A modality‐specific dysfunction of pain processing in schizophrenia

Abstract Clinical observations showed that schizophrenia (SCZ) patients reported little or no pain under various conditions that are commonly associated with intense painful sensations, leading to a higher risk of morbidity and mortality. However, this phenomenon has received little attention and its underlying neural mechanisms remain unclear. Here, we conducted two experiments combining psychophysics, electroencephalography (EEG), and functional magnetic resonance imaging (fMRI) techniques to investigate neural mechanisms of pain insensitivity in SCZ patients. Specifically, we adopted a stimulus–response paradigm with brief stimuli of different sensory modalities (i.e., nociceptive, non‐nociceptive somatosensory, and auditory) to test whether pain insensitivity in SCZ patients is supra‐modal or modality‐specific, and used EEG and fMRI techniques to clarify its neural mechanisms. We observed that perceived intensities to nociceptive stimuli were significantly smaller in SCZ patients than healthy controls, whereas perceived intensities to non‐nociceptive somatosensory and auditory stimuli were not significantly different. The behavioral results were confirmed by stimulus‐evoked brain responses sampled by EEG and fMRI techniques, thus verifying the modality‐specific nature of the modulation of nociceptive information processing in SCZ patients. Additionally, significant group differences were observed in the spectral power of alpha oscillations in prestimulus EEG and the seed‐based functional connectivity in resting‐state fMRI (seeds: the thalamus and periaqueductal gray that are key nodes in ascending and descending pain pathways respectively), suggesting a possible contribution of cortical–subcortical dysfunction to the phenomenon. Overall, our study provides insight into the neural mechanisms of pain insensitivity in SCZ and highlights a need for systematic assessments of their pain‐related diseases.

cephalography (EEG), and functional magnetic resonance imaging (fMRI) techniques to investigate neural mechanisms of pain insensitivity in SCZ patients. Specifically, we adopted a stimulus-response paradigm with brief stimuli of different sensory modalities (i.e., nociceptive, non-nociceptive somatosensory, and auditory) to test whether pain insensitivity in SCZ patients is supra-modal or modality-specific, and used EEG and fMRI techniques to clarify its neural mechanisms. We observed that perceived intensities to nociceptive stimuli were significantly smaller in SCZ patients than healthy controls, whereas perceived intensities to non-nociceptive somatosensory and auditory stimuli were not significantly different. The behavioral results were confirmed by stimulus-evoked brain responses sampled by EEG and fMRI techniques, thus verifying the modality-specific nature of the modulation of nociceptive information processing in SCZ patients. Additionally, significant group differences were observed in the spectral power of alpha oscillations in prestimulus EEG and the seedbased functional connectivity in resting-state fMRI (seeds: the thalamus and periaqueductal gray that are key nodes in ascending and descending pain pathways respectively), suggesting a possible contribution of cortical-subcortical dysfunction to the phenomenon. Overall, our study provides insight into the neural mechanisms of pain insensitivity in SCZ and highlights a need for systematic assessments of their pain-related diseases.

K E Y W O R D S
electroencephalogram, functional magnetic resonance imaging, pain sensitivity, schizophrenia, sensory processing 1 | INTRODUCTION Schizophrenia (SCZ) is a psychiatric disorder marked by a large spectrum of positive symptoms (e.g., hallucination, delusion), negative symptoms (e.g., anhedonia, apathy), and cognitive impairments (Kendler, 2016;Tandon et al., 2013). Despite mental suffering, SCZ patients have an extremely high risk of various pain-related diseases, such as irritable bowel syndrome, cardiovascular diseases, stroke, fractures, and diabetes mellitus (Fan, Wu, Shen, Ji, & Zhan, 2013;Garakani et al., 2003;Li, Fan, Tang, & Cheng, 2014;. However, a lower prevalence of perceived pain is reported in SCZ patients as compared to patients with other psychosis or general population (Chaturvedi, 1987;Engels et al., 2014;Stubbs, Eggermont, et al., 2015;. This observation could be due to the abnormal pain sensitivity in SCZ patients, which is pervasive but an ignored topic in clinic. The abnormal pain response profile could result in lower possibility for SCZ patients to seek medical help under various conditions normally associated with severe pain (Engels et al., 2014), which could lead to higher morbidity and mortality rates De Hert, Correll, et al., 2011;Dworkin, 1994;Jarcho, Mayer, Jiang, Feier, & London, 2012). These facts prompt the urgency to explore the interrelationship between pain and SCZ.
To test whether pain insensitivity in SCZ is a supra-modal or modality-specific phenomenon, we conducted two experiments combining with psychophysics, electroencephalography (EEG), and functional magnetic resonance imaging (fMRI) techniques. In Experiment 1 (n = 42), we used a stimulus-response paradigm to obtain stimulusevoked and prestimulus EEG data and compared sensory processing across three different modalities (i.e., nociceptive, non-nociceptive somatosensory, and auditory) between SCZ and HC. In Experiment 2 (n = 40), we further compared fMRI blood-oxygen-level dependent (BOLD) responses to nociceptive stimuli and seed-based resting-state functional connectivity (RSFC) between SCZ and HC.

| Subjects
Twenty-one right-handed patients, diagnosed with SCZ according to DSM-V (Bhati, 2013) at the inpatient and outpatient psychiatric services of Beijing Huilonguan Hospital, were recruited through clinical assessments in Experiment 1. Further inclusion criteria were as follows: illness duration longer than 2 years, considered clinically stable by their treating physician, no electric compulsive treatment in the past 6 months. Exclusion criteria included concomitant severe medical or neurological illness, comorbidity of any other DSM-V Axis I disorder, past or current alcohol abuse, high suicidal risk, and contraindications of MRI scanning. Twenty-one age-gender matched, righthanded HC without positive personal/family history of psychosis or current concomitant physical pain were recruited from the local communities. Other exclusion criteria for HC were identical to SCZ. All subjects provided written informed consent. The experiment was approved by the Ethics Review Board at the Beijing Huilonguan Hospital and registered with ChiCTR-BOC-17013972 in the Chinese Clinical Trial Registry. All experimental procedures were carried out in accordance with the Declaration of Helsinki.

| Clinical assessments
For SCZ, clinical assessments included age at illness onset, duration of illness, personal and family psychopathology, substance use, and medication. The personal and family psychopathology was assessed using the Positive and Negative Syndrome Scale (PANSS) (Kay, Fiszbein, & Opler, 1987). Substance use, particularly the duration and daily dose of smoke, was evaluated, as nicotine can modify subjective pain sensitivity (Girdler et al., 2005). Type and dosage of antipsychotic were recorded, and the daily antipsychotic dose was converted into chlorpromazine equivalents (Gardner, Murphy, O'Donnell, Centorrino, & Baldessarini, 2010). Clinical assessments on HC, which included personal and family psychopathology and substance use, were conducted using a semi-structural clinical interview.
Sociodemographic and clinical data are summarized in Table 1.
prompted the subjects to verbally rate the perceived intensity within 5 s on an 11-point Numeric Rating Scale (NRS) ranging from 0 to 10.

| Event-related potentials: Time-domain analysis
For each subject, epochs belonging to the same sensory modality were averaged, yielding three average waveforms time-locked to the F I G U R E 1 EEG and fMRI experimental paradigms. Top panel: EEG experiment (Experiment 1) was composed of three blocks, and in each block 30 trials with transient stimuli belonging to three different sensory modalities (i.e., nociceptive, non-nociceptive somatosensory, and auditory) were delivered in a pseudorandom order. Subjects were allowed to have a short break (3-8 min) after each block. Each trial started with a 4-s fixation of white cross-centered on the screen, and followed by the delivery of a sensory stimulus. A visual cue presented 2 s after the sensory stimulus prompted the subjects to verbally rate the perceived intensity within 5 s on a 11-point NRS ranging from 0 (no sensation) to 10 (unbearable sensation). The inter-trial interval (ITI) was 3-5 s. Bottom panel: fMRI experiment (Experiment 2) contained a single block of 30 trials with transient nociceptive stimuli. Each trial started with a 6-s fixation of the white cross-centered on the screen, and followed by the delivery of a nociceptive stimulus. A visual cue presented 14 s after the nociceptive stimulus prompted the subjects to rate the perceived intensity within 6 s on the same 11-point NRS. The ITI was 1-2 s stimulus onset. For laser-evoked potentials (LEPs), peak latencies and amplitudes of N1, N2, and P2 waves were measured from the average waveform. N2 and P2 waves were defined as the most negative and positive deflections between 150 and 500 ms after stimulus onset at the central electrode (Cz-nose, vertex potentials), respectively (Kunde & Treede, 1993;Valentini et al., 2012). N1 wave, defined as the most negative deflection preceding the N2 wave, can be optimally detected at the central electrode contralateral to the stimulated side referenced to Fz (C4-Fz) (Hu, Mouraux, Hu, & Iannetti, 2010;Valentini et al., 2012). For non-nociceptive somatosensoryevoked potentials (SEPs) and auditory-evoked potentials (AEPs), peak latencies and amplitudes of N2 and P2 waves were measured from the average waveforms. For both SEPs and AEPs, N2 and P2 waves were defined as the most negative and positive deflections between 100 and 400 ms after stimulus onset at the central electrode (Cz-nose, vertex potentials), respectively (Mouraux & Iannetti, 2009;Peng, Hu, Zhang, & Hu, 2012). Please note that vertex potentials elicited by intense stimuli belonging to non-nociceptive somatosensory and auditory modalities are functionally similar to the N2-P2 complex in LEP responses (Mouraux & Iannetti, 2009).
For this reason, we used the same nomenclatures (i.e., N2 and P2 waves) for all sensory modalities in the present study. Single-subject average waveforms of each sensory modality were averaged across subjects to obtain group-level waveforms. Group-level scalp topographies at the peak latency of all waves were computed by spline interpolation.

| Subjects
Twenty patients and 20 age-gender matched HC were recruited in Experiment 2, and most of them (15 patients and 17 HC) also participated in Experiment 1. The inclusion/exclusion criteria and other experimental requirements were identical to Experiment 1. Sociodemographic and clinical data are summarized in Table 1.

| fMRI experimental design
The fMRI experiment consisted of a resting-state fMRI session and a task fMRI session ( Figure 1, bottom panel). For the resting-state fMRI scanning session, subjects were required to lay supine in the scanner, and kept their eyes fixed on a white cross centered on the screen for 10 min. For the task fMRI scanning session, the paradigm was similar to that in Experiment 1, but only nociceptive stimuli were delivered as the main aim of this experiment was to explore the neural mechanisms underlying the alterations of pain sensitivity in SCZ. Thirty nociceptive stimuli (laser pulses were also generated by Stimul 1,340, which is an MRI-compatible device) were delivered to a squared area (4 × 4 cm 2 ) on the dorsum of the left hand. Other laser parameters (i.e., wavelength, pulse duration, beam diameter, and stimulus intensity) were identical to Experiment 1. Please note that, the laser beam target was manually shifted by an experimenter in the scanning room for at least 1 cm in a random direction after each stimulus to prevent fatigue or sensitization of the nociceptors. Each trial started with a 6-s fixation of the white cross-centered on the screen, followed by the delivery of a nociceptive stimulus. Fourteen seconds after the delivery of nociceptive stimulus, a visual cue was presented to prompt the subjects to rate the perceived intensity within 6 s on the same 11-point NRS by pressing buttons on a shank in their right hand. The ITI was 1-2 s.

| MRI data acquisition
Both structural and functional MRI data were acquired on a 3.  (Jenkinson et al., 2002) and then transformed to a standard brain (MNI 152 2 mm template) using FLIRT and FMRIB's Nonlinear Image Registration Tool (FNIRT) (Andersson, Jenkinson, & Smith, 2007). Group-level statistical analyses were carried out using a mixed-effect approach (FLAME, FMRIB's Local Analysis of Mixed Effects) (Beckmann, Jenkinson, & Smith, 2003;Woolrich, Behrens, Beckmann, Jenkinson, & Smith, 2004). The single-subject contrast maps entered a one-sample t test to obtain the group-level brain responses to nociceptive stimuli for each group (SCZ and HC). The differences between SCZ and HC were assessed using the independent-sample t test of single-subject contrast maps. The statistical images were thresholded using cluster-forming correction determined by Z > 2.3 and a corrected cluster significance threshold of p < .05 (Worsley, 2003).

| Resting-state fMRI: Functional connectivity analysis
Given that the thalamus and periaqueductal gray (PAG) are key nodes in the ascending and descending pain modulation pathways respectively (Ab Aziz & Ahmad, 2006;Basbaum & Fields, 1984), and showed significant differences in BOLD responses to nociceptive stimuli between SCZ and HC detected in the task fMRI GLM analysis, these two brain regions were defined as two seed ROIs for the functional connectivity analysis of resting-state fMRI data. The ROI of thalamus was defined from the Harvard Oxford subcortical structural atlas (Frazier et al., 2005), which are population-based probability atlas in MNI 152 standard space. The ROI of PAG was defined from Duvernoy's atlas of the Human Brainstem and Cerebellum (Naidich et al., 2009) in MNI 152 standard space (Ezra, Faull, Jbabdi, & Pattinson, 2015). To investigate the RSFC of each ROI in individual functional space using FEAT (FMRI Expert Analysis Tool, Version 6.00), the two seed ROIs in standard space were first transformed into individual low-resolution functional space via inverted registration files with nearest-neighbor interpolation. Voxel-wise seed-based RSFC analyses were completed using standard methods (Segerdahl, Themistocleous, Fido, Bennett, & Tracey, 2018) for each seed ROI as follows. The mean time series of a given ROI was set as a connectivity EV with realignment parameters, averaged white matter signal and cerebrospinal fluid signal as the nuisance regressors.
Functional connectivity maps were transformed straight into standard space following the same registration steps as task fMRI. The differences in the RSFC of each ROI between SCZ and HC were assessed using the independent-sample t test. The statistical images were thresholded using cluster-forming correction determined by Z > 2.3 and a corrected cluster significance threshold of p < .05 (Worsley, 2003).

| Statistics
The average ratings of the perceived intensity and event-related EEG responses were compared using two-way mixed-design analyses of variance (ANOVA), with "modality" (three levels: nociceptive, non-nociceptive somatosensory, and auditory) as a within-subject factor and "group" (two levels: SCZ and HC) as a between-subject factor. The statistical P values were adjusted with Greenhouse-Geisser correction to avoid violation of the sphericity assumption, when necessary. When the main effects or the interaction reach statistical significance, post hoc pairwise comparisons with Bonferroni correction were performed. The peak latency and amplitude of N1 wave in LEPs were compared between groups using an independent-sample t test.
In addition, Pearson's correlation analyses were performed between the ratings of perceived pain and a series of variables, including: (1) laser-evoked EEG responses (i.e., N1, N2, and P2 amplitudes), (2) prestimulus EEG oscillations (i.e., lower and higher α oscillations, which were detailed in the Supplementary materials), (3) laser-evoked BOLD responses (BOLD responses in brain regions showed significant group differences), and (4) seed-based RSFC (RSFC between thalamus/PAG and clusters showed significant group differences). To quantify the relationship between EEG and fMRI measures at resting state, Pearson's correlation analyses were also performed between prestimulus EEG oscillations and seed-based RSFC.

| Psychophysics
For both experiments, no significant differences were observed between SCZ and HC in age, years of education, and substance use (Table 1). For ratings of perceived intensity to sensory stimuli, significant main effect of "group" (F [1,40] = 7.002, p = .012, η 2 p = 0.149) and interaction between "group" and "modality" (F [2,40] = 3.573, p = .033, η 2 p = 0.082) were observed. Post hoc pairwise comparisons revealed that the perceived intensities to nociceptive stimuli in SCZ were significantly smaller than that in HC (p = .006), while the perceived intensities to non-nociceptive somatosensory (p = 1.000) and auditory (p = 1.000) stimuli were not significantly different between the two groups (Table 2;  In line with previous studies (Hu, Cai, Xiao, Luo, & Iannetti, 2014;Valentini et al., 2012), scalp topographies of the N1 wave were maximal at central electrodes contralateral to the stimulated hand, as it has been demonstrated that N1 wave is generated in the contralateral primary somatosensory/motor cortices. Scalp topographies of the N2 wave were maximal at the vertex and extended bilaterally towards temporal regions, and scalp topographies of the P2 wave were more centrally distributed (Mouraux & Iannetti, 2009   In summary, all tested variables in the event-evoked EEG responses to nociceptive stimuli were significantly different between SCZ and HC, represented as smaller peak amplitude and longer peak latency in SCZ than in HC. Whereas N2 latency, P2 latency and amplitude in the event-evoked EEG responses to non-nociceptive somatosensory and auditory stimuli were not significantly different between the two groups, N2 amplitude was significantly smaller in SCZ than HC. These results demonstrated that different from non-nociceptive somatosensory and auditory modalities, there is an overall dysfunction of nociceptive information processing in SCZ, which is in line with the behavioral data.

T A B L E 2 Comparisons of psychophysics and electrophysiological features between SCZ and HC
Note: Data are expressed in mean ± SD.

| Resting-state fMRI functional connectivity
When thalamus was used as the ROI for the functional connectivity analysis of resting state fMRI data, we observed that thalamus exhibited weaker RSFC with the right S1, right S2, left posterior insula (PI) in HC than in SCZ (Z > 2.3, p < .05 corrected; Figure 5, top panel), suggesting an abnormal function of the ascending pain pathway at resting state in SCZ. In contrast, when PAG was used as the ROI for the same functional connectivity analysis, PAG showed stronger RSFC with the supplementary motor area (SMA),

F I G U R E 2 Comparison of behavioral variables and electrophysiological features between SCZ and HC. Top left panel:
Perceived intensities to different sensory stimuli. While the perceived intensities to nociceptive stimuli were significantly smaller for SCZ compared to HC, the perceived intensities to non-nociceptive somatosensory and auditory stimuli showed no significant differences between the two groups. Top right, middle, and bottom panels: Event-related EEG responses to different sensory stimuli in the time domain. Latencies and amplitudes of all LEP waves (i.e., N1, N2, and P2) were significantly different between SCZ and HC. In contrast, no significant differences between SCZ and HC were found in N2 latency, P2 latency, and P2 amplitude of SEPs and AEPs. Notably, for both SEPs and AEPs, N2 amplitudes were significantly smaller in SCZ than HC (*p < .05; **p < .01; ***p < .001; ns: not significant) dACC, and dorsolateral prefrontal cortex (DLPFC) in HC than in SCZ (Z > 2.3, p < .05 corrected; Figure 5, bottom panel), indicating a possible degenerated function of the descending pain modulation pathway in SCZ. (r = −0.396, p = .020; Table S3).

| Correlation results
When assessing the relationship between EEG and fMRI measures at resting state, significant correlations were observed (1) between prestimulus lower alpha oscillations (6-7 Hz) and RSFC of thalamus with S1 (r = 0.533, p = .001) and S2 (r = 0.455, p = .007), and F I G U R E 4 Event-evoked BOLD responses to nociceptive stimuli. Top panel: For HC, nociceptive stimuli elicited significant activations in the PAG, thalamus, S1, S2, insula, and dACC. Middle panel: For SCZ, nociceptive stimuli elicited significant activations in the right S2 and bilateral insula. Bottom panel: Brain activations were significantly smaller in SCZ than HC in almost all brain regions associated with nociceptive information processing, including the PAG, thalamus, S2, insula, and dACC F I G U R E 5 Resting-state fMRI functional connectivity. Top panel: Thalamus showed weaker resting-state functional connectivity with the right S1, right S2, left posterior insula in HC than in SCZ. Bottom panel: PAG had stronger resting-state functional connectivity with the SMA, dACC, and DLPFC in HC than in SCZ (2) between prestimulus higher alpha oscillations (9-10 Hz) and RSFC of thalamus with S1 (r = 0.539, p = .001; Table S3).

| DISCUSSION
In the present study, combining with psychophysics, EEG, and fMRI techniques, we comprehensively investigated the difference of sensory processing across modalities between SCZ and HC. We obtained two main findings. First, SCZ patients were insensitive to nociceptive stimuli compared to HC, as revealed by three lines of evidence: (1) lower pain ratings (Table 2;  showed weaker functional connectivity with the SMA, dACC, and DLPFC in SCZ than HC ( Figure 5). Second, even the perceived intensity ratings and the late part of neural responses (i.e., P2 wave) to non-nociceptive somatosensory and auditory stimuli were not significantly different between SCZ and HC, N2 amplitude was significantly smaller in SCZ than HC (Table 2; Figures 2 and 3). This finding, which showed the dysfunction of sensory information processing across modalities, could be associated with the abnormality of the recurrent neuronal activity within the thalamocortical system in SCZ patients: compared with HC, the magnitudes of prestimulus alpha oscillations at both occipital and central electrodes were significantly larger in SCZ (Table S1; Figure S2, right panel; Figure S3).

| Pain insensitivity in SCZ patients
Our observation that pain insensitivity in SCZ is supported by several previous studies ( patients had a decreased sensitivity to prolonged pain, which was not accompanied by any difference in the nociceptive flexion reflex response. To achieve better understandings of the supra-spinal mechanism, we performed seed-based RSFC analyses for thalamus and PAG, which are key nodes in the ascending and descending pain modulation pathways respectively. We found that thalamus exhibited stronger RSFC with the right S1, right S2, left posterior insula in SCZ than in HC ( Figure 5, top panel), which was consistent with numerous previous studies that highlighted the potential of the abnormal thalamocortical functional connectivity as a promising neurobiological marker to SCZ (Ferrarelli & Tononi, 2011;Giraldo-Chica, Rogers, Damon, Landman, & Woodward, 2018;Welsh, Chen, & Taylor, 2010;Woodward, Karbasforoushan, & Heckers, 2012). Since thalamus is a key node responsible for transmissions of sensory signals in the ascending pain modulation pathway (Andersen & Dafny, 1983), the enhanced thalamocortical functional connectivity could indicate that the ascending pain pathway was hyper-activated during rest in SCZ possibly due to sensory information overload. Therefore, the thalamocortical network dysfunction could provide an alternative explanation of pain insensitivity in SCZ. It is reasonable to speculate that overloaded irrelevant internal information (e.g., hallucination) in the thalamocortical network interrupt the transmission of nociceptive inputs, thus leading to the reduced pain experience. This surmise could be indirectly (at least partly) supported by a recent study that was focused on the role of excitatory and inhibitory systems in the pain modulation in SCZ patients applied a temporal summation paradigm before and after the activation of diffuse noxious inhibitory control (DNIC) system via a cold-pressor test (Potvin et al., 2008). A lack of temporal summation in SCZ was observed in this study, which suggested a lack of central pain sensitization in patients, as temporal summation of pain is thought to reflect the progressive enhancement of C-fiber (involving nociceptive information transmission) evoked responses in the central nervous system (Hu et al., 2014;Iannetti et al., 2003).
Additionally, we observed that PAG had weaker RSFC with the SMA, dACC, and DLPFC in SCZ than in HC ( Figure 5, bottom panel), indicating a possible pain inhibition mechanism of the descending pain modulation pathway in SCZ contributed to their pain insensitivity.
Extensive evidence has highlighted the importance of this descending pathway in modulating pain experience through inhibitory/excitatory mechanisms (De Felice et al., 2011;Tracey, 2017). As a crucial nucleus in the descending pathway, the enhanced RSFC of PAG to key brain regions of pain (e.g., the S1, thalamus and ACC) is responsible to central sensitization in chronic pain patients Segerdahl et al., 2018;Zambreanu, Wise, Brooks, Iannetti, & Tracey, 2005), whose pain sensitivity is increased. Conversely, SCZ patients, whose pain sensitivity was decreased, showed attenuated RSFC of PAG to similar brain regions. Our observation was also supported by Potvin's finding, which DNIC significantly reduced pain perception in both SCZ and HC, but such reduction was more evident in HC than in SCZ (at the end of curve [80-120 s] in figure 1 in Potvin et al. (2008).
Since DNIC involves an endogenous modulation mechanism triggered by nociceptive stimuli (Potvin et al., 2008), pain insensitivity in SCZ could not be due to the enhanced functioning of endogenous inhibitory systems. On the contrary, as the pain sensitivity is diminished in SCZ, there is no need for descending inhibitory system to modulate pain. As a longitudinal consequence, the function of the descending pain modulation pathway in SCZ could be degenerated according to the theory of use and disuse.

| Abnormalities across sensory modalities in SCZ patients
In addition to pain insensitivity, dysfunction of sensory information processing across modalities was also observed in SCZ patients: the early part of brain responses (i.e., N2 amplitude) to non-nociceptive somatosensory and auditory stimuli was significantly smaller in SCZ than HC (Table 2; Figures 2 and 3). Similar to LEP responses, vertex potentials elicited by intense stimuli belonging to different sensory modalities (Mouraux & Iannetti, 2009) largely reflect saliency-related neural processes possibly related to the detection of relevant changes in the sensory environment (Downar, Crawley, Mikulis, & Davis, 2002).
Considering that N2 wave is mainly generated from the insula that is an interoceptive integration brain structure playing a crucial role in the salience network, as it conveys multisensory information about internal body state and external surrounding environment (Craig, 2009), previous studies suggested that dysfunction of sensory information processing across modalities in SCZ patients could represent an epiphenomenon of salience network dysfunctions (Alustiza et al., 2018;Liddle et al., 2016;Minichino et al., 2016;Palaniyappan & Liddle, 2012;Potvin et al., 2008;Smucny, Wylie, Kronberg, Legget, & Tregellas, 2017). The salience network is involved in detecting and filtering salient stimuli and functions to segregate the most prominent information among internal and external stimuli in order to guide behavior (Legrain, Iannetti, Plaghki, & Mouraux, 2011;Mouraux, Diukova, Lee, Wise, & Iannetti, 2011). In accordance with this notion, the salience network dysfunction in SCZ patients would result in reduced ability to distinguish self-initiated neural activity from neural activity evoked by external stimuli, which contributes to the dysfunction of sensory information processing across modalities and some psychotic symptoms, for example, hallucination (Palaniyappan & Liddle, 2012). Notably, we are aware that the majority of LEP responses are nonspecific to pain. However, they can still provide important information related to the state of the afferent nociceptive system, and be potentially useful to better understand the neural mechanisms of pain modulation through well-designed control or longitudinal studies in clinical settings (Mouraux & Iannetti, 2018). For example, LEPs can help document the deficit of the nociceptive system (e.g., lesions in the spinothalamic tract), and thus are recommended as a diagnostic tool to distinguish patients with hyperalgesia or neuropathic pain from healthy populations (Treede, Lorenz, & Baumgartner, 2003).
Importantly, we observed that the magnitudes of prestimulus alpha oscillations were significantly larger in SCZ than HC (Table S1; Figure S2, right panel; Figure S3), which could also be associated with the dysfunction of sensory information processing across modalities.
Noted that such brain oscillations reflect the discharging capacity of action potentials in thalamocortical relay neurons from tonic to burst modes depended on the neuronal membrane potentials (Llinas & Jahnsen, 1982). These state-dependent oscillatory activities could characterize wakefulness/sleep, perceptual, and cognitive states (Buzsaki, Logothetis, & Singer, 2013;Freeman, 2006;Llinas, Urbano, Leznik, Ramirez, & van Marle, 2005) in an evolutionarily preserved way. Thus, our observation that the increased magnitudes of prestimulus alpha oscillations suggested an abnormal mental state at baseline in SCZ, even though such state-dependent alpha oscillations could also be modulated by other factors, for example, level of vigilance, conscious awareness, and endogenous shifts of spatial attention (Linkenkaer-Hansen, Nikulin, Palva, Ilmoniemi, & Palva, 2004;Mathewson, Gratton, Fabiani, Beck, & Ro, 2009;May et al., 2012). In fact, emerging evidence that altered brain oscillations in neurological diseases (e.g., Parkinson's disease and chronic pain) (Llinas, Ribary, Jeanmonod, Kronberg, & Mitra, 1999;Walton, Dubois, & Llinas, 2010) and neuropsychiatry disorders (e.g., depression and SCZ) (Schulman et al., 2011;Vanneste, Song, & De Ridder, 2018) has given rise to a postulated model known as thalamocortical dysrhythmia (TCD) (Llinas et al., 1999;Llinas et al., 2005;Vanneste et al., 2018). The major point behind the TCD model is that the generation of these intrinsic abnormal low-frequency oscillations in the thalamocortical network could interrupt the original state-dependent information flow between thalamus and cerebral cortex (Vanneste et al., 2018). In other words, these low-frequency brain activities could serve as a trigger for the dysfunction of thalamocortical system, in which intrinsic neuronal properties form the substrate of illness-related pathophysiology (Behrendt, 2006;Schulman et al., 2011). Specifically, the persistent thalamic neuronal hyperpolarization induced by the activation of lowthreshold (Cav3, T-type) Ca ++ channel and followed by the lowfrequency resonant recurrent interaction between thalamic and cortical neurons disrupted the normal function of thalamocortical circuit (Llinas, Ribary, Contreras, & Pedroarena, 1998). Notably, such hyperpolarization could occur by blocking the N-methyl-D-aspartic acid receptors (NMDAr) in reticular thalamus and lead to the generation of low-frequency brain oscillations, which is consistent with the major theories about the neuropathology of SCZ, including NMDAr hypofunction (Lindsley et al., 2006;Singh & Singh, 2011;Snyder & Gao, 2013), dopamine hyperfunction (Howes & Kapur, 2009;Lodge & Grace, 2011), and GABAergic neuronal inhibition (Gonzalez-Burgos & Lewis, 2008;Gordon, 2010). Based on the above understandings of the intrinsic oscillatory properties of thalamic neurons and the biochemical effects of neurotransmitter system on the thalamocortical circuit, our results that the magnitude of prestimulus alpha oscillations was significantly larger in SCZ than HC suggested a different explanation: the abnormal sensory information processing across modalities in SCZ is related to the abnormal recurrent neuronal activity evidenced by the thalamocortical dysrhythmia.

| Limitations and future directions
There are several limitations of our study. First, as all patients were receiving antipsychotic medication, we cannot exclude drug effects on pain sensitivity. Previous studies provided evidence showing that pain insensitivity in SCZ was independent of antipsychotic effects: (1) pain insensitivity in SCZ patients was reported before the introduction of antipsychotics (Hall & Stride, 1954;Marchand et al., 1959); (2) pain sensitivity was similarly reduced in both antipsychotic-free and medicated patients (Potvin et al., 2008;; (3) diminished pain sensitivity was observed in firstdegree relatives of SCZ patients (Hooley & Delgado, 2001). Nevertheless, these studies are rare and the antipsychotic effects on sensory perception are frequently addressed by pharmacologist (Catalani et al., 2014;Schreiber, Getslev, Backer, Weizman, & Pick, 1999). Thus, further studies on unmedicated patients are needed. Second, the small number of subjects (particularly the potential impact of patients' heterogeneity due to small sample size) limited the reliability and external validity of our findings. To testify the reliability of our results, we calculated values of Cohen's d for all statistical tests and further performed a post hoc test on the effectiveness of sample size. We obtained a large effect size of the detected group difference (Appelbaum et al., 2018;Cohen, 1992), and confirmed the sufficiency of our sample size (n = 42) with a large statistical power in the EEG experiment (1 − β = .99, determined by a large effect size of η 2 p = 0.1339 or 1 − β = .82, determined by a large effect size of Cohen's d = 0.8 at the significance level of 0.05). Moreover, the fMRI results verify the findings of EEG experiment, thus increasing the reliability of our findings from different aspects. Admittedly, our findings still need to be replicated in a large and independent sample. Third, it is still unclear whether pain sensitivity varies with the development of the disease (i.e., acute, remission, chronic). To address this issue, our findings ought to be tested in prodromal-phase or first-episode patients in a longitudinal study, especially considering the tremendous differences of pain sensitivity across individuals (Hu & Iannetti, 2019).
Forth, we did not detect the relationship between clinical symptoms (illness severity, positive or negative symptom, and cognitive impairments, etc.) and the dysfunction of pain processing due to the small sample size. Future studies could investigate the possible effects of these factors on pain perception in a multifactorial model, as the typical psychotic symptoms could be potential contributors to pain insensitivity in SCZ. Last but not the least, subsequent studies involving comparison of pain sensitivity between patients with SCZ versus other psychosis (e.g., major depressive disorder, bipolar disorder) are needed in the future to establish the specificity of this phenomenon in psychosis spectrum. After all the above-mentioned issues have been clarified, our findings could be of great significance as neural index coded pain insensitivity could be used as a promising and intriguing trait marker for the diagnosis of SCZ in the future (Minichino et al., 2016).

| CONCLUSIONS
Beyond a general dysfunction of cortical sensory information processing across modalities, the pain insensitivity in SCZ also relied on a specific deficit of ascending and descending pathways modulating nociceptive information processing. Our findings provide insights into the neural mechanisms of pain insensitivity in SCZ and highlight a need for systematic assessments of their pain-related diseases.

ACKNOWLEDGMENTS
We thank Dr. Xiaoyun Li for her assistance of the experimental design and procedures, and Wenxin Su, Libo Zhang, Fengrui Zhang, Huijuan Zhang for their help with data acquisition.

AUTHOR DECLARATION
All authors have seen and approved the final version of the manuscript being submitted. The article is an original work and has not received prior publication or been under consideration for publication elsewhere.

CONFLICT OF INTEREST STATEMENT
All authors declared no competing interests.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request. The data are not publicly available due to privacy or ethical restrictions.