweating events were associated with significantly 383 greater activation than thermogenic sweating events in the supplementary 384 motor area, premotor cortex, parietal cortex, parts of the cerebellum and dorsal 385 mid-cingulate cortex. Thus, as would be predicted, the two types of sweating 386 event were linked to distinct patterns of brain activation. 387 388 Both these types of sweating event occur episodically and synchronously across 389 the entire body surface (Hagbarth et al. 1972; van Beaumont et al. 1966) 390 following sweat gland priming (Machado-Moreira and Taylor 2012a; b). The 391 phasic sweating events that informed the analysis used in this study are driven 392 by bursts of sympathetic sudomotor activity (Bini et al. 1980b; Hagbarth et al. 393 1972; Ogawa and Bullard 1972). The measure used in this study, episodic 394 increases in skin conductance, measures the timing and amplitude of sweating 395 events. Although measured here at the fingers, the timing and amplitude of these 396 conductance changes accurately and linearly reflect bursts of sudomotor nerve 397 activity, which are highly correlated across skin regions (Bini et al. 1980a). We 398 are therefore confident that the brain activation patterns we have identified here 399 apply not only to finger sweating but to sweating events across the body. The 400 sudomotor bursts driving the sweating events are likely to be driven by the 401 output of rostral-lateral medullary nuclei, whose event-related activation was 402 identified in this and an earlier study (Farrell et al. 2013), and whose location is 403 homologous with the sympathetic premotor nuclei for sweating identified in the 404 cat medulla (Shafton and McAllen 2013). 405 406 Resistance of the skin is principally due to the stratum corneum (Lawler et al. 407 1960), and skin conductance increases only when fluid-filled sweat ducts pierce 408 that resistive layer. This makes skin conductance measurements insensitive to 409 any change skin in blood flow. In line with this, atropine abolishes sweating 410 events but leaves skin vasomotor responses intact, while bretylium does the 411 reverse (Lader and Montagu 1962). Even though vasomotor nerve volleys may 412 be activated at the same time as sudomotor nerve volleys (Bini et al. 1980a), they 413 should have had no influence on the electrodermal measurements that we used 414 to identify activated brain regions. 415 416 As have others, we found widespread forebrain regions activated in association 417 with psychogenic sweating events (Craig et al. 2000; Fredrikson et al. 1998; 418 Nagai et al. 2004; Patterson et al. 2002; Williams et al. 2000). It is likely that 419 many of these represent other arousal-related cerebral events that co-activate 420 with sweating, (e.g., cognitive, premotor and other autonomic processes). 421 Prevailing views on the interaction between autonomic responses and cognition 422 posit a link between arousal state and the behavioural significance of stimuli 423 (Critchley et al. 2013). Research using similar mental stresses, (i.e., Stroop task), 424 and additional measures of autonomic responses would suggest that 425 performance errors trigger a change in bodily state, especially when there is a 426 conscious awareness of the error (Critchley et al. 2005; Hajcak et al. 2003; 427 Nieuwenhuis et al. 2001). This shift in autonomic responses is conceptualised as 428 a somatic marker that provides cognitive feedback to decrease the probability of 429 future errors. 430 431 The dorsal mid-cingulate cortex has been identified as the common cortical area 432 associated with autonomic activation, including sweating, in response to both 433 physical and mental effort (Critchley et al. 2000). In line with that view, the 434 present study confirmed the expectation that this region would be strongly 435 activated with psychogenic sweating events. Strikingly, however, this was a site 436 of strong functional contrast: it was not activated by thermogenic sweating. If the 437 dorsal mid-cingulate represents a key cortical site controlling several ‘stress438 arousal-related’ sympathetic outflows (Craig 2009; Critchley et al. 2000; 439 Critchley et al. 2003), the source of thermogenic sweating events needs to be 440 sought elsewhere. 441 442 Methodologically, finding such a contrast is reassuring, because it tells us clearly 443 that different brain mechanisms drive the two types of sweating event. 444 Therefore, the thermogenic sweating events were not simply mini-arousals that 445 happened to occur during whole body heating. On the other hand, we cannot 446 exclude the possibility that some such arousals did occur, even though the 447 heated condition was generally relaxing and subjects were left undisturbed. If so, 448 that might explain some of the forebrain sites that were activated with sweating 449 events from both protocols. 450 451 While no site was identified that showed significantly greater activation with 452 thermogenic than psychogenic sweating events, this could be a false-negative 453 conclusion due to noisy data. In this context, the AH/PO area merits mention. 454 This brain region is known from animal studies to play a key role in 455 thermoregulatory processes (Nakamura 2011) and, when locally heated, it can 456 drive sweating (Beaton et al. 1941). A discrete cluster there was activated with 457 thermogenic sweating events, but not psychogenic sweating events (Fig. 1). We 458 have reported elsewhere that this region showed enhanced functional 459 connectivity with other brain regions in humans during whole-body heating 460 compared with during the normothermic state (Farrell et al. 2014). Whether the 461 AH/PO area in humans is truly a source of thermogenic but not psychogenic 462 sweating events must await determination by more detailed study. 463 464 We still do not know exactly where either type of sweating event originates. 465 Presumably in each case there is a network of neurons whose synchronous 466 activity acts as the source of the burst. These in turn may be influenced by such 467 factors as brain temperature, or state of arousal. Other attempts to localise the 468 source of psychogenic sweating events have included the following. Event469 related fMRI studies have identified activation in prefrontal, cingulate, parietal, 470 motor, insular and occipital cortex, and hippocampus, thalamus and cerebellum 471 to psychogenic sweating events (Craig et al. 2000; Fredrikson et al. 1998; Nagai 472 et al. 2004; Patterson et al. 2002; Williams et al. 2000), and additionally 473 activation in the cingulate, superior frontal, precentral and occipital cortices to 474 the mean sweating level by skin resistance (Fan et al. 2012; Nagai et al. 2004; 475 Zhang et al. 2014). James and colleagues sought cerebral signals to variations in 476 skin sympathetic nerve activity (SSNA) (James et al. 2013). The SNAA signal 477 includes vasomotor as well as sudomotor nerve activity (Bini et al. 1980b; 478 Macefield and Wallin 1996), although burst activity associated with these two 479 autonomic responses can be synchronised (Bini et al. 1980a), in which case the 480 associated regional brain activation would be the same. In the case of 481 psychogenic sweating events a related EEG signal has been studied, and its 482 sources located in two subjects to the inferior frontal gyrus, amygdala and 483 hippocampus (Homma et al. 1998). No one has previously attempted to localise 484 the origins of thermogenic sweating events. 485 486 Limitations 487 As with all fMRI studies, the results obtained here are correlative only. This 488 means that the brain regions we identified as activated with sweating events 489 could be due to other processes that may occur consistently at the same time. 490 This may apply, for example, to cutaneous vasoconstrictor traffic that is co491 activated with sudomotor bursts during psychogenic stimuli. It should not apply 492 to thermogenic sweating events. This distinction may explain some of the 493 differences we observe between the brain regions activated with psychogenic 494 versus thermogenic sweating events. 495 496 Second, the study involved two groups of participants. A within-subject design 497 would have been ideal, but a lag between study times prevented us recruiting all 498 the same participants twice. 499 Our experiment was designed to give a clear distinction between thermogenic 500 and psychogenic stimuli, but we cannot eliminate the possibility that some 501 psychogenic sweating events also occurred during the thermal stimulus runs and 502 contributed to the fMRI findings. Against this, the heating protocol we used was 503 mild, and likely to have been relaxing. Moreover, vasoconstrictor sympathetic 504 nerve bursts were not found to occur in resting heated subjects (Bini et al. 505 1980b; Macefield and Wallin 1996). 506 507 Finally, to measure responses through the whole brain we traded spatial for 508 temporal resolution. The size of voxels used could have prevented us from 509 resolving small distinct clusters from a larger mass of activation, and could have 510 made it more difficult to detect local differences between psychogenic and 511 thermogenic sweatingrelated activations. 512 513 Perspectives 514 This study has been the first successful foray into the functional imaging of 515 thermogenic sweating activation in the hemispheres of the human brain. The 516 network of brain regions activated during thermogenic sweating events shares 517 some common regions that are activated also during psychogenic sweating 518 events, but regional differences are also apparent. Notably, neural activation in 519 the dorsal mid-cingulate cortex was linked to psychogenic more than 520 thermogenic sweating events. This difference is likely to reflect greater levels of 521 arousal during psychogenic sweating. Previous studies have linked this region 522 with sympathetic activation (including sweating) during stress and arousal. Our 523 findings support that view. But the sources of thermogenic sweating events are 524 clearly different, and may ultimately be found in subcortical structures such as 525 the AH/PO area. Taken together with previous findings, this study suggests that 526 distinct supratentorial mechanisms generate psychogenic and thermogenic 527 sweating events, but these converge on common brainstem pathways en route to 528 the sweat glands. 529 530 531 532 Acknowledgments 533 We acknowledge the technical expertise provided by Michael Kean of the 534 Children’s Magnetic Resonance Imaging Centre (Melbourne, Australia). This 535 study was supported by the National Health and Medical Research Council of the536 Commonwealth Government of Australia (Project Grant 509089) and the537 Victorian government through the Operational Infrastructure Scheme. RMcA538 received a Principal Research Fellowship (566667) from the National Health and539 Medical Research Council, and MF was supported in part by the Robert540 J. Kleberg, Jr. and Helen C. Kleberg Foundation and the G. Harold and Leila541 Y. Mathers Charitable Foundation.542543 References544545Beaton LE, McKinley WA, Berry CM, and Ranson SW. Localization of cerebral546 center activating heat-loss mechanisms in monkeys. Journal of neurophysiology 4:547 478-485, 1941.548Beckmann CF, Jenkinson M, and Smith SM. General multilevel linear modeling549 for group analysis in FMRI. NeuroImage 20: 1052-1063, 2003.550Bini G, Hagbarth KE, Hynninen P, and Wallin BG. Regional similarities and551 differences in thermoregulatory vasoand sudomotor tone. The Journal of552physiology 306: 553-565, 1980a.553Bini G, Hagbarth KE, Hynninen P, and Wallin BG. Thermoregulatory and554 rhythm-generating mechanisms governing the sudomotor and vasoconstrictor555 outflow in human cutaneous nerves. The Journal of physiology 306: 537-552,556 1980b.557Chalmers TM, and Keele CA. The nervous and chemical control of sweating. The558British journal of dermatology 64: 43-54, 1952.559Craig AD. How do you feel--now? The anterior insula and human awareness.560Nature reviews Neuroscience 10: 59-70, 2009.561Craig AD, Chen K, Bandy D, and Reiman EM. Thermosensory activation of562 insular cortex. Nature neuroscience 3: 184-190, 2000.563Critchley HD, Corfield DR, Chandler MP, Mathias CJ, and Dolan RJ. Cerebral564 correlates of autonomic cardiovascular arousal: a functional neuroimaging565 investigation in humans. The Journal of physiology 523 Pt 1: 259-270, 2000.566Critchley HD, Eccles J, and Garfinkel SN. Interaction between cognition,567 emotion, and the autonomic nervous system. Handbook of clinical neurology 117:568 59-77, 2013.569Critchley HD, Mathias CJ, Josephs O, O'Doherty J, Zanini S, Dewar BK,570Cipolotti L, Shallice T, and Dolan RJ. Human cingulate cortex and autonomic571 control: converging neuroimaging and clinical evidence. Brain : a journal of572neurology 126: 2139-2152, 2003.573Critchley HD, Tang J, Glaser D, Butterworth B, and Dolan RJ. Anterior574 cingulate activity during error and autonomic response. NeuroImage 27: 885-575 895, 2005.576Darrow CW. Neural mechanisms controlling the palmar galvanic skin reflex and577 palmar sweating. Arch Neurol Psychiat 37: 641-663, 1937.578Delius W, Hagbarth KE, Hongell A, and Wallin BG. Manoeuvres affecting579 sympathetic outflow in human skin nerves. Acta physiologica Scandinavica 84:580 177-186, 1972.581Egan GF, Johnson J, Farrell M, McAllen R, Zamarripa F, McKinley MJ,582Lancaster J, Denton D, and Fox PT. Cortical, thalamic, and hypothalamic583 responses to cooling and warming the skin in awake humans: a positron-584 emission tomography study. Proceedings of the National Academy of Sciences of585the United States of America 102: 5262-5267, 2005.586Fan J, Xu P, Van Dam NT, Eilam-Stock T, Gu X, Luo YJ, and Hof PR.587 Spontaneous brain activity relates to autonomic arousal. The Journal of588neuroscience : the official journal of the Society for Neuroscience 32: 11176-589 11186, 2012.590 Farrell MJ, Cole LJ, Chiapoco D, Egan GF, and Mazzone SB. Neural correlates591 coding stimulus level and perception of capsaicin-evoked urge-to-cough in592 humans. NeuroImage 61: 1324-1335, 2012.593Farrell MJ, Trevaks D, and McAllen RM. Preoptic activation and connectivity594 during thermal sweating in humans. Temperature 1: 135-141, 2014.595Farrell MJ, Trevaks D, Taylor NAS, and McAllen RM. Brain stem596 representation of thermal and psychogenic sweating in humans. American597journal of physiology Regulatory, integrative and comparative physiology 304:598 R810-817, 2013.599Fechir M, Klega A, Buchholz HG, Pfeifer N, Balon S, Schlereth T, Geber C,600Breimhorst M, Maihofner C, Birklein F, and Schreckenberger M. Cortical601 control of thermoregulatory sympathetic activation. The European journal of602neuroscience 31: 2101-2111, 2010.603Fechir M, Schlereth T, Purat T, Kritzmann S, Geber C, Eberle T, Gamer M,604and Birklein F. Patterns of sympathetic responses induced by different stress605 tasks. The open neurology journal 2: 25-31, 2008.606Fredrikson M, Furmark T, Olsson MT, Fischer H, Andersson J, and607Langstrom B. Functional neuroanatomical correlates of electrodermal activity: a608 positron emission tomographic study. Psychophysiology 35: 179-185, 1998.609Gilbert SS, van den Heuvel CJ, Ferguson SA, and Dawson D. Thermoregulation610 as a sleep signalling system. Sleep medicine reviews 8: 81-93, 2004.611Greve DN, and Fischl B. Accurate and robust brain image alignment using612 boundary-based registration. NeuroImage 48: 63-72, 2009.613Hagbarth KE, Hallin RG, Hongell A, Torebjork HE, and Wallin BG. General614 characteristics of sympathetic activity in human skin nerves. Acta physiologica615Scandinavica 84: 164-176, 1972.616Hajcak G, McDonald N, and Simons RF. To err is autonomic: error-related617 brain potentials, ANS activity, and post-error compensatory behavior.618Psychophysiology 40: 895-903, 2003.619Henson RN, and Friston KJ. Convolution models for fMRI. In: Statistical620Parametric Mapping: The analysis of functional brainimages, edited by Friston K,621 Ashburner J, Kiebel S, Nichols TE, and Penny W. London: Elsevier, 2006, p. 178-622 192.623Homma S, Nakajima Y, Toma S, Ito T, and Shibata T. Intracerebral source624 localization of mental process-related potentials elicited prior to mental625 sweating response in humans. Neuroscience letters 247: 25-28, 1998.626Iwase S, Ikeda T, Kitazawa H, Hakusui S, Sugenoya J, and Mano T. Altered627 response in cutaneous sympathetic outflow to mental and thermal stimuli in628 primary palmoplantar hyperhidrosis. Journal of the autonomic nervous system629 64: 65-73, 1997.630James C, Henderson L, and Macefield VG. Real-time imaging of brain areas631 involved in the generation of spontaneous skin sympathetic nerve activity at632 rest. NeuroImage 74: 188-194, 2013.633Jenkinson M, Bannister P, Brady M, and Smith S. Improved optimization for634 the robust and accurate linear registration and motion correction of brain635 images. NeuroImage 17: 825-841, 2002.636Jenkinson M, and Smith S. A global optimisation method for robust affine637 registration of brain images. Medical image analysis 5: 143-156, 2001.638 Kunimoto M, Kirno K, Elam M, and Wallin BG. Neuroeffector characteristics of639 sweat glands in the human hand activated by regular neural stimuli. The Journal640of physiology 442: 391-411, 1991.641Kuno Y. Human perspiration. Sprinfield, Illinois: C.C. Thomas, 1956.642Lader MH, and Montagu JD. The psycho-galvanic reflex: a pharmacological643 study of the peripheral mechanism. Journal of neurology, neurosurgery, and644psychiatry 25: 126-133, 1962.645Lawler JC, Davis MJ, and Griffith EC. Electrical characteristics of the skin. The646 impedance of the surface sheath and deep tissues. The Journal of investigative647dermatology 34: 301-308, 1960.648List CF, and Peet MM. Sweat secretion in man. I. Sweating responses in normal649 persons. Arch Neurol & Psychiat 38: 1228-1237, 1938.650Macefield VG, and Wallin BG. The discharge behaviour of single sympathetic651 neurones supplying human sweat glands. Journal of the autonomic nervous652system 61: 277-286, 1996.653Machado-Moreira CA, McLennan PL, Lillioja S, van Dijk W, Caldwell JN, and654Taylor NAS. The cholinergic blockade of both thermally and non-thermally655 induced human eccrine sweating. Experimental physiology 97: 930-942, 2012.656Machado-Moreira CA, and Taylor NAS. Psychological sweating from glabrous657 and nonglabrous skin surfaces under thermoneutral conditions.658Psychophysiology 49: 369-374, 2012a.659Machado-Moreira CA, and Taylor NAS. Sudomotor responses from glabrous660 and non-glabrous skin during cognitive and painful stimulations following661 passive heating. Acta physiologica 204: 571-581, 2012b.662Nagai Y, Critchley HD, Featherstone E, Trimble MR, and Dolan RJ. Activity in663 ventromedial prefrontal cortex covaries with sympathetic skin conductance664 level: a physiological account of a "default mode" of brain function. NeuroImage665 22: 243-251, 2004.666Nakamura K. Central circuitries for body temperature regulation and fever.667American journal of physiology Regulatory, integrative and comparative668physiology 301: R1207-1228, 2011.669Nakazato Y, Tamura N, Ohkuma A, Yoshimaru K, and Shimazu K. Idiopathic670 pure sudomotor failure: anhidrosis due to deficits in cholinergic transmission.671Neurology 63: 1476-1480, 2004.672Nieuwenhuis S, Ridderinkhof KR, Blom J, Band GP, and Kok A. Error-related673 brain potentials are differentially related to awareness of response errors:674 evidence from an antisaccade task. Psychophysiology 38: 752-760, 2001.675Noppen M, Sevens C, and Vincken WG. Effects of non-pharmacological676 sympathetic sudomotor denervation on sweating in humans with essential677 palmar hyperhidrosis. Clinical biochemistry 30: 171-175, 1997.678Nunneley SA, Martin CC, Slauson JW, Hearon CM, Nickerson LD, and Mason679 PA. Changes in regional cerebral metabolism during systemic hyperthermia in680 humans. J Appl Physiol (1985) 92: 846-851, 2002.681Ogawa T. Thermal influence on palmar sweating and mental influence on682 generalized sweating in man. The Japanese journal of physiology 25: 525-536,683 1975.684Ogawa T, and Bullard RW. Characteristics of subthreshold sudomotor neural685 impulses. J Appl Physiol 33: 300-305, 1972.686 Patterson JC, 2nd, Ungerleider LG, and Bandettini PA. Task-independent687 functional brain activity correlation with skin conductance changes: an fMRI688 study. NeuroImage 17: 1797-1806, 2002.689Qian S, Jiang Q, Liu K, Li B, Li M, Li L, Yang X, Yang Z, and Sun G. Effects of690 short-term environmental hyperthermia on patterns of cerebral blood flow.691Physiology & behavior 128: 99-107, 2014.692Robertshaw D. Neuroendocrine control of sweat glands. The Journal of693investigative dermatology 69: 121-129, 1977.694Shafton AD, and McAllen RM. Location of cat brain stem neurons that drive695 sweating. American journal of physiology Regulatory, integrative and comparative696physiology 304: R804-809, 2013.697Smith SM, Jenkinson M, Woolrich MW, Beckmann CF, Behrens TE, Johansen-698Berg H, Bannister PR, De Luca M, Drobnjak I, Flitney DE, Niazy RK, Saunders699J, Vickers J, Zhang Y, De Stefano N, Brady JM, and Matthews PM. Advances in700 functional and structural MR image analysis and implementation as FSL.701NeuroImage 23 Suppl 1: S208-219, 2004.702van Beaumont W, Bullard RW, and Banjeree MR. Observations on human703 sweating by resistance hygrometry. Dermatol Digest 1966.704Williams LM, Brammer MJ, Skerrett D, Lagopolous J, Rennie C, Kozek K,705Olivieri G, Peduto T, and Gordon E. The neural correlates of orienting: an706 integration of fMRI and skin conductance orienting. Neuroreport 11: 3011-3015,707 2000.708Woolrich MW, Jbabdi S, Patenaude B, Chappell M, Makni S, Behrens T,709Beckmann C, Jenkinson M, and Smith SM. Bayesian analysis of neuroimaging710 data in FSL. NeuroImage 45: S173-186, 2009.711Worsley KJ, Evans AC, Marrett S, and Neelin P. A three-dimensional statistical712 analysis for CBF activation studies in human brain. Journal of cerebral blood flow713and metabolism : official journal of the International Society of Cerebral Blood714Flow and Metabolism 12: 900-918, 1992.715Zhang S, Hu S, Chao HH, Ide JS, Luo X, Farr OM, and Li CS. Ventromedial716 prefrontal cortex and the regulation of physiological arousal. Social cognitive and717affective neuroscience9: 900-908, 2014.718719720 Table 1721 Clusters of significant activation during psychogenic sweating722MNI CoordinatesRegionBA Side x y z z ScoreSupplementary Motor Area6 R2 -12 52 4.73Cingulate Cortex24 R6 8 42 4.7732 L -2 12 40 4.71 Middle Frontal Gyrus10 R 34 58 6 4.33Precentral Gyrus6 R 46 -2 50 4.574 L -38 -10 50 4.47 Premotor Cortex6 R 50 6 34 3.996 L -52 4 36 4.77Inferior Parietal Lobule40 R 50 -26 38 4.6540 L -62 -34 40 4.48 Superior Parietal Lobule7 R 38 -66 40 3.777 L -28 -68 34 4.71Precuneus7 L -4 -74 46 4.97Middle Temporal Gyrus37 L -52 -64 2 4.46Lingual Gyrus18 R8 -90 8 3.89Cuneus18 L -8 -96 10 3.63Middle Occipital Gyrus19 R 56 -62 -6 3.8319 L -48 -66 -10 4.60Fusiform Gyrus19 R 34 -68 -16 4.4619 L -48 -64 -14 4.47 Insula13 R 36 22 6 4.2213 L -34 20 6 4.92 ThalamusR 14 -14 6 4.18L -14 -18 8 3.63Globus PallidusR 18 0 -4 3.69