Reward and aversion processing in patients with post-traumatic stress disorder: functional neuroimaging with visual and thermal stimuli

In patients with post-traumatic stress disorder (PTSD), a decrease in the brain reward function was reported in behavioral- and in neuroimaging studies. While pathophysiological mechanisms underlying this response are unclear, there are several lines of evidence suggesting over-recruitment of the brain reward regions by aversive stimuli rendering them unavailable to respond to reward-related content. The purpose of this study was to juxtapose brain responses to functional neuroimaging probes that reliably produce rewarding and aversive experiences in PTSD subjects and in healthy controls. The stimuli used were pleasant, aversive and neutral images selected from the International Affective Picture System (IAPS) along with pain-inducing heat applied to the dorsum of the left hand; all were administered during 3 T functional magnetic resonance imaging. Analyses of IAPS responses for the pleasant images revealed significantly decreased subjective ratings and brain activations in PTSD subjects that included striatum and medial prefrontal-, parietal- and temporal cortices. For the aversive images, decreased activations were observed in the amygdala and in the thalamus. PTSD and healthy subjects provided similar subjective ratings of thermal sensory thresholds and each of the temperatures. When 46 °C (hot) and 42 °C (neutral) temperatures were contrasted, voxelwise between-group comparison revealed greater activations in the striatum, amygdala, hippocampus and medial prefrontal cortex in the PTSD subjects. These latter findings were for the most part mirrored by the 44 vs. 42 °C contrast. Our data suggest different brain alterations patterns in PTSD, namely relatively diminished corticolimbic response to pleasant and aversive psychosocial stimuli in the face of exaggerated response to heat-related pain. The present findings support the hypothesis that brain sensitization to pain in PTSD may interfere with the processing of psychosocial stimuli whether they are of rewarding or aversive valence.


Introduction
Reward deficiency, that is to say, hypofunctionality of the brain reward circuitry manifested in the diminution of drives and in inability to experience joy or pleasure 1 is considered by some 2,3 to be the most specific diagnostic 4 feature of post-traumatic stress disorder (PTSD) documented in preclinical studies 5 along with behavioral 6,7 and neuroimaging 8,9 clinical research. Although such neuropsychopathology is rather resistant to conventional therapies 10,11 and is also associated with chronicity and severe disability 11,12 , its pathophysiological mechanisms remain poorly understood. One possibility is that reward hypo-responsivity is driven by an enduring brain alteration whether it be preexisting or acquired. A second possibility is that it is derived from a functional reciprocity between reward and stress reactivity 8 .
With regard to the former possibility, individuals afflicted with reward deficiency may perceive their life as bland and unfulfilling and possess a character trait of novelty seeking 13,14 , which could drive their engagement in stressogenic situations with an elevated potential for trauma exposure and subsequent PTSD 15,16 . This causality could run in the opposite direction 17 . That is to say, besides potent vasoconstriction 18 , chronic stress can exerts neurotoxic effects 19,20 via a mix of related, but conceptually and operationally different mechanisms such as aggregation of platelets 21 , upsurge of intracellular calcium 22 and acceleration of apoptosis 23 evident in structural gray matter volume changes of the key corticolimbic structures 24,25 . Inherent in these structural changes are alterations in neural connectivity and/or neurochemisitry. For instance, reward deficiency is caused by dampening reward circuitry neurotransmission by way of enhanced dopamine metabolism 26 , its inhibited synthesis 27 or extracellular release 28,29 in conjunction with the reduction in dopamine receptors' number 30 and activity 31,32 .
It may as well be plausible that reward and stress alterations arising in the context of PTSD are temporally related owing to conspicuous neuroanatomical and functional overlap between the respective neurocircuitries 33,34 . Specifically, dopamine terminal fields, including amygdala, striatum and medial prefrontal cortex that are involved in the reward and motivational processing 35 also play key roles in stress and aversion 36 . In patients with PTSD these areas 37,38 become hypersensitive to traumaconditioned environmental cues 8,39 , a mounting process leading to the generalization of fear 40,41 that is added or synergized by the anti-reward cross-sensitization neuroadaptation amplifying responses to other aversive yet not necessarily conditioned stimuli [42][43][44] . Like so, in PTSD the same brain regions may be over-recruited by the aversive stimuli rendering them unavailable to respond to reward-related content and in the reversed order in people with low reward function aversive experiences (e.g., pain) are not buffered by reward and a consequence is the heightened pain experience 45,46 . These are testable hypothesis that could be evaluated by juxtaposing responses to functional neuroimaging probes that reliably produce rewarding and aversive experiences 8 .
Inquiry into aversion mechanisms in humans is limited in part by paucity of laboratory-based procedures that bring about strong and reproducible activation of major systems and that can be controlled with respect to the 'amount' of the administered stimulus. A paradigm well suited for examining aversive responses in humans is a common stressor 47 , experimentally-induced pain 48 . Consistent with the reward-aversion continuum conceptualization 49 , the brain's pain system is embedded within extensive reward/motivation circuitry indispensable for the survival mechanisms via pursuit of nourishment while avoiding/escaping threats 44 . Even mild pain poses a sufficient aversive experience resulting in reliable brain and subjective responses 50 . Moreover, this procedure is not associated with performance confounds, so that equal 'amounts' of aversion are given to, both healthy subjects and to patients with a neuropsychiatric condition potentially entailing motivational 4 and attentional 51 deficits such as PTSD. Pain is also an ecologically valid stimulus to be used in PTSD patients as numerous epidemiological surveys indicate that the prevalence of chronic pain in PTSD patients exceeds that of the general population 52 with up to a third of pain clinics' patients afflicted with comorbid PTSD 53,54 compared to a 4-12% PTSD rate in the general population 55 .
The purpose of the present study was to determine, employing functional magnetic resonance imaging (fMRI), whether PTSD is associated with primary vs. secondary alterations in reward processing. Two challenges used were (1) aversive or pleasant (i.e., rewarding) and neutral images 56 selected from the International Affective Picture System (IAPS) and (2) pain-inducing noxious thermal stimuli 57 . The value of using these types of challenges is a more conclusive interpretation of the findings. Increased aversive stimuli (pain and negative IAPS images) responses in pain-free PTSD patients associated with signal decrements during rewarding (positive IAPS images) stimuli would support the notion that reward responsivity and pain sensitization are inversely related phenomena. Alternatively, if PTSD patients present the same directionality of the fMRI signal changes during both rewarding and aversive visual stimuli, it may be concluded that altered brain reward responses are not secondary to the over-recruitment of the brain reward regions by the aversive stimuli and a case for primary alterations in the brain reward and aversion function may be supported. Moreover, normal activity during pain, but not during aversive images' processing, would suggest intact brain pain mechanisms and that fMRI signal differences are secondary to performance of the visual task. In a similar fashion, control level activity on both challenges would indicate that the respective brain circuitries are essentially intact with regard to their response to diverse rewarding and aversive challenges. Given that theoretical considerations on the above scores are not unambiguous directional prediction on rewarding vs. aversive stimuli responses was not sufficiently justified. Therefore, the hypothesis was formulated in terms of PTSD-related differences in the brain processing of both visual-and thermal-type of stimuli.

Subjects
Twelve subjects meeting the DSM-IV-TR criteria for PTSD, diagnosed via the Structured Clinical Interview for DSM-IV 58 and Clinician-Administered PTSD Scale (CAPS) 59 , and 12 mentally healthy subjects were recruited by advertisement. After the procedures were fully explained, each subject gave written informed consent to the protocol approved by the McLean Hospital Institutional Review Board. All subjects were right-handed as assessed with Edinburgh Handedness Inventory; 60 they were pain-free and in good physical health as determined by respective Brief Pain- 61 and Cornell Medical Index Health Questionnaires 62 . Subjects with cognitive impairment or head trauma accompanied by amnesia or loss of consciousness greater than 10 min were excluded, as well as those with a history of schizophrenic-, paranoid-, other psychotic-, bipolar-, non-PTSD anxiety-, or substance dependence disorder. Given the high rate of depressive comorbidity in PTSD 63 , subjects with onset of major depressive disorder after the traumatic event that caused the PTSD were allowed to participate. Recent drug and alcohol consumption was ruled out by negative results on urine toxicology screen and breathalyzer. We also excluded the use within the previous month of any potentially confounding medications or drugs (e.g., opioids, psychostimulants, cannabinoids, dopaminergic or antidopaminergic agents, and mood stabilizers, antidepressants with prominent catecholaminergic effects such as tricylclics, buproprion, mirtazepine, venlafaxine, and duloxetine).

Visual stimulation
Similar to our prior studies in mentally healthy subjects, emotional responses were probed using images selected from the IAPS 64 . Based on normative ratings for affective valance (unpleasant to pleasant) and arousal (calm to excited), three categories of images were selected: "pleasant", "neutral", and "aversive" categories. The pleasant images were the 90 pictures with the highest normative arousal scores selected from the 120 pictures with the highest normative valence intensity scores. Similarly, the aversive images were the 90 pictures with the highest normative arousal score selected from the 120 pictures with the lowest normative valence scores. Neutral images were 120 pictures with the highest normative arousal score selected from pictures with valence scores between 4.5 and 5.5 (range 1-9).
IAPS images were presented in blocks of nine for each of the three categories (Fig. 1). Every subject had three fMRI scans, each with a total of nine visual stimulation blocks; three blocks of positive images, three blocks of aversive images and three blocks of neutral images. Each scan consisted of a 60 s baseline followed by nine visual stimulation blocks (20 s long) presented in pseudorandom order. Each image was only presented once. After each scan, subjects verbally rated the average valence experienced for the Pleasant and Aversive blocks using visual analog scale (VAS).

Quantitative sensory testing
Prior to scanning, heat and cold thresholds were determined using a 3 × 3 cm contact thermode (TSA-II, Medoc Advanced Medical Systems). The temperature increased from a 32°C baseline at the 1°C/s rate until stopped by the subject at the first onset of pain. To determine cold pain thresholds, the skin was cooled down linearly at a slow rate (1°C/s) until pain sensation was perceived (method of limits).
Similar to our prior studies in healthy subjects 65 , pain responses were probed by heat stimuli to the dorsum of the left hand delivered with a 3 × 3 cm contact thermode (TSA-II, Medoc Advanced Medical Systems). The thermode had a baseline temperature of 32°C, and was rapidly heated (temperature rise = +4°C/s) to 42, 44, or 46°C. The target temperature was maintained for 20 s and then returned to baseline (-4°C/s) to end the stimulus event. Every subject received a total of nine thermal stimuli, three at each temperature, with an inter-stimulus interval of 30 s. During each thermal stimulus, subjects rated pain intensity and unpleasantness using a rating dial in their right hand to adjust a VAS presented using the software package LabVIEW 5.1 (National Instruments Corp). Pain intensity was rated on a 0 to 10 VAS anchored at "No Pain" to "Max Pain", unpleasantness was anchored at "Min" 0 to "Max" 10. To reduce expectancy confounds the stimuli were presented in a random order.

Imaging protocol
A Siemens Trio 3 Tesla MRI scanner with a circularly polarized head coil was used for all scans. Brain structure was acquired with a magnetization prepared rapid gradient echo (MPRAGE) sequence [128 slices 1.33 mm thick, with an in-plane resolution of 1 mm (256 × 256)]. Blood-Oxygen-Level Dependent (BOLD) contrast functional scans were collected using an echo planar imaging sequence (echo time/repetition time (TE/TR) = 30/ 2500 ms for heat pain runs, TE/TR = 30/3000 ms for IAPS). The repetition times were optimized to the timing of the heat and visual probes. Both heat pain and visual functional scans consisted of 41 slices, with 3.5 mm isometric resolution. Eighty-four volumes were captured for each of the 42, 44 and 46°C fMRI scans (3:30 each), and 199 volumes were captured for the IAPS fMRI scans (9:57). Visual and thermal stimuli were administered in a double blind counterbalanced fashion at least 15 min apart (Fig. 1).

Data processing and voxelwise statistical analyses
Analysis was carried out using FSL tools release 5.0 (FMRIB Analysis Group, Oxford University; http://www. fmrib.ox.ac.uk/fsl/), specifically FEAT version number 5.92. Functional images were pre-processed using standard pipelines: motion correction, high pass temporal filtering (100 s), spatial smoothing (5 mm). Scans were inspected for gross motion with a threshold of 3 mm for elimination of the scan from further analysis. Images were registered to a standard atlas provided by FSL (MNI152 standard brain).
Statistical Analysis was carried out using a univariate general linear model approach; explanatory variables were created to represent the temporal presentation of thermal and visual stimuli. The resulting spatial parameter estimates were registered to standard atlas for group analysis. Group statistical analysis was carried out using a mixed-model approach as implemented in FSL; parameter estimate and variance images were included to perform the group comparisons described below. Inference was carried out using a Gaussian mixture model approach as described in ref. 66 . Group and comparison statistical maps were subjected to alternative hypothesis testing without assuming normal distribution. The Gaussian mixture model approach produces posterior probability maps for the different classes of the original z-statistics map. Each voxel is associated with different classes with a specific (posterior) probability of belong to each class.  . The thermode had a baseline temperature of 32°C, and was rapidly heated (temperature rise = + 4°C/s) to 42, 44 or 46°C. The target temperature was maintained for 20 s and then returned to baseline (−4°C/s) to end the stimulus event. Every subject received a total of nine thermal stimuli, three at each temperature, with an interstimulus interval of 30 s. To identify brain regions that differed between PTSD and healthy subjects, we contrasted 46 and 44°C evoked responses to 42°C. IAPS images data comparison was performed in an analogous fashion i.e., positive images minus neutral images and negative images minus neutral images. Both comparisons were calculated for each group (patients and controls) and for between group differences Posterior probability maps were thresholded at 0.5 to determine brain regions statistically significant differences between the groups. To search for brain regions that differed between PTSD and healthy subjects, we contrasted rewarding (pleasant) versus aversive (unpleasant) IAPS images (rewarding images minus neutral images and aversive images minus neutral images). Similarly, evoked responses to 46 and 44°C were compared to 42°C. Both comparisons were calculated for each group (patients and controls) and for between group differences. The t test results for each voxel were converted to z scores and thresholded to p < .01, at first uncorrected for multiple comparisons. All voxels with less significant activations (or deactivations of any magnitude) were excluded from further study. Remaining voxels were then collected into contiguous clusters. With Gaussian random field theory 67 , a significance level was associated with each cluster, this time correcting for multiple comparisons across the whole brain. Clusters with corrected significance at z > 2.3 and p < .05 were rendered as colored regions, with the color at each voxel indicating the corresponding z score.
The power analysis was based upon testing the IAPS response differences, which were likely to require more subjects consistent with a weaker response to psychosocial vs. physiological stimuli 68 . In our prior experiment with a psychosocial task 8 the mean striatal BOLD signal changes in response to monetary reward in PTSD subjects was 0.05 ± 0.17 (SD) compared to 0.33 ± 0.35 in the healthy controls, yielding an effect size of 1.02d. We assumed the effect size for rewarding IAPS images to be comparably large. With 12 subjects in each group, we had 80% power at the p < 0.05 significance level to detect such an effect size for lower responses to reward in PTSD subjects. Table 1 presents demographic and clinical data for the study groups. These data demonstrate that PTSD subjects were not significantly different from healthy controls with respect to age, gender, years of education and performance on the quantitative sensory testing, but they scored significantly higher on the Harm Avoidance and Self-Transcendence and lower on Self-Directedness. The PTSD subjects also rated pleasant images significantly lower than healthy controls. As planned, there were conspicuous differences in the CAPS and Beck Depression Inventory-II 69 scores.

Imaging data
Imaging data are displayed in Figs. 2-5 and Tables 2-5 as regions within the brain divided into Cortical, Subcortical, and Brainstem/Cerebellum with x, y, and z coordinates in millimeters of the peak voxel and cluster volumes. Significant activations are noted in terms of zstatistics (z-stat). Because our prior work implicated striatum in hypofunctional reward responsivity in PTSD patients 8 , an a priori emphasis was placed on potential activations and deactivations in that area.

Visual stimuli
For the processing of reward (Fig. 2, Table 2), betweengroup analyses of responses to the presentation of rewarding vs. neutral IAPS images in PTSD-vs. healthy subjects displayed 31 clusters of deactivation including cortical (cingulate, frontal occipital, parahippocampal, parietal and temporal), sub-cortical (right putamen and left pallidum), brainstem and cerebellum areas. Separate analyses in healthy and PTSD subjects revealed significant clusters of activations in the above regions for both groups; the clusters volumes and the level of significance were smaller in the PTSD group. For psychosocially aversive stimuli (Fig. 3, Table 3), between-group analyses of negative versus neutral IAPS images produced 12 clusters of activation (frontal occipital, parietal and temporal cortices and in cerebellum) and 9 clusters of deactivation (frontal parietal and temporal cortices, bilateral amygdala and thalamus). Separate analyses in healthy subjects detected large significant clusters of activation to negative minus neutral IAPS images that comprised

Thermal stimuli
Pair-wise group (PTSD-subjects vs. healthy subjects) comparison between noxious heat (46°C) and mildly warm temperature (42°C) (Fig. 4, Table 4) uncovered 60 clusters of activation in cortical (cingulate, frontal, hippocampal, occipital, parahippocampal, parietal, temporal and insular), sub-cortical (right amygdala, left caudate, left nucleus accumbens, right pallidum and right putamen), brainstem (periaqueductal gray) and cerebellar areas and 32 clusters of deactivation in the cortical (cingular, frontal, insular, occipital, parietal, temporal), sub-cortical (left amygdala and right hypothalamus) and cerebellar areas. In healthy subjects the 46 vs. 42°C contrast resulted in a    When the PTSD group was considered in isolation, the 46°C vs. 42°C contrast detected bilateral activation clusters in the ventral and dorsal striatum comprised of nucleus accumbens and pallidum along with the clusters in the anterior cingulate and other cortical areas; deactivations were observed bilaterally in the hippocampus. Other than the prominent thalamic activations in the PTSD group that were not apparent in the 46°C vs. 42°C contrast potentially due to activation of the descending modulation system 70 , the 44°C vs. 42°C contrast (Fig. 5, Table 5) produced by and large a similar to the 46°C vs. 42°C contrast pattern of activations (62 clusters) and deactivations (20 clusters) in the cortical, subcortical and brainstem regions on both between groups and within group analyses.

Discussion
To our knowledge, this is the first study to integrate reward and aversion subjective rating and neuroimaging data in patients with PTSD. The present results replicate others' 2 and our earlier behavioral 6 , self-report 7 and neuroimaging 8 work uncovering PTSD-related      decrements in response to rewarding visual stimuli and extend these prior findings by suggesting that, in addition to been numb to rewards, PTSD subjects may also be indifferent to some of the life's discontents operationalized via aversive IAPS images as evidenced by bilateral deactivations in the key reward and aversion structure, amygdala. PTSD neuropsychopathology may thus encompass both positive and negative valence processing whether it is subserved by the same or by a different set of neurons 71 . Decrease in cerebral metabolism and blood flow when exposed to natural reinforcers has been observed in a number of neuropsychiatric conditions (e.g., addiction and schizophrenia) characterized, like PTSD 8 , by diminished dopaminergic tone with corresponding decreases in the tonic glutamatergic activity due to drugs or to the disease process per se 35,72,73 . On the background of this diminished activity, respective exposure to drugs, to conditioned cues or to psychotic contents leads to robust augmentations of phasic corticolimbic responses 72 akin to pain-induced activations on the present study.
A prior neuroimaging investigation with the laboratorybased pain induction found greater activations in hippocampus, putamen and insula and less activations in the amygdala and prefrontal cortex of combat PTSD Veterans during their exposure to a fixed and customized (to subjective ratings) temperatures 74 . That study did not, however, obtain baseline pain assessments and subjective pain thresholds. A subsequent study in women only replicated the insular activations finding 75 . The direction of PTSD subjects' subjective responses to experimentallyinduced pain varied and resting state hyperalgesia 76,77 , hypoalgesia 74,76,78 and no differences 79 when compared to healthy subjects have been reported. Methodological factors 80 such as inter-subject pain threshold variability 74,76 , individualized vs. standardized magnitude of the pain stimuli 74 , concurrent PTSD symptoms reactivation 79 , pain expectancy context 81 and presence of comorbid pain conditions 76 may explain the divergent pain effects in PTSD.
It has been previously suggested that PTSD patients are not actually numb and that their capacity to experience positive emotions is rather constrained by preferential allocation of emotional, motivational and cognitive resources to environmental threats including reexperiencing of the traumatic episodes 82,83 . Partially overlapping hypo and hyper in the PTSD subjects (e.g., left pallidum 84 ) respectively produced by the positive IAPS images and by pain supports the possibility that PTSD patients deactivate and activate the same brain structures to respective rewarding and aversive stimuli. However, even if such structures are identified in this and prior functional and/or structural neuroimaging studies, the microcircuits located within those structures may actually carry out discrete and non-overlapping tasks. Emerging neuroscience technologies integrating viral vectors with optogenetics in combination with in vivo single cell recording, electrophysiology and neuroanatomical analyses 85,86 afford higher (than human neuroimaging) resolution of neural underpinning of normal function and of pathopysiological processes. The present findings thus provide a foundation for preclinical studies applying concurrent reward and stress measurements in PTSD models to further address the questions of reward and stress circuitries' interactions.
We also observed dissociation between brain activations and quantified measures of pain valuation. Specifically, pain free subjects with PTSD rated painful stimuli similarly to healthy controls, but displayed greater brain activations to the same stimuli. This group difference was not explained by variability of pain thresholds. Such heightened brain pain responses notwithstanding regular self-reports may point to enhanced brain's ability to screen out/suppress responses to seemingly irrelevant 87 noxious and other types of stimuli from reaching conscious awareness 88,89 i.e., "gating" 87 . While disrupted sensorimotor gating plays an important role in the course of PTSD 90 , the present finding of similar unpleasantness ratings of the aversive IAPS images in the face of decreased activations in the PTSD group renders enhanced gating an unlikely mechanism of the observed dissociations between neuroimaging findings and subjective ratings. Nonetheless, electroencephalography 91 and magnetoencephalography 92 could be used in conjunction with pain probes to examine further questions concerning sensorimotor gating mechanisms underlying PTSD symptomatology.
Another issue to consider is the cross-sensitization phenomena 44,93 . This term pertains to a situation where prior exposure to one stimulus (e.g., trauma and its consequent re-experiencing) increases subsequent response to itself and to a different stimulus (e.g., pain). The cross-sensitization did not seem to include brain responses to another aversive stimulus used on the study, i.e., negative IAPS images, which may have been attenuated because of a possible 'floor effect' given the low subjective ratings. Emotional processing may be attributed to a two-system construct 94 comprised of corticolimbic circuits mediating valence (ranging from aversive to rewarding) in conjunction with closely linked networks coding intensity-related arousal 71 . Future research may consider matching negative stimuli by the level of intensity to address the generalizability of the crosssensitization processes.
The mechanisms of cross-sensitization may involve conditioning. Thus, pain, paired with emotional trauma and its recollections, can become a conditioned stimulus that evokes fear and anxiety responses that in turn augment subjective pain perception and its neural correlates 95,96 , and so mounts the "mutual maintenance" 97 cycle, leading to additional deterioration and avoidance of pain-and trauma-related situations [98][99][100][101] . Formulation of PTSD treatment plans targeting emotional numbing might then benefit from the habituation and extinction of stressful re-experience techniques 102 along with provision of potent positive stimuli 83 .
In addition, increased central opiodergic tone 79,103 along with robust elevations of endogenous opiates concentrations in the cerebral spinal fluid 104 and in plasma 105,106 is a relatively consistent clinical finding in PTSD. Therefore, similarly to chronic users of opioid pain relievers 107,108 , PTSD-related exaggerated CNS opioidergic activity could contribute to sensitized brain pain responses mediated via the amplification of the excitatory (e.g., glutamtergic) neurotransmission [109][110][111] . If such neurobiologic vulnerability factors could be identified, they might be used to screen patients at risk for the development of pain condition. Patients found to possess high vulnerability for the development of pain owing to PTSD-related heightened opioidergic tone function might be counseled to avoid opioids (primary prevention), or targeted for early intervention with non-opioid agents 49 even in the presence of mild pain problems (secondary prevention).
Yet, in order to prevent sensitization of the healthy brain aggressive and timely analgesic treatment may actually be indicated. In fact, peritraumatic pain is a stressor recognized than an independent PTSD risk factor 112 whereas chronic pain may be construed as a variant of PTSD due to persistent relieving of stress, avoidance of pain-related situations and negative cognitions and affective states 44 . This may be why adequate morphine 113,114 or ketamine 115 analgesia reduces the severity and may even prevent the appearance of PTSD. An additional therapeutic implication of the opioidergic mechanisms' involvement in PTSD pathophysiology 79,116,117 is the clinical use of opioid antagonists 103,118,119 in some pain-free patients, that on the whole appears to be safe and well tolerated and results in significant improvements of various aspects of PTSD symptomatology such as emotional numbing, startle response, nightmares, flashbacks, intrusive thoughts and comorbid alcoholism 103,118,120 .

Caveats
Caveats that should be considered in interpreting our data refer to the type of stimuli, the duration of the study and the pilot nature of the study design. First, although the aversive state created by the thermal stimuli is qualitatively different from environmentally-induced pain that is implicated in PTSD pathophysiology 112 , we believe that our results may have clinical significance because real life pain affects similar brain areas to those produced by the heated thermode 121 . Likewise, both aversive stimuli employed may have been quantitatively different from environmental stressors that have been implicated in initiation and exacerbation of PTSD. While we involved both psychosocial and sensory components, the subjective ratings of averseness were only moderately affected. Because various stressors may have diverse effects on regulatory systems, future studies employing other types of aversive stimuli than the ones previously used by our group e.g., glucoprivation with 2-deoxyglucose 122 or adrenergic stimulation with yohimbine 123 may provide unique information pertaining to general stress and corticolimbic responsiveness. Also, even though we employed visual and sensory stimuli of aversive quality they engage different behavioral and emotional systems the overlap of which may not necessary be aversion processing per se. This systems' parameter can be isolated by comparing brain response in subjects who do experience versus who do not experience aversion from the presented stimuli.
Second, the observed group differences in reward processing may reflect a pre-existing risk factor rather than an acquired neuropsychopathology resulting from trauma exposure and subsequent PTSD. If this were the case, the PTSD subjects would have displayed purportedly heritable personality traits that are suggestive of the reward deficiency 124 . The Temperament and Character Inventory's Novelty Seeking and Reward Dependence data render this option unlikely and suggest that premorbid reward function in PTSD subjects was similar to that in the control group. Yet, the effects of premorbid factors particularly related to Harm Avoidance, Self-Directedness and Self-Transcendence that differentiated PTSD and control groups on this study is an important consideration for the future research regarding the origin of PTSDrelated reward deficits. Third, this study assessed only acute pain response while evidence suggests that such response tend to sensitize over time 125 . Therefore, longer study periods may have yielded different results. Finally, these findings should be considered as preliminary pending replication with a larger sample.

Conclusions
In conclusion, pilot data presented here suggest that reward and pain activate partially overlapping corticolimbic areas. Patients with PTSD display reward hyporesponsivity notwithstanding excessive responses to pain. At the same time, subjective group differences in response to aversive psychosocial images are not obvious. These data shed light on pathophysiology of reward and aversion disturbances to suggest their reciprocity in PTSD and call for further research aimed at understanding the distinctive features of reward vis-à-vis pain alterations and their potential role in preventive efforts and in therapeutic armamentarium for the respective patients.