Brain structural changes in patients with chronic methylmercury poisoning in Minamata

Exploratory whole-brain studies in patients suffering from methylmercury (MeHg) poisoning have not been conducted. We aimed to evaluate the neuroanatomical differences between patients with chronic MeHg poisoning and healthy volunteers via magnetic resonance (MR) imaging. Patients included in this case-control study were divided into three categories based on whether MeHg exposure occurred in utero, under 15 years of age, or over 15 years of age, as fetal-, pediatric-, and adult-type patients, respectively. This study analyzed MR imaging data from 10 patients each of fetal, pediatric, and adult types of chronic MeHg poisoning in Minamata and corresponding 53, 37, and 15 age- and sex-matched healthy volunteers. Whole-brain voxel-based morphometry (VBM) analysis was used to determine the volumetric gray and white matter (GM and WM) differences in patients with chronic MeHg poisoning. Compared to healthy individuals, VBM revealed a significant reduction in GM in the cerebellar and calcarine areas in pediatric- and adult-type cases and in the thalamus of fetal-type cases. A significant reduction in WM volume was also noted in the cerebral and the cerebellar regions, especially in pediatric-type cases. Patients with chronic MeHg poisoning develop structural differences in the GM of the calcarine, the cerebellum, and the thalamus and in the WM of the cerebrum and cerebellum. These changes can appear, depending on the timing of MeHg exposure.


Introduction
Among environmentally hazardous metals, mercury (Hg) has been identified as one of the most highly toxic elements to humans (Massányi et al., 2020). There are three main forms of mercury: elemental mercury (Hg 0 ), inorganic mercury (Hg 2+ ), and organic mercury (e.g., methylmercury); of these, methylmercury (MeHg) is a known artificial byproduct of acetaldehyde synthesis but can also occur naturally due to microbial methylation of mercury. The first large-scale disaster from artificially produced MeHg was discovered in Minamata, Japan, in 1956; the resulting illness reached epidemic proportions, later named as Minamata disease (Harada, 1995). Although extensive artificial MeHg pollution has been reduced, naturally occurring environmental MeHg has seen a steady increase worldwide due to mercury release associated with human activity (Gibb and O'Leary, 2014;Kolipinski et al., 2020;Streets et al., 2011).
MeHg at toxic exposure levels can cause central nervous system dysfunction in humans, and clinical manifestations of MeHg intoxication in adults include cerebellar ataxia, concentric constriction of visual fields, and sensory and auditory disturbances (Hunter and Russell, 1954;Matsumoto et al., 1965;Takeuchi, 1982;Takeuchi and Eto, 1999;Eto, 2000). Additionally, specific symptoms depend on the location of the lesions, which can often damage the cerebellum and the occipital lobe (Hunter and Russell, 1954;Matsumoto et al., 1965;Takeuchi, 1982;Takeuchi and Eto, 1999;Eto, 2000).
Although a few reports have described radiologic imaging findings in patients with MeHg poisoning (Korogi et al., 1994;Matsumoto et al., 1988), these studies have provided no objective criteria for brain atrophy and have been subjective assessments. No exploratory whole-brain studies with objective criteria in patients with MeHg poisoning have been performed to date. Therefore, using MR imaging data from survivors of Minamata disease, in this study, we aimed to determine whether voxel-based morphometry (VBM) analysis can identify regional gray and white matter differences in the brain of patients with chronic MeHg poisoning.

Study subjects and patient group classification
All participants provided informed consent at the National Institute for Minamata Disease in Minamata, and written informed consent was obtained, either from them or their caregivers. This study was approved by the Ethics Committee for Epidemiological and General Research at the Faculty of Life Science, Kumamoto University (approval number: 2177).
Data from 37 survivors of Minamata disease and 141 healthy controls who had undergone 3 T MR imaging at the National Institute for Minamata Disease from October 2014 to April 2021 were retrieved from PACS storage and used in this study. All patients were officially diagnosed and certified based on the Diagnostic Guidelines for Minamata Disease, issued by the Environment Agency, Japan (Director General of Environmental Health Department, 1986), which were derived based on epidemiological and neurological evidence. Epidemiological evidence included (a) presence of MeHg in hair, blood, urine, or umbilical cord; (b) history of ingestion of fish and shellfish contaminated with MeHg; (c) history of residence in Minamata, occupation, and health status of family members; and (d) time of onset and disease course. Neurological evidence included any one of the combinations of signs: (a) sensory disturbance and ataxia; (b) sensory disturbance, suspected ataxia, and dysequilibrium or bilateral concentric constriction of the visual field; (c) sensory disturbance, bilateral concentric constriction of the visual field, and other ophthalmological or otological signs indicating central nervous system dysfunction; and (d) sensory disturbance, suspected ataxia, and co-existence of other signs, which, collectively, can be judged to be the effects of MeHg. Patients suspected of placental MeHg exposure due to maternal Minamata were included if (a) more than 50 ppm total mercury was estimated to be present in the hair of the mother during pregnancy or the mother had been diagnosed with MeHg poisoning or (b) more than 1 ppm MeHg was present in the conserved umbilical cord. The exclusion criteria were as follows: (a) born in or after 1969, (b) considerable intellectual disability but without ataxia, (c) progressive symptoms, (d) unilateral symptoms, (e) multiple malformations, and (f) all signs and symptoms that can be explained by other causes. Official records such as patient age and neurological findings at the time of MeHg poisoning recognition were retrieved from Ministry of the Environment related facilities. As neuropathology due to MeHg poisoning can depend on the timing of MeHg exposure (Matsumoto et al., 1965;Takeuchi, 1982;Takeuchi and Eto, 1999;Eto, 2000), patients included in this study were divided into three categories based on the above guidelines and whether exposure occurred in utero, under 15 years of age, or over 15 years of age, as fetal-, pediatric-, and adult-type patients, respectively.
Patients were excluded if MR images had artifacts or structural abnormalities that could interfere with VBM analysis (e.g., brain infarction, Fazekas grade 2 or higher white matter lesions (Fazekas et al., 1987), brain atrophy, or deformation due to other pathologies, among others). T2-weighted and three-dimensional (3D) spoiled gradient recalled echo (SPGR) images were consensually inspected by two experienced neuroradiologists. We excluded seven patients due to (i) severe motion artifacts that would make VBM analysis difficult (n = 2), (ii) old brain infarction and Fazekas grade 2 or higher white matter lesions on MR images (n = 2), (iii) idiopathic normal pressure hydrocephalus-like MR imaging features (n = 1), (iv) cortical dysplasia (n = 1), and (v) Chiari I malformation (n = 1). Consequently, clinical and MR imaging data of 30 patients, corresponding to 10 patients each of the fetal, the pediatric, and the adult types, were analyzed in this study ( Table 1). Activities of daily living at the time of MR imaging were evaluated using the Nishimura's Activities of Daily Living (N-ADL) scale in all patients (Nishimura et al., 1993).
The selection criteria for healthy controls were as follows: (i) volunteers who lived in areas other than Minamata where no outbreak of MeHg poisoning had been reported, (ii) no history of diabetes and neurological disorder, and (iii) no use of medications that can act on the central nervous system. An experienced neurologist interviewed all potential participants to screen for the presence or absence of neurological disorders, and these interviews were conducted on the same day as MR imaging. Similar to patient exclusion criteria, healthy controls were excepted if there were artifacts and structural abnormalities that could interfere with VBM analysis on MR images. We excluded 36 volunteers due to (i) the presence of old brain infarction and Fazekas grade 2 or higher white matter lesions on MR images (n = 16), (ii) severe image motion artifacts that would make VBM analysis difficult (n = 5), and (iii) a diagnosis of focal or diffuse brain atrophy (n = 4), venous malformation (n = 3), markedly dilated Virchow-Robin spaces (n = 3), cavum vergae (n = 3), or arachnoid cyst (n = 2). Consequently, 53, 37, and 15 age-and sex-matched healthy controls were assigned to the fetal-, pediatric-, and adult-type patient groups, respectively.

MR image acquisition
An identical MRI scanning protocol was employed in both patient *Age at the time of disease certification; ** neurological findings at the time of disease certification; N-ADL = Nishimura's Activities of Daily Living scale; 1 = present; 0 = absent; -= lost or unknown records or missing data from unexaminable or non-assessable tests; SD = sensory disturbance including tactile disorder, pain disorder, and Romberg sign; CS = cerebellar signs including limb ataxia, truncal ataxia, gait disturbance, and tandem gait; AD = auditory disturbance; VFI = visual field impairment. and control groups. All scans were performed on a 3 T MR unit (Discovery MR750 3.0 T, GE Healthcare) with a 12-channel head coil.
In brief, data processing in CAT12 can be separated into an initial voxel-based processing and the main voxel-based processing, wherein initial voxel-based processing begins with a spatial adaptive non-local means denoizing filter (Manjón et al., 2010), followed by internal resampling to accurately accommodate low-resolution images and anisotropic spatial resolutions. The data are then bias-corrected and affine-registered to improve outcomes of the subsequent steps, followed by the standard SPM unified segmentation (Ashburner and Friston, 2005). The outcomes of the latter step provide starting estimates for subsequent refined voxel-based processing, which then uses output from the unified segmentation and proceeds with skull-stripping of the brain. The brain is then parcellated into the left and right hemispheres, subcortical areas, and the cerebellum. Next, local white matter hyperintensities are detected but are accounted for later during the spatial normalization. Subsequently, a local intensity transformation of all tissue classes is performed, which is particularly helpful in terms of reducing the effects of higher gray matter intensities in the motor cortex, the basal ganglia, or the occipital lobe before the final adaptive maximum a posteriori (AMAP) segmentation. This final AMAP segmentation step (Rajapakse et al., 1997), which does not rely on a priori information of tissue probabilities, is then refined by applying a partial volume estimation (Tohka et al., 2004), which can effectively estimate the fractional content for each tissue type per voxel. In the last default step, the tissue segments are spatially normalized to a common reference space using optimized geodesic shooting registrations (Ashburner and Friston, 2011). Furthermore, voxel values in the tissue maps are modulated by the Jacobian determinant calculated during spatial normalization. Total intracranial volume (TIV) was calculated for each participant within MATLAB from gray matter (GM), white matter (WM), and cerebrospinal fluid (CSF) tissue components. Finally, all modulated and normalized GM and WM segments were smoothed with full width at half-maximum isotropic Gaussian kernel of 8 mm.

Statistical analyses
For whole-brain voxel-wise analysis, GM and WM volumes between all patients and all controls, and between subgroup of MeHg patients and their corresponding controls, were compared using the two-sample t-test at an initial cluster forming voxel threshold of p < 0.001. All results were family-wise error (FWE)-corrected for multiple comparisons at p < 0.05 using the Gaussian random field theory on the basis of cluster extent. Participant age, gender, and TIV were added as covariates in our model.
Using SPSS Statistics version 24.0 for Windows (IBM, Armonk, NY; spss.com), differences in terms of gender ratio and mean age between each group of MeHg patients and their corresponding controls, and between all patients and all controls, were assessed using Fisher's exact test and the Mann-Whitney U test, respectively. To compare mean N-ADL scores among the three patient groups, we used one-way analysis of variance and the Scheffe test. A p-value of < 0.05 was considered significant.

Whole-brain VBM analysis
The results of whole-brain VBM GM analysis in patients with MeHg Abbreviations: FWE = family-wise error correction; k = voxel size; MNI = Montreal Neurological Institute.
poisoning, compared to healthy volunteers, are shown in Table 2. A comparison of all patients with all controls revealed a significant reduction in GM volume in the left thalamus (p < 0.001 FWE corrected), the right cerebellum (p < 0.001 FWE corrected), and the right calcarine (p < 0.001 FWE corrected; Fig. 1) in all patients. Among fetal-type patients, a significant reduction in GM volume in the left thalamus was noted (p < 0.001 FWE corrected; Fig. 2). In pediatric-type patients, a significant reduction in GM volume was seen in the right calcarine (p = 0.001 FWE corrected) and the left cerebellum (p < 0.001 FWE corrected; Fig. 3), while in adult-type patients, volume reduction was evident in the right calcarine (p = 0.003 FWE corrected; Fig. 4) and the right cerebellum (p = 0.038 FWE corrected; Fig. 4). In the patient groups, there were no areas with a significant increase in GM volume. The findings from whole-brain VBM WM analysis of patients with MeHg poisoning and corresponding healthy volunteers are listed in Table 2. Compared with all the controls, significant WM volume reduction was found in the corpus callosum (p < 0.001 FWE corrected), the right cerebral peduncle (p < 0.001 FWE corrected), and the left temporal WM (p = 0.007 FWE corrected) in all patients (Fig. 5). Among fetal-type patients, a significant reduction in WM volume was found in the midbrain (p < 0.001 FWE corrected; Fig. 6). In pediatric-type patients, significant WM volume reduction was observed in the left calcarine WM (p < 0.001 FWE corrected), left and right frontal WM (p = 0.001 and p < 0.001 FWE corrected, respectively), left and right cerebellar WM (p < 0.001 FWE corrected), left temporal WM (p = 0.005 FWE corrected), and right frontal WM (p < 0.001 FWE corrected) (Fig. 7). No significant differences in WM volume were seen between adult-type patients and their controls. In the patient groups, there were no areas with a significant increase in WM volume.

Discussion
The results of our exploratory whole-brain VBM study in patients with chronic MeHg poisoning and controls imply that volume reduction occurs in the thalamus and cerebral/cerebellar WM as well as the calcarine and cerebellum. In addition, specific patterns of morphological changes were seen in the brain depending on when exposure occurred, i. e., GM volume reduction in the calcarine and cerebellum for pediatricand adult-type patients but in the thalamus for fetal-type patients. Although atrophy in the calcarine and cerebellum on MRI in patients with chronic MeHg poisoning has been reported (Korogi et al., 1994), this study provided no objective criteria for brain atrophy and was subjective assessments. To date, no exploratory whole-brain studies with objective criteria have been conducted in patients with MeHg poisoning. To the best of our knowledge, this is the first report describing the results

Fig. 2.
Gray matter (GM) volume differences between the fetal-type patients and its corresponding controls, outlined over T1 template image (uncorrected p < 0.001) (A) The left thalamus shows a reduction in GM volume in fetal-type patients compared to the corresponding control group (x = − 10, y = − 26, z = 18; k = 2060; FWE corrected p = 0.000); Bar represents T values. (B) Regional GM volume reduction in the whole brain is shown. Fig. 3. Gray matter (GM) volume differences between pediatric-type patients and corresponding controls, outlined over a T1 template image (uncorrected p < 0.001) (A) Compared to the corresponding controls, a decrease in GM volume in pediatric-type patients was noted in the right calcarine (x = 12, y = -69, z = 15; k = 1737; FWE corrected p = 0.001) and in the left cerebellum (x = -28, y = -66, z = -51; k = 18174; FWE corrected p = 0.000). Bar represents T values. (B) Regional GM volume reduction in the whole brain is shown. Fig. 4. Gray matter (GM) volume differences between adult-type patients and corresponding controls, outlined over T1 template image (uncorrected p < 0.001) (A) Compared to the controls, a decrease in GM volume in adult-type patients was seen in the right calcarine (x = 10, y = -69, z = 14; k = 1232; FWE corrected p = 0.003) and in the right cerebellum (x = 9, y = -82, z = -32; k = 695; FWE corrected p = 0.038). Bar represents T values. (B) Regional GM volume reduction in the whole brain is shown.
of an exploratory whole-brain study using VBM analysis in patients with chronic MeHg poisoning.
In all patients and fetal-type patients, a significant, extensive reduction in GM volume in the thalamus was observed (Figs. 1 and 2, respectively). Although thalamic atrophy has not been fully reported previously in patients with MeHg poisoning, a study on transplacental MeHg neurotoxicity in the fetal rat brain consistently demonstrated neuronal degeneration of varying degrees in the thalamus (Kakita et al., 2000). All sensory information, except olfaction, reaches the primary sensory cortex via a relay in one of the nuclei of the sensory thalamus, and the primary cortical area projects back to the same and other thalamic nuclei (Peyrache et al., 2019). As sensory cortices are predominantly involved in patients with MeHg poisoning (Hunter and Russell, 1954;Matsumoto et al., 1965;Takeuchi, 1982;Takeuchi and Eto, 1999;Eto, 2000), the thalamus might have been affected by secondary degeneration from the sensory cortices.
We also demonstrate significant volume reduction in the frontal, temporal, calcarine, cerebellar WM, and the corpus callosum (Fig. 5). A previous study has reported that, in the cerebral cortices showing a macroscopic spongy appearance to a thinning-out decrease of neurons, secondary degeneration of the pyramidal tracts, internal sagittal stratum, and central parts of the cerebral WM were also often observed (Takeuchi and Eto, 1999); notably, our VBM assessment results support these WM findings.
Previous pathological studies on chronic cases of MeHg poisoning (Hunter and Russell, 1954;Matsumoto et al., 1965;Takeuchi, 1982;Takeuchi and Eto, 1999;Eto, 2000) have documented different changes in the brain between fetal-and postnatal-type cases with a predilection toward the calcarine and the cerebellum in postnatal-type cases. Specifically, while changes in the calcarine cortex range from a macroscopic spongy appearance to a thinning-out decrease in neurons in severely affected cases (Takeuchi and Eto, 1999), the cerebellum shows central atrophy involving the vermis and its surrounding regions in chronic cases with characteristic loss of granule cells and the presence of relatively well-preserved Purkinje cells (Takeuchi and Eto, 1999). In contrast, changes in the fetal-type cases are less often localized to a specific region (Matsumoto et al., 1965;Takeuchi, 1982;Takeuchi and Eto, 1999;Eto, 2000). Moreover, in line with these findings, our VBM results also demonstrate that brain structural changes may depend on the timing of MeHg exposure.
Pathological changes in the cerebral cortices are most severe in the calcarine cortex of patients with MeHg poisoning, with less severe alterations in the precentral, postcentral, and temporal transverse cortices (Eto, 2000;Takeuchi, 1982;Takeuchi and Eto, 1999). Congruently, VBM might not have revealed a significant reduction in GM in the precentral, postcentral, and temporal transverse cortices, probably because these regions were less severely involved compared to the calcarine cortex. However, WM volume loss, which is thought to be secondary degeneration of the pyramidal tracts, was observed (Figs. 5 and 7), suggesting changes in the precentral and postcentral cortices.
This study has some limitations. First, the population in each patient group was small. Decades have passed since the large-scale MeHg disaster in Minamata in 1956, and the number of surviving patients is decreasing. Although we included as many patients as possible, our Fig. 6. White matter (WM) volume differences between fetal-type patients and the corresponding controls, outlined over T1 template image (uncorrected p < 0.001) Regional WM volume reduction in the whole brain is shown. Compared to the corresponding controls, a decrease in WM volume in pediatric-type patients was seen in the midbrain (x = 0, y = -15, z = -6, k = 2482, FWE corrected p = 0.000) extending to the pons. Bar represents T values.
inclusion and exclusion criteria limited the study population. Second, we did not evaluate lifetime history of exposure to other chemicals based on occupational and environmental exposure histories. Some of these subjects may have had different lifetime histories due to occupations. Additionally, asymmetry in GM volume loss was observed in some brain structures. We do not exactly know why this occurred following a generalized systemic exposure. Handedness has been reported to affect the thickness of the cerebral cortex, resulting in structural asymmetries of the cerebral cortex (Sha et al. 2021). Although we did not evaluate their handedness, the factor might have affected the asymmetry in GM volume loss.
To summarize, this exploratory whole-brain VBM study documents a significant volume reduction in the thalamus, cerebellum, calcarine and cerebral/cerebellar WM in patients with chronic MeHg poisoning. A significant reduction in GM volume was observed in the calcarine area and in the cerebellum in pediatric-and adult-type patients and in the thalamus among fetal-type patients. A significant reduction in WM volume was seen in the cerebral and the cerebellar regions, especially in pediatric-type patients. Our results provide helpful insight into structural changes in the brain depending on the timing of MeHg exposure. We posit that these results can contribute to the objective diagnosis using MRI in patients with chronic MeHg poisoning.

Funding/Support
This work was supported by the Study of the Health Effects of Heavy Metals organized by the Ministry of the Environment, Japan.

Data Sharing Statement
All data that support the findings of this study are available from the study team upon reasonable request.