Altered mismatch response of inferior parietal lobule in amnestic mild cognitive impairment: A magnetoencephalographic study

Abstract Background Mismatch negativity (MMN) reflects the functional integrity of sensory memory function. With the advantages of independence of individual's focused attention and behavioral cooperation, this neurophysiological signal is particularly suitable for investigating elderly with cognitive decline such as amnestic mild cognitive impairment (aMCI). However, the existing results remain substantially inconsistent whether these patients show deficits of MMN. In order to reconcile the previous disputes, the present study used magnetoencephalography combined with distributed source imaging methods to determine the source‐level magnetic mismatch negativity (MMNm) in aMCI. Methods A total of 26 healthy controls (HC) and 26 patients with aMCI underwent an auditory oddball paradigm during the MEG recordings. MMNm amplitudes and latencies in the bilateral superior temporal gyrus, inferior frontal gyrus, and inferior parietal lobule (IPL) were compared between HC and aMCI groups. The correlations of MMNm responses with performance of auditory/verbal memory tests were examined. Finally, MMNm and its combination with verbal/auditory memory tests were submitted to receiver operating characteristic (ROC) curve analysis. Results Compared to HC, patients with aMCI showed significantly delayed MMNm latencies in the IPL. Among the patients with aMCI, longer MMNm latencies of left IPL were associated with lower scores of Chinese Version Verbal Learning Test (CVVLT). The ROC curve analysis revealed that the combination of MMNm latencies of left IPL and CVVLT scores yielded a moderate accuracy in the discrimination of aMCI from HC at an individual level. Conclusions Our data suggest dysfunctional MMNm in patients with aMCI, particularly in the IPL.


| INTRODUC TI ON
Mild cognitive impairment (MCI) has been proposed to be an intermediate stage of cognitive dysfunction between normal aging and dementia. 1 Unlike patients with dementia, those with MCI still maintain their independence in most of activities of daily living (ADL) with minimal aids or assistance; however, cognitive deficits are noticeable and detected by neuropsychological assessments in these patients. [2][3][4] The concept of MCI is further categorized as amnestic MCI (aMCI) and non-amnestic MCI (naMCI). 2,5 Clinically, aMCI is characterized as apparent deficits in memory rather than other cognitive functions and is shown to be strongly associated with the development of Alzheimer's disease (AD), the most common type of dementia. 6,7 Since AD would cause severe ADL dysfunction which in turn increases the health care and economic burdens, the early and accurate diagnosis of aMCI prior to AD is imperative.
Up to the date, the diagnosis of aMCI is based on the clinical criteria. [5][6][7] Although some biomarkers, such as beta amyloid or tau levels of cerebrospinal fluid (CSF) and positron emission tomography (PET) imaging with C-labeled Pittsburgh compound-B, have been proposed for their utility in the diagnosis of aMCI due to AD, the existing data do not provide strong evidence for the routine use of these biomarkers in clinical practice. [8][9][10] Furthermore, with the fact that PET imaging is expensive and CSF is not easy to collect from each patient with aMCI, the search for other markers is needed.
Neurophysiological recordings using electroencephalography (EEG) or magnetoencephalography (MEG) have been demonstrated as promising tools in the AD research [11][12][13] and can also be considered suitable for the studies of aMCI. When applying neurophysiological recordings in the studies of neurodegenerative diseases, a task or a signal that is largely independent of the individual's focused attention, motivation, or behavioral cooperation is a key element for its clinical utility and future application. Mismatch negativity (MMN), or its magnetic counterpart (MMNm), is one of the neurophysiological signals that possess these advantages. [14][15][16][17] MMN/MMNm is an automatic cortical activity elicited by a passive oddball task, in which a sequence of identical auditory stimuli (standards) is occasionally interrupted by deviant sounds (deviants) differing in any of perceptual characteristics, such as pitch, duration, intensity, or location. [16][17][18][19] The generation of the MMN/ MMNm has been interpreted as a pre-attentive cognitive process indexing functional integrity of sensory memory and accuracy in detecting changes. 20 30,31 The second one is related to the different task instructions among the studies. MMN is conventionally obtained through a passive oddball paradigm, while Tsolaki et al. used an active oddball task in which the subjects were asked to respond to infrequent targets. 31 Finally, all the existing studies investigated MMN by means of EEG, whose signals are potentially distorted by different tissues (e.g., brain, CSF, and skull). Also, due to the limited spatial resolution, the brain activities recorded from the scalp electrodes provide little information of the source activation of MMN. Therefore, in order to address these caveats, the present study Results: Compared to HC, patients with aMCI showed significantly delayed MMNm latencies in the IPL. Among the patients with aMCI, longer MMNm latencies of left IPL were associated with lower scores of Chinese Version Verbal Learning Test (CVVLT).
The ROC curve analysis revealed that the combination of MMNm latencies of left IPL and CVVLT scores yielded a moderate accuracy in the discrimination of aMCI from HC at an individual level.

Conclusions:
Our data suggest dysfunctional MMNm in patients with aMCI, particularly in the IPL.

K E Y W O R D S
mismatch negativity (MMN), amnestic mild cognitive impairment (aMCI), inferior parietal lobule (IPL), magnetoencephalography (MEG) aimed to use a whole-head MEG, which has a spatial resolution with a millimeter scale and a temporal resolution with a millisecond scale, to examine the spatiotemporal dynamics of MMNm activities in patients with aMCI.
The specific goals of this study were three-fold. Firstly, based on the selected regions of interest (ROIs), we attempted to determine whether, at a group level, MMNm amplitudes and latencies at the cortical level would be reduced in the patients with aMCI versus HC group. Secondly, we sought to examine whether the MMNm in the ROIs, which exhibited significant between-group differences (if detected), would show significant associations with cognitive performance related to auditory/verbal memory tests, including Chinese  ≧24 and without dementia. 33 A total of 26 community-dwelling older adults (9 males, mean age = 65.81 ± 1.45 years) were recruited as the HC group. All the subjects had no history of major psychiatric disorders, alcoholism, epilepsy, polypharmacy, or other systematic diseases that potentially have detrimental effects on cognitive function. They also self-reported no hearing impairment, and normal or corrected-to-normal vision.

The Institutional Review Board of Taipei Veterans General
Hospital (Taipei, Taiwan) approved this research project. Written informed consent was obtained from each subject after a complete explanation of the study procedure.  (Table 1).

| Neuropsychological assessments
Based on that the MMNm reflects functional integrity of auditory sensory memory and the patients with aMCI exhibit impairments in memory function, we specifically examined the following auditory/verbal memory tests in relation to MMNm: 1. CVVLT: Nine two-character nouns were spoken to the subject to measure total (learning over 4 trials) and delayed recall scores. 34 2. LM: a brief story was spoken to the subject to measure immediate and delayed recall memory function. 35 3. DSB: a series of digits were spoken to the subject to measure the auditory working memory. 36

| MEG recordings
The subjects were presented with an auditory oddball paradigm consisting of 85% standard stimuli (1000 Hz, 70 dB, 100 ms) and 15% deviant stimuli (900 Hz, 70 dB, 100 ms) in a pseudo-random order in which two deviants were separated by at least one standard. The interstimulus interval was 1000 ms. During the whole experiment, the subjects were asked to watch a silent movie with subtitles and ignore the auditory stimuli.
The neuromagnetic responses to standards and deviants were recorded by a whole-head 306-channel MEG system (Elekta-Neuromag), consisting of 102 magnetometers and 204 gradiometer.
The online sampling rate and bandpass filter were set at 1000 Hz and 0.1-200 Hz, respectively. In addition to three anatomical landmarks and four head position indicators, further ~100 head points were uniformly digitized on the head surface with a 3D digitizer.

| MEG data analysis
The MEG raw data were pre-processed by MaxFilter to suppress external and internal interference 37,38 and by signal space projections (SSP) to correct the trials contaminated by eye and cardiac artifacts. 39,40 The artifact-free standards and deviants were separately epoched into 500 ms, including a 100-ms pre-stimulus baseline. The offline bandpass filter was set at 0.1-30 Hz. In an attempt to achieve an equivalent signal-to-noise between standards and deviants, the number of standards was randomly selected to match the number of deviants. The MMNm was measured as a difference waveform obtained by subtracting neuromagnetic responses to standards from those to deviants. 16,17,22 The time window of MMNm was defined as 150-300 ms after the stimulus onset.
An overlapping-sphere method was used to resolve the forward problem. The cortically constrained source imaging was subsequently performed by means of depth-weighted minimum norm estimate (MNE) over a set of ~15,000 dipoles distributed on the cortex. 41 The individual's MNE map was then rescaled to fit the previously defined head points, with the default parameter settings in the brainstorm software. 42 The MNE maps of MMNm from each group were averaged onto the ICBM152 brain template for further analysis. The ROI identification on the Desikan-Killiany brain template has been reported in other studies. 47,48 The MNE magnitude of each dipole (i.e., vertice) was normalized in relation to its fluctuations over 100-ms, yielding a z-score map.
The peak amplitudes (z-score) and peak latencies (ms) of MMNm were derived from each ROI.

| Statistical analysis
All the data were presented as mean ± one standard error of the mean (SEM), and p values less than 0.05 were considered as statistically significant. The numerical data were normally distributed as verified by the Kolmogorov-Smirnov tests (Z < 0.999, p > 0.271).
Between-group differences of demographic and neuropsychological data were compared using independent two-sample t-tests or chisquare tests as appropriate. Two-way mixed ANOVAs, with group (HC and aMCI) as a between-subject factor and hemisphere (left and right) as a within-subject factor, were applied to compare MMNm amplitudes and latencies in each identified ROI. Furthermore, age and gender were entered as covariate variables. Greenhouse-Geisser epsilon correction was applied when the assumption of sphericity was violated. Partial eta squared was used to estimate the effect size.
Based on the ROIs with significant between-group differences, partial correlations, with age and gender as covariates, were used to determine the associations between MMNm responses and scores of auditory/verbal memory tests (i.e., CVVLT_Total, CVVLT_Delayed, LM_Immediate, LM_Delayed, DSB) among the patients with aMCI.
The significance was further adjusted for multiple comparisons by means of Bonferroni approach.
Finally, based on the significant between-group differences of MMNm responses and significant correlational data, MMNm and its combination with verbal/auditory memory tests were submitted to receiver operating characteristic (ROC) curve analysis. For the area under the curve (AUC), an AUC between 0.5 and 0.7 was considered less accurate, an AUC between 0.7 and 0.9 was considered moderately accurate, and an AUC above 0.9 was considered very accurate. 49     and deviants (15%) might cause the non-memory-based N100/ N100m to contaminated the memory-based MMN/MMNm. [66][67][68] Over the past decades, a controlled block along with the oddball paradigm has been designed to successfully address this issue.

| DISCUSS ION
However, in terms of future clinical application, we considered that the traditional oddball paradigm would be easier to operate and the recording time would be short. Secondly, we could not entirely preclude the effects of medicine (e.g., benzodiazepines) on the MMNm responses though these patients were instructed to refrain from taking their medication 24 h prior to MEG recordings. Finally, we did not carefully control the hearing threshold from each subject. However, it has been evident that there was no age-related difference of hearing threshold at the frequency of 1000 Hz. 69 Considering that the tone frequencies we used in the present study were 1000 and 900 Hz, we reasoned that all the participants could successfully encode the auditory stimuli for the basic processing.
Furthermore, since both of the HC and aMCI groups were composed of older adults and there was no significant difference of age, thus our sample was considered homogeneous in terms of auditory acuity.
In conclusion, we found that patients with aMCI showed MMNm prolongation selectively in the IPL compared with HC, and such aberration was associated with the performance of auditory/verbal memory (i.e., CVVLT). Our results also indicated that the combination of MMNm latencies of left IPL with CVVLT could adequately discriminate patients with aMCI from HC at an individual level.

CO N FLI C T O F I NTE R E S T
The authors declare that they have no conflict of interest.

AUTH O R CO NTR I B UTI O N S
CHC and PNW conceived and designed the work and acquired the data. PYC, HYH, YPC, and CHC analyzed the data. PYC, HYH, YPC, RN, and PNW participated in the discussion and provided the comments. PYC, HYH, and CHC wrote the paper. All of the authors have read and approved the manuscript.

DATA AVA I L A B I L I T Y S TAT E M E N T
Data available on request from the authors.