The evaluation of Hashimoto's thyroiditis with event-related potentials and magnetic resonance spectroscopy and its relation to cognitive function

Thyroid dysfunction is very often accompanied by cognitive and affective disorders. The frequency of these disorders in patients with compensated Hashimoto’s thyroiditis (HT) is unknown. The aim of the present study was to evaluate brain dysfunction in euthyroid HT patients by means of event-related potentials (ERP) and magnetic resonance spectroscopy (MRS) and to correlate it with cognitive function. 68 patients with HT (59 female, 9 male) and 45 healthy controls were included in the study. All the patients underwent ERP including an analysis of N200 and P300 response parameters. MRS voxels were located in the posterior cingulate gyrus (PCG) and the left parietal white matter (PWM). The NAA/Cr, mI/Cr, and Cho/Cr ratios were analysed. The ERP parameters, MRS metabolite ratios and hormonal concentrations (TSH, fT3, fT4) as well as TGAb and TPOAb titer were also correlated. There was a significant prolongation of the latencies of N200 and P300 potentials and a significant decrease of P300 amplitude in HT patients than in the control group. There was a significant positive correlation between the mI/Cr ratio in the PCG area and P300 latencies. NAA/Cr ratio in the PCG region showed significant negative correlations with all N200 latencies. The results may suggest brain dysfunction in neurologically asymptomatic HT patients. ERPs undergo significant changes in patients with HT and may, in combination with MRS, constitute an important element in the recognition and monitoring of cognitive functions in this group of patients.


Material and methods
The study comprised 68 patients with HT (59 women and 9 men, aged 20-63 years, mean 44.3) who met the criteria for the diagnosis of HT 1 . All the patients underwent ultrasound examination of the thyroid gland, were in the euthyreosis phase and treated with levothyroxine.
Patients with the presence of any neurological, psychiatric, cognitive, and autoimmune diseases, history of traumatic brain injury, underweight (BMI < 18.5 kg/m 2 ) or obesity with a BMI of ≥ 35 kg/m 2 , defects in hearing, and administration of medications which change brain bioelectrical activity (e.g. neuroleptics, steroids, antiepileptic) were excluded. The control group (CG) consisted of 45 age and sex-matched healthy volunteers (39 women and 6 men, aged 22-64 years, mean 44.1). The exclusion criteria and the study protocol were the same in the CG.
The study protocol included a detailed neurological examination, with assessment of mental state using the Montreal Cognitive Assessment test (MoCA), Trail Making Test (TMT-A), Symbol Digit Modalities Test (SDMD) and Clock-Drawing Test (CDT) to screen for cognitive impairment, ERP and MRI examination. Laboratory tests included serum concentration of thyrotropin (TSH), free tri-iodothyronine (fT3), free thyroxine (fT4), and serum auto-antibodies against thyroid peroxidase (TPOAb) and thyroglobulin (TGAb) titers.
Approval for this research was given by the Commission of Ethics at the Wrocław Medical University (number of permission: KB-313/2013). Informed consent was obtained from each patient to participate in this study. Event related potentials protocol. Brain bioelectrical activity was recorded using superficial Ag/AgCl electrodes placed in the frontal region (Fz), central region (Cz) and parietal region (Pz), according to the international 10-20 system, with reference to linked earlobes and with a forearm ground. Impedance of all electrodes was maintained below 5 kΩ. ERP were elicited using the classic auditory "oddball paradigm". The target tones (high frequency: 2000 Hz) occurred 20% of the time in each trial and non-target tones (low frequency: 1000 Hz) 80% of the time. Auditory stimuli were tones of 70 dB intensity and 200 ms duration. The patients were sitting in a comfortable position in a semi-darkened room and were asked to mentally count the stimuli. At least 30 target trials were averaged in each run and two runs were performed with each patient.
The responses were evaluated by means of a Nicolet 1000 Viking, with a 0.30/s, 70 Hz bandpass filter, sweep time 1000 ms and pre-stimulus baseline 250 ms. N200 was identified as a negative component with a latency of 180-320 ms and P300 was identified as a positive component with a latency of 280-450 ms after start of the stimulus. The latencies and amplitudes of N200 and P300 waves ("peak to peak") were analysed. The procedure was compliant with the International Federation of Clinical Neurophysiology (IFCN) recommended standards 11,15 . We used the same methods as in our previous paper 18 .
MR imaging. Imaging was performed with a 1.5 T SignaHdx MR scanner (GE Healthcare) using a head and neck 16-channel coil. The conventional MR sequences included: axial T1-weighted, axial, sagittal, and coronal T2-weighted images, FLAIR (fluid-attenuated inversion recovery sequence) and diffusion-weighted imaging (DWI) images. The protocol of magnetic resonance spectroscopy (MRS) examination used in our study was exactly the same as in previously published papers 21 . Post-processing of the images was performed using updated ReadyView software (GE Healthcare, ADW 4.6).
Magnetic resonance spectroscopy protocol. The MRS examinations were conducted using the Single Voxel Spectroscopy (SVS) method (PRESS sequence) with the following acquisition parameters: TE = 35 ms, TR = 1500 ms, 128 acquisitions, number of excitations = 8. Using the localising axial T2-weighted images, voxels of 2 × 2 × 2 cm (8 cm 3 ) size were placed in the normal appearing brain in 2 regions: the posterior cingulate gyrus www.nature.com/scientificreports/ cortex-PCG ( Fig. 1) and the left parietal white matter-PWM (Fig. 2). The total acquisition time was 3 min 45 s for each voxel. MR spectroscopy studies were conducted by using the automated single-voxel MR spectroscopy package Proton Brain Examination/Single Voxel (PROBE/SV; GE Medical Systems, Milwaukee, WI). The preimaging algorithm of the PROBE software automatically adjusted the transmitter and receiver gains and centre frequency. The local magnetic field homogeneity was optimised with the three-plane auto-shim procedure with linear gradient shimming, and the flip angle of the third water-suppression pulse was adjusted for chemical shift water suppression (CHESS) before PRESS acquisition. Each spectrum was automatically fitted to four peaks corresponding to the levels of N-acetylaspartate (NAA) (2.02 ppm), total creatine (Cr) (3.03 ppm), cholinecontaining compounds (Cho) (3.23 ppm) and myo-inositol (mI) (3.56 ppm). Metabolite intensity ratios of NAA/ Cr, Cho/Cr, and mI/Cr were automatically calculated at the end of each PROBE/SV acquisition. MRS data were post-processed by means of software provided by the manufacturer (GE workstation, ADW 4.4). The ratios of NAA, Cho, and mI to creatine (NAA/Cr, Cho/Cr, mI/Cr, respectively) were assessed. We used the same methods as in our previous paper 21,22 . Statistical analysis. Statistical computations were performed using the Statistica PL software package version 13.0, and p < 0.05 was set as a significant level.
The event-related potentials (ERP) values in patients with HT and healthy subjects were compared using the Student T-test. Additionally, in order to assess sensitivity, specificity, and accuracy of ERP in distinguishing   Informed consent. Informed consent was obtained from all individual participants included in the study.
The neurological examination and the results of the MoCA, TMT-A, SDMT and CDT were normal in all patients (100%).
Event-related potentials. The mean values for the latency of N200 and P300 potentials recorded from all electrodes (Fz, Cz, Pz) were significantly longer in HT patients than in the control group (p < 0.001). The mean amplitude of P300 potentials recorded from all electrodes was significantly lower in HT patients than in the control ( Table 2).

Event-related potentials in HT patients and control group according to age.
In order to compare HT patients to control group according to their age, HT patients and CG were divided into three groups as follows: 20-39 years, 40-59 years and over 60 years old.
HT patients showed a statistically significant longer N200 and P300 latencies and lower P300 amplitudes from all electrodes in all aged groups compared to the control group (Table 3).

Event-related potentials in HT patients depending on the length of the disease. In order to
compare HT patients to the control group according to the duration of the disease, HT patients were divided into four groups as follows: duration of the disease up to 10 years (19 subjects), duration of the disease 11-20 years (20 subjects), duration of the disease 21-30 years (16 subjects), duration of the disease over 30 years (13 subjects).
HT patients showed a statistically significant longer N200 and P300 latencies and lower P300 amplitudes from all electrodes in all groups compared to the control group (Table 4).
ROC analysis for event-related potentials. ROC analysis was performed separately for the N200 and P300 ERP components (Table 5  www.nature.com/scientificreports/ c). ROC analysis for the value of the amplitude showed good accuracy with the area under the curve above 0.70 only for P300 components (Fig. 4d-f).

Correlations of EPR and thyroid hormones.
There were no statistically significant correlations between the mean ERP parameters, thyroid hormones levels, and TGAb and TPOAb either.

Correlations of MRS measurements and thyroid hormones.
The MRS parameters showed no statistically significant correlations neither with thyroid hormones levels nor TGAb and TPOAb . www.nature.com/scientificreports/

Discussion
The clinical manifestations of autoimmune thyroiditis are not specific and are often low. However, many patients report numerous subjective complaints that have a significant impact on their quality of life 5,7  Up to now, ERP and MRS studies have not been conducted in large groups of patients with hormonally compensated Hashimoto's thyroiditis. It is worth mentioning that Bladowska et al. assessed metabolic alterations within the normal appearing brain in 55 subjects with HT using MRS and they also correlated MRS measurements with hormonal concentrations 22 . They found significantly decreased NAA/Cr ratios in PCG and parietal white matter (PWM) in HT patients. Their results also showed significant positive correlations between the NAA/ Cr ratio and fT3 level. Since the reduction of NAA/Cr ratios suggests a decrease of neuronal activity, the study by Bladowska et al. indicated the early cerebral metabolic disturbances associated with Hashimoto's thyroiditis and thus suggesting that MRS could be a sensitive marker of cognitive decline in presymptomatic subjects 22 . A few, isolated studies describe cases of a rare complication of Hashimoto's encephalopathy 25 . A report describing the results of MRS in a 52-year-old woman with HE was published by Chinese authors. In the patient's brain they found a reduced level of NAA and mI, in addition-diffuse, atypical inflammatory changes 26 . In a SPECT study, features of diffuse hypoperfusion were described in patients with HE 27 .
We have found significantly longer latencies in the N200 and P300 components in HT patients, and reduced P300 wave amplitudes in all leads. During the ERP and brain imaging examinations, all patients remained in euthyreosis and none of them had signs of CNS damage in the neurological examination. People with conditions that could affect brain bioelectrical activity and/or damage the nervous system were excluded from the analysis.
Changes in ERP parameters are a measure of the depth of cognitive impairment, however, their bundles with the severity and extent of changes in the CNS remain ambiguous 14 . Despite specific CNS structures generating individual ERP components such as prefrontal, temporo-parietal cortex and hippocampal region, their result is considered to reflect the overall functioning of extensive neural networks 14,16 . Analysing the correlations of various neuropsychological tests with ERP, no specific tests were identified that would best correlate with ERP results. The results of various authors are ambiguous and most often relate to tests used to assess global cognitive functioning 28,29 .
Interestingly, we found significant correlations of ERP and MRS measurements. NAA/Cr ratio in the PCG region showed significant negative correlations with all N200 latencies. NAA is a neuronal marker, thus its concentration correlates with neuronal density and neuronal function. As mentioned above, the decrease of NAA/ Cr ratio may indicate a reduction of neuronal activity. We observed that the lower NAA/Cr ratios in the PCG region, the longer are the values for the latency of N200 potentials recorded from all electrodes. Furthermore, it should be stressed that the PCG is a very special brain region involved in cognitive function. The PCG is a part of the memory-related default mode network (DMN), an important system of interacting brain regions associated with Alzheimer's Disease (AD). Disorders of the DMN are considered to be an important feature of AD. Moreover, the PCG cortex plays important roles in episodic memory, spatial attention, self-evaluation, and other cognitive functions 30 . Studies have been suggested that abnormal outgoing connections from the PCG cortex to other brain areas may be an imaging biomarker for the early stages of cognitive impairment, especially AD. The most important goal is to perform medical interventions as early as possible. It has been reported that early appropriate management of risk patients can not only effectively delay the progression of the disease and extend survival, but also improve the quality of life of the affected subjects and reduce the burden on the whole of society 30 .
According to these findings our results suggest that both the decrease of NAA/Cr ratio and the longer values for the latency of N200 potentials may indicate the functional deterioration in the PCG cortex.
Furthermore, there was also a significant positive correlation between the mI/Cr ratio in the PCG area and P300 latencies. It is known that mI is an in vivo biomarker of glial cell inflammation and activation. Increased mI levels are thought to be connected with CNS inflammation and gliosis, which may be responsible for neuronal dysfunction 21,22 . It should be emphasised that we did not observe any significant correlations between ERP and metabolite ratios in the white matter of the PWM region, which means that functional impairment may begin in the cortex. Thus early functional disorders can be evaluated with assessing bioelectrical activity of cortical areas of the brain using the cognitive event-related potentials.
Our findings showing the significant relationship between the increased mI/Cr ratio in the PCG cortex and the longer values for the latency of P300 potentials are in accordance with our other results described above, suggesting that MRS and ERP may serve as a useful tool in indicating the risk group of patients prone to developing cognitive impairment and thus to be taken care of as soon as possible.
Although studies have shown an increased risk of cognitive impairment in HT patients, even in the euthyroid state, the mechanisms behind it remain unclear 3,8 . Long-term potentiation in the Schaffer collateral-CA1 pathway of the hippocampus may to play a key role in the development of these disorders. In a study on an experimental model of euthyroid HT mice, abnormalities in synaptic plasticity of the Schaffer collateral-CA1 synapses in the www.nature.com/scientificreports/ hippocampus, worse memory and spatial learning were shown. In transmission electron microscopy the synapses and astrocytes in the hippocampus was damaged. The elements in glutamate-glutamine circulation located in astrocytes, were downregulated and elevated levels of glutamate in the hippocampus was observed 30,31 .
Other studies suggested that inflammation characterised by elevated cytokines plays a key role in the pathogenesis of cognitive impairment in HT 7,24 . Previous studies have shown increased production of cytokines, including interleukin-1β, interleukin-6, tumour necrosis factor alpha and monocyte chemotactic protein-1 in patients with HT. Vitamin D participates in modulating the secretion of cytokines and these results suggest that vitamin D deficiency may play a key role in the cognitive impairment of HT patients 7 . Similar conclusions can be drawn from the study by Cai et al., who demonstrated astrocyte and microglia activation with increased expression of proinflammatory cytokine genes in the frontal cortex on experimental models of an HT mouse, which influenced the serotonin system in this region of the brain 32 . No significant direct effect of the anti-thyroid antibodies themselves on the structures of the central nervous system has been demonstrated, although some researchers have suggested that they may weaken the integrity of the myelin sheath and affect many neurotransmitters 3 .
As mentioned above, we observed significant metabolic changes in the PCG region as well as significant correlations between MRS parameters and ERP. Our results indicate that there are alterations in the important brain region involved in the cognition process, thus obviously leading to cognitive impairment in HT patients. However, it should be stressed that the advanced MR techniques, although able to show even subtle preclinical changes, do not allow us to speculate as to which mechanism could be responsible for brain injury and finally cognitive impairment, as their results are non-specific. By using the MRS method it is possible to assess the metabolic impairment in specific brain regions such as the PCG region, which has already been proven to be associated with cognition, but these alterations may not presume the underlying mechanism 22,33 .
The limitations include conducting the study in only one time point in HT patients, which is known to have a fluctuating hormonal course. However, our previous report was the first pilot study in the world literature whose findings might encourage further research. Our future studies will focus on monitoring ERP and MRS parameters in the course of the disease and analysing the relationship between cognitive impairment and their predictive cognitive outcome of HT patients.

Conclusions
To our knowledge, this is the first study examining ERP and MRS findings in euthyroid patients with HT without cognitive impairment.
We found significant correlations of ERP and MRS measurements in the PCG region. The NAA/Cr ratio showed significant negative correlations with all N200 latencies, while the mI/Cr ratio revealed a significant positive correlation with P300 latencies, which means that decreased values of the N200 and P300 latencies are associated with metabolic alterations in the PCG cortex. Both findings may indicate important functional changes in the cortex of the PCG which can be assessed with ERP as well as MRS measurements. Therefore functional disorders observed in the PCG cortex may suggest some brain dysfunction in neurologically asymptomatic HT patients. The mandatory goal of everyday clinical practice is to assess the group of HT subjects who are at risk of cognitive disorders and accordingly they need early and dedicated medical care in order to achieve cerebral function recovery or at least to stop progression of brain damage in the course of HT.
The lack of significant differences between the ERP and MRS parameters in HT patients with different initial diagnosis may indicate the influence of the autoimmune process itself, and not of the severity of pre-therapy thyroid dysfunction, on the development of cognitive impairment.
ERPs undergo significant changes in patients with HT and may, apart from neuroimaging studies, constitute an important element of recognising and monitoring cognitive functions in this group of patients. The significant correlations we observed with the MRS parameters justify further research in this area. www.nature.com/scientificreports/