Speech and Short-term Memory Functions in Dyslexia: a Combined MEG and EEG Study

1. Background Developmental dyslexia is the most prevalent learning disorder impairing reading skill even in individuals having normal intelligence and full availability of education (Lyon et al., 2003). It is highly heritable, and several candidate genes contributing to dyslexia have been identified (Scerri \& Schulte-Körne, 2010). A wide range of deficits have been associated with dyslexia, but according to the most prevalent theory its major cause is impaired phonological processing (Ramus, 2001). The MMN, the generators of which overlap with brain areas reported to have anatomical abnormalities in dyslexia, has supported this notion. Diminished MMN amplitudes have systematically been reported in dyslexics for certain speech and non-speech sound features (for a review Kujala \& Näätänen, 2001). These effects were even found in children and infants having an inherited risk for dyslexia (Leppänen et al., 2002; Lovio et al., 2010). Furthermore, a recent study in our group has shown that dyslexic children have problems in forming memory traces for words (Kimppa et al., in prep.), whereas both the auditory system of normal-reading adults (Shtyrov et al., 2010) and children (Kimppa et al., in prep.) was shown to rapidly form representations for novel words. It was recently suggested that phonological deficits in dyslexia can occur due to impairments in different steps of sound processing. According to this theory, dyslexic individuals can be divided into different subgroups. Ramus and colleagues (2013) suggested that dyslexic individuals show an impairment in phonological representations or at later processing stages as an impairment in phonological skills. The current project will utilize combined electroencephalographic (EEG) and magnetoencephalographic (MEG) recordings and neuropsychological and perceptual testing in order to determine how impaired phonological neural representations vs. phonological skills contribute to dyslexia. Defined as time-locked changes to external stimuli in EEG and magnetoencephalogram MEG, event-related potentials (ERPs) and event-related fields (ERFs), respectively, could provide an objective index of information processing in the human brain. Both EEG and MEG methods offer a high temporal resolution. The benefit of MEG is offered by a more exact localization of the activated neural sources due to a diminished effect of distortions caused by the skull and tissue than in the process of EEG source localization. It will complement and add more specific information to dyslexia ERP research additional to the previous studies that were mostly conducted with EEG. For a summary of dyslexia studies conducted with MEG, see the review from Salmelin (2007). Neural responses recorded with EEG and MEG have widely been used both in the service of basic research and clinically-oriented research. During the recent years, a cortical response called the mismatch negativity (MMN) has been in intensive use in investigating auditory perception and its deficits. MMN is an ERP component elicited by any change in some repetitive aspect of auditory stimulation, peaking at 100-200 ms from change onset and detectable both electrically (MMN) and magnetically (MMNm). It was suggested that MMN provides an index of the auditory sensory ("echoic") memory and automatic (involuntary) change-detection. It also reflects native-language specific speech-sound memory traces (Näätänen et al., 1997). MMN is elicited even when the subject is not attending the auditory stimuli. Therefore, it has been popular in investigating a variety of patient groups during the recent years (for reviews, see Näätänen et al., 2007; Kujala et al., 2007). In addition to the processing of syllables or (pseudo-)words, the neural activity during comprehension of complex real-life speech by recording single-trial MEG will be characterized. A model-free analysis method investigates inter-subject correlations (ISC) between individuals with and without dyslexia. This approach will give insight about auditory processing characteristics in a more natural condition. The method of ISC has been proven viable in research during natural conditions, e.g. movie watching or listening to music, mainly using functional magnetic resonance imaging (fMRI). It has been introduced by Hasson and colleagues in 2004, who found that brain regions synchronize between subjects during movie viewing. In MEG, this approach has only been used scarcely (e.g. Lankinen et al., 2014; Suppanen, 2014; Thiede, 2014). However, it confirms and complements the results from fMRI research. In language research, temporal correlations that were investigated with the ISC method, have only been reported in resting state networks in children with reading difficulties (Dimitriadis et al., 2013). The results confirm findings from fMRI studies. Thus, the method offers new, promising insights into the neural underpinnings of dyslexia. Developmental dyslexia is a heritable disorder with polygenic origin. Several candidate genes have been detected in the recent years (Kere, 2014 for a review), associated with functions of neuronal migration and auditory processing. However, the connection between the auditory responses of the brain and the genetic cause of the disorder has not been confirmed yet. More research on the connections between the genetic and neural markers of dyslexia is needed in order to verify different existing hypotheses. Previous activity Research in the experimenters group has shown that at group level, cortical low-level discrimination of sounds and speech sounds is impaired in dyslexia (Kujala, 2007 for a review). This was reflected in diminished MMN responses (e.g., Kujala et al., 2006; Schulte-Körne et al., 1998; Neuhoff et al., 2012) and in enhancements of MMNs as a result of dyslexia intervention (Kujala et al., 2001; Lovio et al., 2012). These results suggest a strong connection between this neural response and dyslexia. Furthermore, the investigators' results suggest weak associations between the neural representations of speech sounds and written letters in the brains of dyslexics (Mittag et al., 2013), which could reflect deficits in phonological skills in dyslexia. Furthermore, studies carried out in the current research group have lead to the identification of candidate genes of dyslexia (e.g. Hannula-Jouppi et al., 2005; Schumacher et al., 2006). In a recent review, nine genes and four gene loci have been listed in a summary of genetic loci associated with developmental dyslexia (Kere, 2014). Some candidate genes were shown to be linked with axonal connections and others with neuronal migration functions.
2. Objectives and methods

This study aims at determining neurocognitive underpinnings of dyslexia and their connection with genes. The target neural processes impaired in different subgroups of dyslexic individuals involve phonological representations vs. phonological skills. Phonological representations are reflected in the low-level discrimination of speech sounds, whereas phonological skills can be mirrored in audiovisual integration of written and spoken letters. With neuropsychological tests, it will be determined which participants have primarily weak speech representations or poor phonological skills. It is hypothesized that participants with weak phonological representations have diminished MMN/MMNm responses. However, participants who do not have these problems, but instead have poor phonological skills, are expected to have normal-like MMN/MMNm responses but deficient responses reflecting audiovisual integration, such as no difference in effects of printed text versus nonsense visual material on early auditory speech discrimination. These two types of dyslexic groups are hypothesized to have alterations in partly different dyslexia candidate genes.

Auditory processing in the brain under more complex, real-world conditions will be investigated using single-trial MEG during presentation of natural speech sounds. It is expected that the synchronized brain activity differs between the dyslexic and control group. Specifically, it is expected that dyslexic subjects show decreased synchrony in the left temporoparietal cortex (Temple, 2002). Single-trial data could furthermore show a connection to genes, as suggested by Giraud and Ramus (2013), who hypothesized that a disruption of auditory cortical oscillations modifies the access to phonological representations.

Dyslexia is known to be associated with several candidate genes (for a review, see Kere, 2014). The candidate genes are associated with language skills and with the brain's ERPs. The gene research in cooperation with Prof. Juha Kere's laboratory in Folkhälsan research centre in Biomedicum, Helsinki or Karolinska Institutet, Stockholm, will aim at proofing the connection between dyslexia candidate genes and neural event-related activity of the brain to pseudoword stimuli.

Stimuli and procedure

MMN and MMNm responses to sound changes in pseudoword /tata/ (vowel, vowel duration, and syllable frequency changes) will be recorded while participants attend a movie or see concurrent visually-presented pseudoword stimuli. During the auditory condition, subjects will watch a silenced movie while being presented frequent "standard" stimuli, namely a pseudoword (/tata/) and infrequent "deviant" (see below) auditory stimuli. The task is to attend the movie and to ignore the background sounds. The audiovisual condition includes the same pseudoword stimuli, but instead of watching a movie, subjects will see the written letters of the presented pseudoword or a scrambled picture of the pseudoword letters. Distractors, such as counting deviant stimuli or different shapes and colours of the visual stimulus, will be given as the main task to the participants. They will be instructed to ignore the sounds.

The pseudoword stimuli are as follows:

• standard: /tata/
• vowel deviant: /tato/
• frequency deviant - higher frequency in the 2nd syllable
• duration deviant: /tataa/ - 2nd syllable twice as long as in the standard

The stimuli will be presented in a combined multi-feature and oddball design (Näätänen et al., 2004). All deviant types will be presented in the same sequence with 1-4 standards between each successive 2 deviants.

In addition to recording MMNm responses, a single-trial, continuous recording of brain activity for a real-world stimulus will be recorded. Approximately 8 min of natural speech (Finnish speaker) will be presented to both groups with the task of mere listening and keeping the eyes open. A recording of 8 min during rest with eyes open will complete the addition.

In order to examine differences in general cognitive abilities and performance profile between groups, the participants will undergo behavioural tests. Dyslexia characteristics will be assessed with parts of the Nevala dyslexia test (Nevala et al., 2006). General and performance Intelligence Quotient (IQ) as well as phonological and working memory will be tested using the Wechsler Intelligence Scale (WAIS-III; Wechsler, 1997a) and subtests of Wechsler Memory Scale (WMS-III; Wechsler, 1997b). Phonological naming will be assessed with the rapid alternate stimulus naming (RAS) test for speed and accuracy (Wolf, 1986). These or corresponding neuropsychological tests will be executed in max. 2 hours in an independent test session from the MEG session.

Additionally, saliva or blood samples (2x9 ml blood) will be collected by a trained nurse after subject's approval. DNA will be extracted from these samples and stored in Folkhälsan research centre in Juha Kere's laboratory. The DNA analysis focuses on any related candidate genes in their different variants using DNA sequencing techniques to determine the genotypes (Taqman, Sequenom). Possible links between the electric and magnetic activity of the brain and candidate genes for dyslexia are searched for.