Patients with glioma localized in right cerebral hemisphere who underwent awake surgery in Kanazawa University Hospital between August 2013 and March 2016 participated in this study (n = 20; aged 47.4 ± 10.3 years; range, 31–68 years). Patient characteristics are presented in Table 1. To note, patients with recurrence at postoperative 6-month were excluded from this study. Written informed consent was obtained from all individual participants. This study was performed according to the guidelines of the Internal Review Board of the Kanazawa University, and was approved by the medical ethics committee at Kanazawa University (No.1797-2).

High-level mentalizing was assessed during awake surgery in eight patients who made up the “with intraoperative high-level mentalizing test (HLM test)” group. In this group, the regions with positive mapping that were elicited by direct electrical stimulation (DES) were preserved. Importantly, we resected the central part of the tumor in all patients to fulfill our oncological priorities and performed the intraoperative assessments on the extended resection using the HLM test. Twelve patients who were included in the “without intraoperative HLM test” group were not assessed with the HLM test during surgery because they underwent treatment prior to our hospital’s implementation of the intraoperative HLM test (eight cases) or they were not able to complete the test due to fatigue or other surgical events (one case). Additionally, we decided not to perform the intraoperative HLM test in three cases due to oncological reasons. No significant between-group differences, namely the with and without intraoperative HLM test group, in basic characteristics were observed: age (years), 42.1 ± 6.9; 47.9 ± 11.4; WHO grade, II 2, III 4, IV 2; II 6, III 3, IV 3; volume of resection cavity (cm³), 75.7 ± 58.9; 61.6 ± 28.0; total score of Mini-mental state examination (MMSE), 28.9 ± 1.5; 27.7 ± 2.5. Moreover, 18 normal healthy volunteers (aged 46.0 ± 9.0 years; total score of MMSE, 29.3 ± 1.2) were recruited as a control group, and there was no significant difference among groups.

We used the cartoon format of the picture arrangement (PA) task of the Wechsler Adult Intelligence Scale—third edition (WAIS-III; Wechsler, 2006) as an assessment of high-level mentalizing. The WAIS-III, including the PA task, is a worldwide assessment tool for overall intelligence. The cartoon format of the WAIS PA task has been widely used in the study for social cognition (Sarfati et al., 1997; Happe et al., 1999). The task consists of 11 items, and it includes intention condition task, which is required for understanding others’ mental state or intentions, namely high-level mentalizing. During the task, 4–6 cards were presented at random in front of patients, who were asked to sort these cards into the correct order based on the story. A score was given for each item according to the WAIS-III manual (Wechsler, 2006): a score of 1 or 2 for a correct answer, and no score for incorrect answers or time-out. Through this procedure, a total score was calculated as the WAIS PA task score. All patients were assessed three times: before surgery, 1 week after surgery and 3 months after surgery. These assessments were performed by the same trained occupational therapist (R. N.). The data were collected from medical records retrospectively.

Prior to the study, we studied the relationship between the scores between the WAIS PA and the high-level mentalizing test with 65 normal healthy volunteers (aged 47.3 ± 20.8 years). The carton format HLM test we used was a typical task for high-level mentalizing, namely the false belief task (Takamiya et al., 2009). The story is as follows: “Mary, a girl, had a doll. Mary put the doll on the ground because she saw a flower and wanted to pick it. While Mary was looking at the flower, a boy (Ken) took Mary’s doll. When Mary looked for her doll where she last put it, she was surprised to find that the doll was missing”. Using Pearson’s correlation analysis, a significant positive correlation was observed (R = 0.65, p < 0.001). This result enables to confirm the adequacy of the WAIS PA task as an assessment of high-level mentalizing ability. Here are reasons why we used whole WAIS PA task to assess high-level mentalizing: (1) validity and reliability of WAIS-III including PA task has been confirmed as an entire test (Wechsler, 2006); and (2) we have found that the score of WAIS PA task well reflected high-level mentalizing accuracy from our preliminarily research as mentioned above. In addition, we performed other neuropsychological examinations including processing speed, attention, visuospatial cognition, low-level mentalizing and executive function for each patient.

Structural MR images were acquired during the 3-month postoperative period. The images had been acquired using conventional high-resolution 3DT1-weighted sequences on a 3.0 Tesla MRI scanner (igna Excite HDx 3.0T, General Electric Medical Systems). MR images were normalized to the Montreal Neurological Institute (MNI) template using SPM12¹ implemented in the Matlab environment² and then resection cavities in all patients were reconstructed using MRIcron software³. Each reconstruction was first achieved by R. N. and inspected by a neurosurgeon (M. K.).

All patients were operated using an asleep-awake-asleep technique with cortical and subcortical brain mapping achieved by electrical stimulation (Duffau et al., 2002). After a dural incision, the cortical and subcortical areas were evaluated using DES, which was delivered via a bipolar probe (5-mm tip) with a biphasic current (pulse frequency, 60 Hz; single-pulse phase duration, 1 ms; amplitude of biphasic current, 3–6 mA). Several tasks were performed intraoperatively, and were selected carefully considering optimal onco-functional balance in each patient (e.g., picture naming and movement task, the expression recognition test namely low-level mentalizing test, the HLM test, and the line bisection test, etc.). Operative view and the intraoperative assessments were recorded on a hard disk via video camera. In the intraoperative mentalizing test, we tried to perform both low- and HLM tests to optimize the postoperative social lives of the patients. However, certain criteria, including preoperative neuropsychological examinations and intraoperative patient condition assessments, are required to perform these intraoperative neuropsychological assessments. Generally, because the HLM test is more difficult than the low-level test, the HLM test is more difficult to perform intraoperatively. Therefore, sometimes we decided to only perform the low-level mentalizing task during surgery.

The typical task of high-level mentalizing, namely the false belief task, was used as the intraoperative HLM test as mentioned above (Takamiya et al., 2009). The method of presentation was adapted for awake surgery as described previously (Sarfati et al., 1997; see Supplementary Figure S1). First, three comic strips appeared automatically (1 strip/2 s) on PowerPoint^®. Next, two choices were presented on the fourth strip, and patients had to choose which one of the two answers was the most logical to complete the comic strip sequence. Cortical or subcortical areas were stimulated directly when 2nd and 3rd strips are presented. Since intraoperative assessment should be uncomplicated and patients should certainly give correct response in normal condition, we decided to use the simple HLM test for intraoperative task, which is not same as pre- and post-operative assessment tool.

Some medical professionals have the opinion that emotional deficits recover spontaneously within a few weeks after surgery, even if cortex is damaged by surgery. In line with this clinical impression, our patient groups also recovered within 3 months following surgery, even though high-level mentalizing was disturbed before and just after surgery. In general, higher brain functions consist of multiple component, therefore they are hardly damaged completely even if one component is damaged, since other regions can compensate for damaged regions (Price and Friston, 2002). Previous neuroimaging studies have demonstrated that the medial frontal cortices, dorsal temporal cortices, basal ganglia, and amygdala are related to low-level mentalizing processing. The regions related to high-level mentalizing are more widespread and include the medial and dorsal frontal, parietal, and temporal cortices in both hemispheres, but more so on the right side (Satpute and Lieberman, 2006; Lieberman, 2007). In other words, the high-level metalizing network may more easily recover, even after the damage induced by surgery (Herbet et al., 2013). However, only patients who underwent intraoperative assessment of high-level mentalizing had the same score as healthy volunteers at 3 months. Our results might provide evidence for the usefulness of intraoperative assessment of high-level mentalizing by awake surgery. Understanding others’ mental state is a basic ability required in our social life, including empathy for others, adaptation to culture, and moral reasoning (Baimel et al., 2015). Thus, preserving high-level mentalizing by awake surgery may be important for social life of patients.

Results of the VLSM analysis indicated the most significant voxels were gathered largely in the OFC, and partly in the frontal inferior triangles. To note, in line with previous study (Herbet et al., 2014), we found second-significant region in deep part of the middle frontal gyrus, though statistical power was relatively low compared with the OFC. It is now well established that the OFC and mPFC including cingulate cortex regard as central component among several associated cortices for emotional and mentalizing process (Kringelbach, 2005; Philippi et al., 2009). Especially, the OFC involved in decision-making, and its damage induce disorder of social and emotional judgment of behavior (Kringelbach, 2005; Völlm et al., 2006). Moreover, together with the amygdala, the OFC plays a role in reading others emotional state from eye expression (Gallagher and Frith, 2003). However, as mentioned above, the OFC regards as the core region for mentalizing processing that works under various mental state conditions (e.g., false belief, deception, empathy and pretense) and various task condition (e.g., simple question, stories, static images, comic strips and so on; Philippi et al., 2009). Our result also provides an evidence that the OFC plays a critical role in high-level mentalizing and its damage may cause persistent mentalizing dysfunction.

The SLF III is an association pathway which communicate the inferior parietal lobe and middle and inferior frontal gyri. Our data showed strong correlation between damage of the SLF III and mentalizing accuracy, and previous some direct and indirect studies support our results (Barbey et al., 2014; Herbet et al., 2014; Parkinson and Wheatley, 2014; Cabinio et al., 2015; Yordanova et al., 2017). For instance, involvement of the right SLF III to the reading the mind in the eye’s test, which one estimate other’s mental state from eye gaze of facial expression, has been reported using neuro-imaging analysis (Herbet et al., 2014; Cabinio et al., 2015). More recently, a direct evidence using DES during awake surgery revealed that along with the IFOF, the SLF/AF plays important role in accurately inferring complex mental states from human face (Yordanova et al., 2017). These previous studies indicated the involvement of the SLF III to low-level mentalizing. Recently, some studies raised the possibility that the SLF III also relates to higher level emotional process. The SLF III relates to emotional empathy, which means understanding the internal expression of others through visceral or affected reactions (Parkinson and Wheatley, 2014). In addition, anti-social behavior, namely disorder of emotional process relate to abnormality of the SLF III (Karlsgodt et al., 2015; Lee et al., 2016). Taken together with these recent reports and our results, we suggested that the SLF III plays critical role for high-level mentalizing process.

In our patient series, postoperative high-level mentalizing was disturbed in some patients, although the low-level one was normal. Patients whose low-level mentalizing was disturbed always showed high-level mentalizing deficit. This suggested not only low-level but also high-level mentalizing sites should be preserved. We speculate that certain fibers relate to both low- and high- level mentalizing, whereas there are fibers which relates to only high-level mentalizing. Since it is beyond the scope of our current study, further research is required.

In line with a previous study using the TBLS analysis (Herbet et al., 2014), our results indicated that the right cingulum anterior segment was associated with mentalizing. Although our finding was less significant, and did not survive under the Bonferroni correction, it is important to mention the involvement of the cingulum in mentalizing. Within some regions that were activated during the mentalizing tasks, the cortical regions connected by the cingulum, including the anterior cingulate cortex (ACC), mPFC, and OFC, in the right or both hemispheres are considered the “core region” of mentalizing processing (Völlm et al., 2006; Carrington and Bailey, 2009). Other functional MRI (fMRI) studies have demonstrated that the mPFC and ACC were associated with mental state, and strong activations of these regions were found during a task that required distinguishing intentions of others’ behavior (Mar, 2011). Our findings are thus consistent with the previous study focused on white matter, and functional role of their cortical terminations also support our results.

The FST has been implicated in the attentional network (Chen et al., 2016), language network (Duffau et al., 2002), and network of motor control (Kinoshita et al., 2015). To best of our knowledge, this study is the first to demonstrate the involvement of the FST in mentalizing processing, whereas evidence from previous studies has indicated that cortices connected by the FST relate to networks of lower and higher levels of mentalizing (Carrington and Bailey, 2009; Schurz and Perner, 2015).

The FST plays a major role in communication between the motor area and SMA proper with the posterior putamen, and the pre-SMA with rostral parts of the striatum, through the anterior limb of the internal capsule (Lehéricy et al., 2004). In addition, the FST projects into a more extended region including the mPFC (Rojkova et al., 2016). In our patients group, the mPFC and SMA were resected with high probability within these cortices. Other neuroimaging studies have demonstrated that each cortex plays a different role in mentalizing processing. For example, activation of the SMA is observed in situations requiring prediction of others’ mental state (Carrington and Bailey, 2009; Schurz and Perner, 2015), in particular in relating to the understanding of the intentions and actions of others; so-called higher level of mentalizing (Spunt and Lieberman, 2013). As mentioned above, the mPFC is one area of a core region of mentalizing and is associated with reflective reasoning concerning actions and judgments of others (Carrington and Bailey, 2009; Schurz and Perner, 2015). Even more interestingly, the mPFC may be associated with understanding the emotional state of others (de Lange et al., 2008; Van Overwalle and Baetens, 2009), since this region is activated during mentalizing processing (Saxe and Wexler, 2005; Schlaffke et al., 2015). On the other hand, the basal ganglia seem to be involved in a lower level of mentalizing processing. The basal ganglia receive affected input from the limbic cortex which is involved in emotion and motivation (Byrum et al., 1999). Moreover, the basal ganglia automatically codes the emotional trait and evaluates implications of observed behaviors (Satpute and Lieberman, 2006; Lieberman, 2007). These previous findings regard for functional role of cortical termination of the FST support our finding that the FST is associated with mentalizing processing.

High-level mentalizing and low-level mentalizing involve a dual stream network, and interact with each other (Herbet et al., 2014). However, it is not known whether these two networks can work independently or concertedly (Bohl and van den Bos, 2012). Interestingly, our study raises the possibility that the FST, which relates to both high-level and low-level mentalizing processing, is part of a common network of mentalizing processing. Further studies are required to identify the precise role of the FST in communication between these networks.

The current study has some limitations. First, our patients group includes patients who performed radiotherapy (RT). Previous studies reported that RT cause early or late cognitive dysfunction, especially attention and executive functional deficit (Klein et al., 2002; Douw et al., 2009). Though there is no evidence that whether RT influence on emotional process, we cannot deny the possibility that RT cause early or late emotional disorder. In the same line, our patients group includes patients with glioblastoma, which is not biologically same as low and intermediate grade glioma. We have excluded the patients with recurrence at postoperative 6-month. It means tumor was well controlled at least postoperative 3-month when neuropsychological examination was performed. However, further research will be required to study differences between high and low/intermediate grade glioma regards for involvement of white matter in mentalizing. Similarly, the number of patients in the patient group was not sufficient for strong statistical power. However, some neuroimaging studies have reported successful results with a relatively small number of patients (Teichmann et al., 2015; Kinoshita et al., 2016). Additional studies with more cases are required to strengthen our current results.

As previously described, activated region in mentalizing task might be different depending on task condition in cortical level (Carrington and Bailey, 2009). Nevertheless, we used only cartoon format of the PA task to confirm high-level mentalizing accuracy, because almost all regions related to high-level mentalizing were activated during comic strip task (Carrington and Bailey, 2009). In addition, from the clinical point of view, since it is essential to take into account patients’ condition after surgery, we chose PA task as optimal postoperative assessment. However, if distinct task condition was used, results might be a little different.

The regions found to be significant in the VLSM analysis, namely, the inferior frontal gyrus and orbitofrontal cortex, mainly correspond to the location of the termination of the SLF III and partially to the uncinate fasciculus (UF). Previous studies have suggested that the UF is involved in sociality because the structural connectivity of this area is abnormal in patients with autism and antisocial disorder (Sundaram et al., 2008; Thomas et al., 2011; Sarkar et al., 2013). Moreover, emotional processes, including emotional regulation and empathy, are related to the UF (Herbet et al., 2015; Olson et al., 2015). Based on these findings, a significant correlation was expected between the resected volume of the UF and the high-level mentalizing scores in the current study. However, a significant correlation was found only between the scores and the resected volume of the SLF III. These results might have been due to the small number of patients (only seven) who had a resected UF (Supplementary Table S4). The small number of patients in this study was a methodological limitation and might explain why we did not find a significant correlation between the resected volume of the UF and mentalizing accuracy.

Finally, the high-level mentalizing of some patients was disturbed, even when intraoperative assessments were performed. Two explanations are proposed. First, the two (out of three) patients with deficits remaining until the chronic phase of our study exhibited the mentalizing disorder preoperatively (see Supplementary Table S2). Consistently, our recent study revealed that the deficits in patients with preoperative neuropsychological deficits tend to remain until the chronic phase, regardless of performance on the intraoperative assessments (Nakajima et al., 2017). Second, as mentioned above, the intraoperative HLM test was not performed on all patients to optimize oncofunctional balance and due to individual factors. Hence, we might have resected the brain region responsible for high-level mentalizing because the intraoperative assessment was not performed, in order to set the priority in the onco-functional balance on the extent of resection.