Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 May 15.
Published in final edited form as: Brain Res. 2021 Feb 23;1759:147386. doi: 10.1016/j.brainres.2021.147386

Greater Reading Gain Following Intervention is Associated with Low Magnetic Resonance Spectroscopy Derived Concentrations in the Anterior Cingulate Cortex in Children with Dyslexia

Kim M Cecil 1,2,3, Kelly J Brunst 1, Tzipi Horowitz-Kraus 4,5,6
PMCID: PMC7980091  NIHMSID: NIHMS1681149  PMID: 33631208

Abstract

Background/Objective:

The “neural noise” hypothesis suggests that individuals with dyslexia have high glutamate concentrations associated with their reading challenges. Different reading intervention programs have showed low GLX (a combined measure for glutamine and glutamate obtained with in vivo magnetic resonance spectroscopy) in association with reading improvement. Several studies demonstrated improved reading and increased activation in the anterior cingulate cortex following an-executive-function (EF)-based reading intervention. The goals of the current study are two-fold: 1) to determine if the effect of the EF-based reading program extends also to the metabolite concentrations and in particular, on the GLX concentrations in the anterior cingulate cortex; 2) to expand the neural noise hypothesis in dyslexia also to neural networks supporting additional parts of the reading networks, i.e. in specific regions related to executive function skills.

Methods:

Children with dyslexia and typical readers were trained on the EF-based reading program. Reading ability was assessed before and after training while spectroscopy data was obtained at the end of the program. The association between change in reading scores following intervention and GLX concentrations was examined.

Results:

Greater “gains” in word reading were associated with low GLX, Glu, Cr, and NAA concentrations for children with dyslexia compared to typical readers.

Conclusions:

These results suggest that the improvement reported following the EF-based reading intervention program also involved a low GLX concentration, as well as additional metabolites previously associated with better reading ability (Glx, Cr, NAA) which may point at the decreased neural noise, especially in the anterior cingulate cortex, as a possible mechanism for the effect of this program.

Keywords: Glutamate-glutamine, executive function, dyslexia, neural noise hypothesis

Introduction

Neural noise hypothesis

Developmental dyslexia is a developmental reading disorder (RD) characterized by a specific deficit in reading accuracy and/or speed that continues into adulthood despite an intact IQ, repeated remediation and exposure to written language(IDA, 2011). Several theories were raised regarding the basic mechanism causing dyslexia. Amongst them, a deficit in phonological processing (Snowling et al., 1997), challenges in orthography (Brunswick et al., 1999), speed of processing (Breznitz and Misra, 2003), and specific challenges in executive function skills (EF) (Horowitz-Kraus, 2014). Due to these multiple challenges observed in individuals with RD, a recent concept called the “neural noise” hypothesis was suggested, which captures all of the above mentioned challenges reported in this disorder (Hancock et al., 2017). Traditionally, the “neural noise” hypothesis proposes that the variability in neuronal signal generation or transmission can result from every step in the cascade of neuronal signaling (ion channels, neurotransmitters to network connectivity)(Hancock et al., 2017). The noise can result from extreme excitability or challenges in balancing the excitatory and inhibitory signals (Hancock et al., 2017). This theory related to dyslexia, posits that the excessive activation of the glutamatergic system, might be associated with the cortical excitability, as observed in the visual system in these readers (Hancock et al., 2017). In return, this increased excitability is related increased neural noise. This increased neural noise in individuals with dyslexia is related to decreased synchronization between all language components, lexical access (orthographical processing), noise exclusion from the auditory and visual system, the integration of information and eventually reading impairment (Hancock et al., 2017). This theory is unique in integrating many clinical aspects related to the source of dyslexia as well as several etiologies aiming to explain it. This hypothesis is supported by several intervention studies suggesting that greater reading gains were associated with low GLX (a combined glutamate and glutamine measurement with in vivo magnetic resonance spectroscopy (Kossowski et al., 2019; Ramadan et al., 2013)). In line with the hypothesis, these studies examined the concentration of GLX in language visual processing-related cortices (Kossowski et al., 2019). Additional spectroscopy studies examining the relationship between metabolites levels and reading abilities, suggested that better reading abilities in those with dyslexia are associated with lower Glutamate (Glu)(Del Tufo et al., 2018; Hancock et al., 2017). Creatine (Cr) phosphate buffers the adenosine triphosphate concentration in cells with dynamic and/or high energy demands, as well as providing an energy shuttle from production in the mitochondria to use within nerve terminals (Laycock et al., 2008). N-acetylaspartate (NAA), a marker for neuronal viability, density and mitochondrial activity in the brain was positively related to functional MRI signal level in the occipital cortex in children with dyslexia, i.e. reduced activation was related to reduced NAA concentration (Ibrahim, 2008).

However, these studies focused solely on the visual, auditory and sensory systems and neural circuits. As the reading network also includes regions related to EF, and as children with dyslexia also exhibit additional impairments in EF (Horowitz-Kraus, 2014), we postulated that this neural noise hypothesis extends the classical language and visual processing also to EF-related brain regions and networks.

Hence, low GLX and Glu concentrations in the anterior cingulate cortex (ACC) following training with the above mentioned EF-based intervention may be present. We also suggest that better reading abilities following training will also be associated with lower concentrations in metabolites previously related to higher reading abilities in dyslexia (Cr) and higher NAA concentration representing an increased neuronal viability.

Executive dysfunction in dyslexia

Accumulated data suggested that individuals with dyslexia suffer from challenges in EF (Horowitz-Kraus, 2014; Horowitz-Kraus, 2016b; Horowitz-Kraus et al., 2016; Horowitz-Kraus et al., 2018b; Levinson et al., 2018). In addition to decreased visual attention, inhibition, switching and speed of processing observed in children with dyslexia (Horowitz-Kraus, 2014), the alterations in neural circuits supporting executive function skills were observed during both linguistic tasks (Horowitz-Kraus et al., 2016), reading (Freedman et al., 2020) and non-linguistic tasks (Horowitz-Kraus, 2014; Levinson et al., 2018). Recently, differences in metabolite concentrations were observed in a key-region of the EF system, the ACC in children with dyslexia (Horowitz-Kraus et al., 2018a), supporting the extension of this alteration in regions responsible for EF.

Findings from executive function-based reading intervention

Several studies examined the effect of EF-based interventions on reading abilities in individuals with dyslexia (Horowitz-Kraus, 2012; Horowitz-Kraus, 2013; Horowitz-Kraus and Breznitz, 2014; Horowitz-Kraus et al., 2014; Horowitz-Kraus, 2015; Horowitz-Kraus et al., 2015a; Horowitz-Kraus et al., 2015b; Horowitz-Kraus, 2016a). These interventions indicated that an EF-based reading intervention that “forces” the reader to read faster and to process more letters at a given time, not only improved reading and EF in children (Horowitz-Kraus, 2013; Horowitz-Kraus et al., 2015b) and adults with dyslexia (Horowitz-Kraus, 2016a) but also showed increased activation in neural circuits supporting EF (such as the ACC) during a lexical decision task while reading words (Horowitz-Kraus and Holland, 2015). In addition, the improved behavioral outcomes were associated with increased functional connections within the cingulo-opercular network (including the ACC hub) during a resting state condition (Horowitz-Kraus et al., 2015b), increased functional connections between the ACC and the visual word for area during a reading task (Horowitz-Kraus and Holland, 2015) and increased electroencephalogram (EEG) signal while committing reading errors (error related negativity, ERN, representing error detection)(Horowitz-Kraus and Breznitz, 2014). These studies suggested that the EF-based reading intervention trains both their visual attention, working memory and speed of processing and allows the trainees to establish a more stable mental lexicon (Breznitz, 1997b). Hence, when they perform reading errors, their monitoring is much more efficient and their error monitoring is enhanced. Additional studies suggested that this effect is not specific to word reading/linguistic stimulation, as improved activation of the EF system was observed in these readers during the Wisconsin task (non-linguistic task) (Horowitz-Kraus, 2015) and resting state functional magnetic resonance imaging (Horowitz-Kraus et al., 2015b), respectively. Therefore, an assumption related to an overall improvement in EF and neural circuits supporting EF which is not specific to reading arises. In the current study, we suggest that the neural noise hypothesis in individuals with dyslexia may not be restricted to visual and audio/language or sensory regions as suggested by Hancock et al. (Hancock et al., 2017), and that it might be extended also to EF regions. Hence, low GLX and Glu concentrations in the ACC following training with the above mentioned EF-based intervention may be present. We also suggest that better reading abilities following training will also be associated with lower concentrations in metabolites previously related to higher reading abilities in dyslexia (Cr) and higher NAA concentration representing an increased neuronal viability, and connectivity.

The goal of the current study was to determine if training with the EF-based reading intervention will also be related to low GLX in key regions in the monitoring system, i.e. in the ACC. We hypothesize that better gains from intervention, measured using the Letter-Word reading task, will be associated with low GLX for children with dyslexia. We suggest that this will represent the decreased “neural noise” in the ACC and represents an extension of the neural noise theory.

Results

Behavioral measures

Overall, children with dyslexia demonstrated significantly lower scores in all reading measures taken in the study: these readers showed lower timed and non-timed phonological and orthographical abilities. General cognitive abilities and attention did not differ between the groups (Table 1).

Table 1.

Differences in baseline general cognitive, reading and reading-related measures in children with dyslexia and typical readers.

Children with dyslexia Typical readers T(p)
Assessment Mean Standard Deviation Mean Standard Deviation
General cognitive ability (TONI, percentile) 99.24 11.53 103.45 8.37 −1.527, ns
Attention (Conners, Parent, T score) 69.43 17.02 52.48 14.91 3.657**
Phonological processing (Ellison, CTOPP, Scaled Score) 7.48 1.77 11.71 2.39 −6.9***
Orthographic abilities-timed (TOWRE, SWE, Scaled score) 82.05 12.63 108.55 11.6 −7.797***
Orthographic abilities not-times (Letter-Word, Woodcock Johnson, Standard score) 89.43 12.59 115 7.56 −9.148***
Decoding-timed (TOWRE, PDE, Scaled score) 80.43 10.57 109.8 9.43 −10.492***
Open in a new tab

Mean and standard deviation scores for cognitive and reading measures before intervention (Test 1) in children with dyslexia and typical readers. T scores representing the differences and results significant differences are marked in the right handed column (***p<.001; **p<.01). Abbreviations: TONI, Comprehensive Test of Phonological Processing (CTOPP), Testing of Word Reading Efficiency- sight word efficiency (TOWRE -SWE), Testing of Word Reading Efficiency- phonetic decoding efficiency (TOWRE-PDE)

Differences in gain following intervention in children with dyslexia and typical readers

The change in Letter-Word scores did not significantly differ by group status (t-value −0.62, p-value 0.54). Table 2 illustrates the differences in Letter-Word reading gain among children with dyslexia and typical readers.

Table 2.

Differences in gain among children with dyslexia and typical readers

Letter-Word, gain (Mean) Letter-Word, gain (Standard deviation) Letter-Word, gain (Minimum) Letter-Word, gain (Maximum)
Children with Dyslexia (n=21) 3.81 8.93 −12.00 18.00
Typical Readers (n=31) 3.92 10.30 −27.00 25.00
Open in a new tab

Mean, standard deviation, minimum and maximum scores of the gain in Letter-Word standard scores in children with dyslexia and typical readers. There were no significant differences in Letter-Word gain between children with dyslexia and typical readers (t-value = −0.04, p-value = 0.95).

Association between reading gains following intervention and metabolites concentrations in children with dyslexia and typical readers

Letter-Word reading gain was not associated with metabolite concentrations in unadjusted or adjusted models (Table 3). However, the associations between gain in Letter-Word scores and metabolite concentrations did vary between children with dyslexia and typical readers. Glu (pinteraction=0.06), GLX (pinteraction =0.007), Cr (pinteraction =0.009), and NAA (pinteraction =0.05) concentrations decreased among children with dyslexia and increased among typical readers following the intervention (Figure 1); controlling for global executive function (using the BRIEF General cognitive measure (Baron, 2000)) did not alter these results in a significant manner; differences in Cho (pinteraction =0.28) and mI (pinteraction =0.31) were not observed.

Table 3.

Association between changes in Letter-Word gain and metabolite concentrations

Letter-Word GAIN (unadjusted) Letter-Word GAIN (adjusted for age & sex)a
Outcome (Metabolite) β 95%CI p-value β 95%CI p-value
mI 0.002 −0.01, 0.02 0.83 0.002 −0.01, 0.02 0.79
NAA −0.006 −0.03, 0.02 0.65 −0.008 −0.03, 0.02 0.55
Cr 0.01 −0.001, 0.03 0.07 0.009 −0.006, 0.02 0.25
Cho 0.0004 −0.0035 0.84 −0.001 −0.005, 0.003 0.66
Glu 0.007 −0.009, 0.02 0.41 0.009 −0.008, 0.03 0.29
GLX 0.01 −0.011, 0.03 0.35 0.01 −0.01, 0.03 0.41
Open in a new tab

Unadjusted and adjusted (for sex and age) associations between changes in Letter-Word gain and metabolite concentrations. Metabolite concentrations are modeled separately and each metabolite’s parameter estimate (β), 95% confidence interval (95% CI), p-value is shown with significant results are bolded. Abbreviations: mI (myo-inositol), NAA (N-Acetylaspartate), Cr (Creatine), Cho (choline), Glu (glutamate), GLX (glutamate + glutamine).

a

Additional adjustment for executive function ability did not alter the results.

Figure 1.

Figure 1.

Open in a new tab

Association between Letter-Word gain and metabolite concentrations significantly varied amongst children with dyslexia (dashed) and typical readers (solid)

Association between GLX concentrations and other metabolites detected in the ACC

Low GLX concentrations were positively associated with low Glu and Cr concentrations across groups (Table 4). Low Glu concentrations were positively associated with low GLX, NAA and Cr concentrations.

Table 4.

Partial Pearson correlations between metabolite concentrations

mI NAA Cho Cr Glu GLX
mI 1 0.58183 0.60267 0.66073 0.3167 0.22821
NAA 0.58183 1 0.27886 0.52295 0.41278 0.35042
Cho 0.60267 0.27886 1 0.61949 0.30639 0.2794
Cr 0.66073 0.52295 0.61949 1 0.55175 0.50349
Glu 0.3167 0.41278 0.30639 0.55175 1 0.80152
GLX 0.22821 0.35042 0.2794 0.50349 0.80152 1
Open in a new tab

Correction for multiple comparisons was conducted by Bonferroni Adjusted p-value (0.05/6=0.0008). Correlations are adjusted for sex and age at first visit. Bolded correlations are significant (p-value < 0.0008). Association between metabolites Correction for multiple comparisons was conducted by Bonferroni Adjusted p-value (0.05/6=0.0008). Abbreviations: mI (myo-inositol), NAA (N-Acetylaspartate), Cr (Creatine), Cho (choline), Glu (glutamate), GLX (glutamate + glutamine).

Discussion

The goals of the current study were two-fold: 1) to determine if the effect of the EF-based reading program extends also to the metabolite concentrations and in particular, on the GLX concentrations in the anterior cingulate cortex; 2) to expand the neural noise hypothesis in dyslexia also to neural networks supporting additional parts of the reading networks, i.e. in specific regions related to executive function skills. The results of this study concur with our hypothesis focused on low GLX and Glu concentrations in the ACC following intervention were related to increased word reading abilities for children with dyslexia, even when controlling for basic executive functions abilities. The low GLX and Glu in the ACC, adds to previous findings of increased activation of the ACC (Horowitz-Kraus et al., 2014) of functional connectivity within the cingulo-opercular network (Horowitz-Kraus and Holland, 2015; Horowitz-Kraus et al., 2015b) increased event-related potential associated with error monitoring (error related negativity, ERN) (Horowitz-Kraus and Breznitz, 2014; Horowitz-Kraus, 2016a) and hence, strengthens the second hypothesis of a possible extension of the neural noise theory to regions supporting executive function skills as well. In line with previous studies examining the metabolites levels related to reading showing that lower Cr concentrations are related to better reading (Laycock et al., 2008), and opposed to findings relating high NAA concentrations to better reading (Ibrahim, 2008; Kossowski et al., 2019), children with dyslexia in the current study showed associations between low Cr and low NAA and better reading gain when controlling for basic EF skills. Studies suggest that NAA concentration is related to neuronal viability, and white matter integrity as it may reflect connectivity and myelination (Del Tufo et al., 2018). The uncinate fasciculus is the white matter tract crossing the ACC, which may not be affected by the intervention used in the current study. Hence, no increased NAA was found as originally postulated. Also, most studies evaluating reading disorders with magnetic resonance spectroscopy report NAA in terms of ratios to either creatine or choline as few studies have performed individual metabolite determination as done in the current study. This complicates interpretation especially as Cr is often used as the denominator in the ratio values. A future study should examine the change following this intervention in fractional anisotropy in white matter tracts passing through the ACC in children with dyslexia. Another explanation is that the low NAA and the Cr in the ACC are part of the pathology characteristics (similarly to attention deficit hyperactive disorder (Hesslinger et al., 2001)), hence is not affected by this intervention.

EF-based reading program decreased neural noise

Reading interventions previously reported for children with reading difficulties have focused on the different components of reading: phonology (Graphogame (Lyytinen et al., 2007), Fast-forward (Corporation, 2004; Hook et al., 2001)) or reading comprehension (Ehren, 2005) but did not target EF directly. In this study, we provide additional support for the positive effect of the EF based reading intervention program on reading ability in children with and without reading difficulties. We also demonstrated, for the first time, that the positive effect of reading intervention on metabolite concentrations can be detected not only in neural circuits supporting visual processing (Pugh et al., 2014) but also in EF-based cortical regions, such as the ACC. In line with our previous assumptions, the current results confirm that a reading manipulation of deleting the letters from the screen in an increasing rate while monitoring comprehension, is related to increased words representations in the mental lexicon (Breznitz and Share, 1992) (previously validated by increased activation in the fusiform gyrus (Horowitz-Kraus et al., 2014)) as well as increased functional connections between these visual processing brain cortices and error monitoring ones (i.e. ACC) (Horowitz-Kraus et al., 2015a; Horowitz-Kraus and Holland, 2015). Based on these findings, we could therefore assume that this effect is mediated through the effect on visual cortices directly. However, additional findings have suggested increased functional connections within EF networks even during a resting condition (Horowitz-Kraus et al., 2015b) as well as increased error monitoring during a non-linguistic task such as the Wisconsin task (Horowitz-Kraus, 2015). The results of the current study extend these previous findings by demonstrating an association of reduced metabolites related to “excitation” (i.e. GLX, Glu) in the ACC together with reading improvement for children with dyslexia. As the ACC is related to error monitoring and on-going monitoring of task-performance, the results may suggest an overall “noise” reduction, previously related to GLX (Hancock et al., 2017). Hence, our study’s results extend these findings also to EF-related regions, specifically to the ACC, which can be directly mediated by the EF-based reading program.

Mechanistically, while the letters are being deleted in a faster manner, the reader has to process more characters (letters) in a given time, forcing the reader to 1) capture the words holistically and 2) visually track the letters according to the reading direction, without the ability to regress (which was found to be associated with reduced functional connections between EF-regions in dyslexia (Horowitz-Kraus et al., 2018b). Hence, it might be that this manipulation of fast processing and one-directional visual scanning (without the ability to regress) is related to “noise reduction” manifested by low GLX and Glu in children with dyslexia that is generalized beyond the reading domain, as was previously found in non-linguistic conditions.

Conclusions

Based on accumulated findings outlined above, we suggest that the neural noise hypothesis previously suggested based on changes in GLX and Glu in reading related regions (Hancock et al., 2017) may also be extended to include the ACC and EF, addition to reading ability (See Figure 2). As EF are key components associated with reading ability (Horowitz-Kraus, 2016b) and reading fluency (Freedman et al., 2020) it might be that a “noise reduction ”, both for typical readers and those with dyslexia is related to better performance in both EF and reading. Determining if a focal EF-training will be associated with even more robust GLX and Glu reduction and improved performance, or whether the GLX reduction in the ACC is secondary or primary to changes in the visual cortex are questions that still need to be answered.

Figure 2.

Figure 2.

Open in a new tab

A proposed model for the extension of the “neural noise” hypothesis to executive function and executive function-based neural circuits

Study limitations

Despite the promising nature of the results descried in the current study, there are some limitations that warrant consideration. The study sample size is modest, with participants having similar socioeconomic status and race, which may limit the generalizability of these findings to other populations. The current study did not include a baseline (pre-intervention) metabolic assessment (rather, it included only the “post assessment”) due to time constraints in the scanner during the first session. The visual system and phonological processing regions were not sampled, and hence we cannot exclusively state that the low GLX and other metabolite concentrations were solely in the ACC. The current study evaluated metabolite concentrations instead of ratio levels. This may complicate comparisons and interpretation with other published studies. Also, technical limitations for the MRS assessment, such as acquisition errors with voxel localization, incomplete water suppression, effects of motion, post-processing assumptions for relaxation rates, estimations of tissue concentration may have impacted our results. However, all attempts to minimize these errors and apply consistency across all participants were performed. Future studies should examine all reading-related regions and subject into a hierarchical model to demonstrate how GLX and other metabolites within the different components on the reading network best related to improved (or worsened) reading ability.

Methods and Materials

Participants

Twenty-one children with dyslexia (mean age = 9.73 years, SD = 1.07, 14 males) and 31 age matched typical readers (mean age=10.26 years, SD=1.126 year; 14 males) participated in the current study (t(50)=1.694, ns). All participants were native English speakers, with average socioeconomic status, White, right-handed, and displayed normal or corrected-to-normal vision in both eyes as well as had normal hearing. None had a history of neurological, mood disorders, or attention difficulties. Recruitment used emails, posted and commercial advertisements. All parents and children provided informed written consents and assents respectively, prior to inclusion in the study. Participants were compensated with $30 gift cards for completion of the study protocol. Cincinnati Children’s Hospital Medical Center Institutional Review Board approved this study. The research was performed in accordance with relevant guidelines/regulations.

Reading assessment

During the first meeting (Test 1), we confirmed the diagnosis of dyslexia using a battery of normative reading tests conducted in English. Inclusion criteria for the dyslexia group were standard score of −1 and below or meeting the 25% or below cutoff in words reading, decoding and comprehension abilities (as previously described (Kovelman et al., 2012)). The reading battery included a) words reading accuracy/orthography: the “Letter-Word” subset from the Woodcock & Johnson-III battery (Woodcock and Johnson, 1989) b) phonological processing: Elision subtest (from the Comprehensive Test of Phonological Processing (CTOPP battery (Wagner et al., 1999)), c) automatic orthographical abilities: the timed Testing of Word Reading Efficiency- sight word efficiency (TOWRE-SWE) (Torgesen et al., 1999) d) decoding: the timed Testing of Word Reading Efficiency-phonetic decoding efficiency TOWRE-PDE” subset (Torgesen et al., 1999).

Participants in the typical reader group were age-matched students who volunteered for the study with fluent and accurate reading (according to established normative concentrations). The behavioral (reading and executive function skills) assessment lasted approximately two hours. This data was also used for the Test 1 assessment (pre-intervention). To evaluate the effect of intervention, we used the Letter-Word reading test following intervention (Test 2).

Study procedure

Following the administration of the first testing (i.e. baseline behavioral testing) and assent/consent signing procedure, the reading intervention program was administered via a computer using internet-based software. Following training, reading outcome (i.e. Letter-Word task) was administered to all participants using a pencil and paper testing format.

The EF-based reading program

Stimuli.

The EF-based reading intervention program features a bank of 1500 sentences with moderate-to-high frequency encountered vocabulary words in the English language (http://www.wordfrequency.info/). Each stimulus included a sentence with a multiple-choice question followed by four possible answers. Each sentence length spanned 9 – 12 words, comprised of 45–70 letters, letter width of 5 mm, extending over 1 to 2 lines, and with 18 mm between lines. Each sentence appeared once during the entire training intervention. The sentences were tested and verified for their level of difficulty in previous studies (Breznitz, 2006; Breznitz et al., 2013; Horowitz-Kraus and Breznitz, 2014).

Training procedure.

Reading training was administered via the internet using a computer in the participant’s home. The participants’ compliance was monitored by remote access to the training records, with verification of the record of five training sessions per week. Only datasets of participants who completed at least 18 training sessions (out of a total of 20 sessions) were included in the study. The participants were trained for 4 weeks, five times each week at 15–20 min per session for a total of 20 sessions and reading a different set of 50 randomly presented sentences in each session. The initial and the final reading pace and comprehension were measured by the evaluation mode of the program, which measures these variables in a self-paced reading condition (Breznitz et al., 2013).

The duration of a sentence display on the screen was calculated individually for each participant based on the evaluation mode and controlled by text erasure, starting from the beginning of the sentence and advancing at a given per-character rate. All participants were presented with the same sets of sentences, in the same order. They were instructed to read the sentence silently and while doing so, the sentence disappeared from the computer screen and a multiple-choice comprehension question appeared and remained on the screen until the participant responded. They were instructed to choose the correct answer by pushing the corresponding number on the numeric keypad of the computer. The disappearance of the question from the computer screen prompted appearance of the next sentence.

Presentation rate and evaluation mode.

The initial text erasure rate was determined specifically for each participant based on a pre-test evaluation mode administered prior to training. The evaluation mode consists of 12 sentences and 12 multiple-choice questions (Breznitz and Leikin, 2001). The sentences in the evaluation mode remained on the screen until the participants finished reading them. The participants were instructed to read the sentences silently and to push the space button on the keyboard when finished reading, which prompted a comprehension question. The mean reading rate (milliseconds (ms) per letter) for the sentence correctly answered determined the initial presentation rate of the program for that participant.

Accelerated training condition.

The initial reading rate in the training mode is determined based on the reading rate calculated in the evaluation mode (based on the reading rate of 12 sentences). In the first training session, 50 sentences were presented consecutively on the screen. The letters in each sentence disappeared one after the other, according to the mean reading time (ms per letter) recorded on the pre-test. Following the disappearance of the sentence from the computer screen, participants were instructed to answer a comprehension question. The per-letter “presentation rate” decreased from one sentence to the next by 2% for each subsequent sentence (Breznitz, 1997a; Breznitz, 1997b) and the “disappearance rate” increased only when the participant’s answers to the probe questions were correct on 10 consecutive sentences. In other words, the computer software pacing is modified periodically based on participant performance with the goal of increasing the pace over what would be chosen by the participant.

Magnetic Resonance Measures

Acquisition Methods.

As also reported in (Horowitz-Kraus et al., 2018a) brain magnetic resonance imaging (MRI) and spectroscopy (MRS) were acquired using a Philips Achieva MR scanner operating at 3 Tesla (3T) and equipped with a 32-channel head coil. A three-dimensional (3D), high-resolution, isotropic, T1-weighted fast Fourier echo anatomical imaging sequence was performed using 8.2 ms repetition time, 3.7 ms echo time, 1057 ms inversion time, 8 degree flip angle, sensitivity encoding factor (SENSE) of 2, contiguous slices with a 1 mm thickness, and 1 mm × 1 mm voxel size. A single voxel, point resolved spectroscopy (PRESS) sequence was conducted using a 2000 ms repetition time, 30 ms echo time, and 96 averages with water suppression along with an embedded unsuppressed water reference series of 16 averages. The 8 cubic centimeter single voxel was prescribed about the ACC within the medial frontal lobe localized from the 3D T1 anatomical imaging sequence (Figure 3).

Figure 3.

Figure 3.

Open in a new tab

A. Representative voxel placement for acquisition of MRS (shown in yellow on a T1-weighted 3D anatomical image) within the perigenual anterior cingulate cortex. B. Representative short echo spectrum with labels for N-acetyl aspartate (NAA), glutamate and glutamine (GLX), creatine (Cr), cholines (Cho) and myo-inositol (mI). X-axis is chemical shift shown in parts per million (ppm).

Participants were acclimated and desensitized to condition them for comfort inside the scanner (see (Byars et al., 2002) for details). Head motions were controlled using elastic straps that were attached to either side of the head-coil apparatus. An MRI-compatible audio/visual system (Avotec, SS3150/ SS7100) was used for the presentation of a movie during the session.

MRS data analysis.

The raw spectroscopy data were imported into LCModel (Provencher, 1993) commercial software for quantitative processing. Figure 4 illustrates for a child with dyslexia the LCModel fitting and output to illustrate the fit quality with Cramer Rao-Lower Bounds (CRLBs). Metabolite concentrations were determined in millimolar (mM) units. The FMRIB Software Library (FSL, http://www.fmrib.ox.ac.uk/fsl) brain extraction and segmentation tools were then used to segment each T1 image and calculate the percentage of each tissue type (grey matter, white matter and cerebrospinal fluid) within each MRS voxel. The raw metabolite concentrations were adjusted for the tissue contributions (Woolrich et al., 2009), adjusted to the T1 and T2 relaxation decay rate of the corrected water concentration and corrected for literature reported T1 and T2 relaxation decay rates of the primary metabolites including N-acetyl aspartate (NAA), choline (Cho), creatine (Cr) and myo-inositol (mI) (Tal et al., 2012; Traber et al., 2004; Wansapura et al., 1999). The relaxation rates depend on tissue type. However, glutamate and GLX concentrations were corrected for tissue contribution and water relaxation but remained unadjusted for metabolite T1 and T2 relaxation decay (Gussew et al., 2012). Supplemental Table 1 reports the mean and standard deviation of key spectral features including signal to noise ratio, full width half maximum peak width, adjusted metabolite concentrations, unadjusted metabolite ratios and CRLBs for metabolites by group. Figure 5 displays the composite of spectra from each participant group (children with dyslexia and typical readers) illustrating the mean and standard deviation.

Figure 4.

Figure 4.

Open in a new tab

Representative MR spectrum with LCModel output acquired from the perigenual anterior cingulate cortex for a child with dyslexia.

This spectrum was selected as it has a signal to noise ratio value of 27, which corresponds to the mean value for the group of readers with dyslexia (27.29). This spectrum does have a higher full width half-maximum (FWHM) value than the mean. It also demonstrates typical Cramer Rao Lower Bounds (as %SD) found in the study.

Figure 5.

Figure 5.

Open in a new tab

A composite from individual spectra for each group is presented to show the mean (shown with the red line) and standard deviation (shaded region) for each group (A children with dyslexia and B typical readers).

Statistical Analyses

First, two sample t-tests were used to examine differences in age and IQ measures, reading and executive function measures as well as metabolites concentrations among children with dyslexia and typical readers.

Second, to determine differences in gains following intervention between children with dyslexia and typical readers, a gain measure for the difference between the Letter-Word standard score before (Test 1) vs after (Test 2) intervention was calculated (i.e. Test 1 scores were subtracted from Test 2 scores). We then used two-sample t-tests to determine if the intervention resulted in significant improvements/changes in Letter-Word scores among children with dyslexia vs typical readers.

Third, to determine the association between metabolites following intervention, partial Pearson (r) correlations were calculated while controlling for sex and age at first visit. Correction for multiple comparisons was conducted by Bonferroni adjusted p-value (0.05/6=0.0008).

Fourth, linear regression was conducted to determine if a change in Letter-Word scores predicted metabolite concentrations following the intervention. Each metabolite was modeled individually and the change in Letter-Word scores was treated as a continuous predictor. Unadjusted and adjusted (for sex and age at first visit) models are presented; as a secondary analysis we controlled for global executive function using the Behavioral Rating Inventory of Executive functions (BRIEF)(Baron, 2000).

Lastly, we investigated whether the relationship between a change in Letter-Word score and metabolites level varied among children with dyslexia and typical readers. Thus, we included an interaction term (Letter-Word×dyslexia status) to our regression model described above. Interaction p-values less than 0.05 were considered significant. Analyses were conducted with SAS 9.4 (SAS Institute Inc., Cary, NC, USA) and SPSS13.

Supplementary Material

1

Highlights.

  • An executive function-based reading intervention is related to an increased reading ability in children.

  • Greater “gains” in word reading were associated with low GLX, Glu, Cr, and NAA concentrations for children with dyslexia compared to typical readers.

  • This study supports the extension of the neural noise hypothesis in dyslexia also to cortical regions responsible for executive function skills.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Baron IS, 2000. Behavior Rating Inventory of Executive Function. Child Neuropsychol. 6, 235–238. 10.1076/chin.6.3.235.3152. [DOI] [PubMed] [Google Scholar]
  2. Breznitz Z, Share DL, 1992. Effects of accelerated reading rate on memory for text. J Educ Psychol. 84, 193–199. 10.1037/0022-0663.84.2.193. [DOI] [Google Scholar]
  3. Breznitz Z, 1997a. Enhancing the reading of dyslexic children by reading acceleration and auditory masking. J Educ Psychol. 89, 103–113. 10.1037/0022-0663.89.1.103. [DOI] [Google Scholar]
  4. Breznitz Z, 1997b. Effects of accelerated reading rate on memory for text among dyslexic readers. J Educ Psychol. 89, 289–297. 10.1037/0022-0663.89.2.289. [DOI] [Google Scholar]
  5. Breznitz Z, Leikin M, 2001. Effects of accelerated reading rate on processing words’ syntactic functions by normal and dyslexic readers: event related potentials evidence. J Genet Psychol. 162, 276–296. 10.1080/00221320109597484. [DOI] [PubMed] [Google Scholar]
  6. Breznitz Z, Misra M, 2003. Speed of processing of the visual-orthographic and auditory-phonological systems in adult dyslexics: the contribution of “asynchrony” to word recognition deficits. Brain Lang. 85, 486–502. 10.1016/s0093-934x(03)00071-3. [DOI] [PubMed] [Google Scholar]
  7. Breznitz Z, 2006. Fluency in Reading: Synchronization of Processes. Lawrence Erlbaum Associates, Mahwah, New Jersey. [Google Scholar]
  8. Breznitz Z, et al. , 2013. Enhanced reading by training with imposed time constraint in typical and dyslexic adults. Nat Commun. 4, 1486. 10.1038/ncomms2488. [DOI] [PubMed] [Google Scholar]
  9. Brunswick N, et al. , 1999. Explicit and implicit processing of words and pseudowords by adult developmental dyslexics: A search for Wernicke’s Wortschatz? Brain. 122 (Pt 10), 1901–1917. 10.1093/brain/122.10.1901. [DOI] [PubMed] [Google Scholar]
  10. Byars AW, et al. , 2002. Practical aspects of conducting large-scale functional magnetic resonance imaging studies in children. J Child Neurol. 17, 885–890. 10.1177/08830738020170122201. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Scientific Learning Corporation, 2004. Fast ForWord: Gateway edition protocols. http://www.scilearn.com/alldocs/training accessed 2014.
  12. Del Tufo SN, et al. , 2018. Neurochemistry Predicts Convergence of Written and Spoken Language: A Proton Magnetic Resonance Spectroscopy Study of Cross-Modal Language Integration. Front Psychol. 9, 1507. 10.3389/fpsyg.2018.01507. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Ehren B, 2005. Looking for Evidence-Based Practice in Reading Comprehension Instruction. Top Lang Disord. 25, 310–321. 10.1097/00011363-200510000-00005. [DOI] [Google Scholar]
  14. Freedman L, et al. , 2020. Greater functional connectivity within the cingulo-opercular and ventral attention networks is related to better fluent reading: A resting-state functional connectivity study. Neuroimage Clin. 26, 102214. 10.1016/j.nicl.2020.102214. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Gussew A, et al. , 2012. Absolute quantitation of brain metabolites with respect to heterogeneous tissue compositions in (1)H-MR spectroscopic volumes. MAGMA. 25, 321–333. 10.1007/s10334-012-0305-z. [DOI] [PubMed] [Google Scholar]
  16. Hancock R, et al. , 2017. Neural Noise Hypothesis of Developmental Dyslexia. Trends Cogn Sci. 21, 434–448. 10.1016/j.tics.2017.03.008. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Hesslinger B, et al. , 2001. Attention-deficit disorder in adults with or without hyperactivity: where is the difference? A study in humans using short echo (1)H-magnetic resonance spectroscopy. Neurosci Lett. 304, 117–119. 10.1016/s0304-3940(01)01730-x. [DOI] [PubMed] [Google Scholar]
  18. Hook PM, et al. , 2001. Efficacy of Fast ForWord training on facilitating acquisition of reading skills by children with reading difficulties—A longitudinal study. Annals of Dyslexia. 51, 73–96. 10.1007/s11881-001-0006-1. [DOI] [Google Scholar]
  19. Horowitz-Kraus T, 2012. The Error Detection Mechanism Among Dyslexic and Skilled Readers: Characterization and Plasticity. In Reading, Writing, Mathematics and the Developing Brain: Listening to Many Voices. Literacy Studies, Vol. 6, Breznitz Z, Rubinsten O, Molfese VJ, Molfese DL, eds. Dordrecht: Springer, pp. 113–130. [Google Scholar]
  20. Horowitz-Kraus T, 2013. Can we train the dyslexic readers to read like a typical readers? EEG and fMRI study using the Reading Acceleration Program. In Brain, Mind and Fluency Conference. Edmond J. Safra Brain Research Center for the Study of Learning Disabilities, Haifa, Israel. [Google Scholar]
  21. Horowitz-Kraus T, 2014. Pinpointing the deficit in executive functions in adolescents with dyslexia performing the Wisconsin card sorting test: an ERP study. J Learn Disabil. 47, 208–223. 10.1177/0022219412453084. [DOI] [PubMed] [Google Scholar]
  22. Horowitz-Kraus T, Breznitz Z, 2014. Can reading rate acceleration improve error monitoring and cognitive abilities underlying reading in adolescents with reading difficulties and in typical readers? Brain Res. 1544, 1–14. 10.1016/j.brainres.2013.11.027. [DOI] [PubMed] [Google Scholar]
  23. Horowitz-Kraus T, et al. , 2014. Reading acceleration training changes brain circuitry in children with reading difficulties. Brain Behav. 4, 886–902. 10.1002/brb3.281. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Horowitz-Kraus T, 2015. Improvement in non-linguistic executive functions following reading acceleration training in children with reading difficulties: An ERP study. Trends Neurosci Educ. 4, 77–86. doi: 10.1016/j.tine.2015.06.002. [DOI] [Google Scholar]
  25. Horowitz-Kraus T, et al. , 2015a. Increased resting-state functional connectivity of visual- and cognitive-control brain networks after training in children with reading difficulties. Neuroimage Clin. 8 619–630. 10.1016/j.nicl.2015.06.010. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Horowitz-Kraus T, Holland SK, 2015. Greater functional connectivity between reading and error-detection regions following training with the reading acceleration program in children with reading difficulties. Ann Dyslexia. 65, 1–23. 10.1007/s11881-015-0096-9. [DOI] [PubMed] [Google Scholar]
  27. Horowitz-Kraus T, et al. , 2015b. Increased Resting-State Functional Connectivity in the Cingulo-Opercular Cognitive-Control Network after Intervention in Children with Reading Difficulties. PLoS One. 10, e0133762. 10.1371/journal.pone.0133762. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Horowitz-Kraus T, 2016a. Improvement of the Error-detection Mechanism in Adults with Dyslexia Following Reading Acceleration Training. Dyslexia. 22, 173–189. 10.1002/dys.1523. [DOI] [PubMed] [Google Scholar]
  29. Horowitz-Kraus T, 2016b. The Role of Executive Functions in the Reading Process. In Reading Fluency: Current Insights from Neuro-Cognitive Research and Intervention Studies. Khateb A, B.-K. I, ed. Springer, Netherlands, pp. 51–63. [Google Scholar]
  30. Horowitz-Kraus T, et al. , 2016. Altered neural circuits accompany lower performance during narrative comprehension in children with reading difficulties: an fMRI study. Ann Dyslexia. 301–318. 10.1007/s11881-016-0124-4. [DOI] [PubMed] [Google Scholar]
  31. Horowitz-Kraus T, et al. , 2018a. Children With Dyslexia and Typical Readers: Sex-Based Choline Differences Revealed Using Proton Magnetic Resonance Spectroscopy Acquired Within Anterior Cingulate Cortex. Front Hum Neurosci. 12, 466. 10.3389/fnhum.2018.00466. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Horowitz-Kraus T, et al. , 2018b. Longer Fixation Times During Reading Are Correlated With Decreased Connectivity in Cognitive-Control Brain Regions During Rest in Children. Mind Brain Educ. 49–60. 10.1111/mbe.12168. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Ibrahim H, 2008. Interactions between fMRI BOLD-activation during Reading Tasks and MRS-measured Metabolite Levels. Yale University School of Medicine, Yale Medicine Thesis Digital Library. [Google Scholar]
  34. IDA, 2011. Definition of dyslexia-Based in the initial definition of the Research Committee of the Orton Dyslexia Society, former name of the IDA, done in 1994. In International Dyslexia Association. [Google Scholar]
  35. Kossowski B, et al. , 2019. Dyslexia and age related effects in the neurometabolites concentration in the visual and temporo-parietal cortex. Sci Rep. 9, 5096. 10.1038/s41598-019-41473-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Kovelman I, et al. , 2012. Brain basis of phonological awareness for spoken language in children and its disruption in dyslexia. Cereb Cortex. 22, 754–764. 10.1093/cercor/bhr094. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Laycock SK, et al. , 2008. Cerebellar volume and cerebellar metabolic characteristics in adults with dyslexia. Ann NY Acad Sci 1145, 222–236. 10.1196/annals.1416.002. [DOI] [PubMed] [Google Scholar]
  38. Levinson O, et al. , 2018. Altered functional connectivity of the executive-functions network during a Stroop task in children with reading difficulties. Brain Connect. 8, 516–525. 10.1089/brain.2018.0595. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Lyytinen H, et al. , 2007. Early identification of dyslexia and the use of computer game-based practice to support reading acquisition. Nord Psychol. 59, 109–126. 10.1027/1901-2276.59.2.109. [DOI] [Google Scholar]
  40. Provencher SW, 1993. Estimation of metabolite concentrations from localized in vivo proton NMR spectra. Magn Reson Med. 30, 672–679. 10.1002/mrm.1910300604. [DOI] [PubMed] [Google Scholar]
  41. Pugh KR, et al. , 2014. Glutamate and choline levels predict individual differences in reading ability in emergent readers. J Neurosci. 34, 4082–4089. 10.1523/JNEUROSCI.3907-13.2014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  42. Ramadan S, et al. , 2013. Glutamate and glutamine: a review of in vivo MRS in the human brain. NMR Biomed. 26, 1630–1646. 10.1002/nbm.3045. [DOI] [PMC free article] [PubMed] [Google Scholar]
  43. Snowling M, et al. , 1997. Phonological processing skills of dyslexic students in higher education: a preliminary report. J Res Read. 20, 31–41. 10.1111/1467-9817.00018. [DOI] [Google Scholar]
  44. Tal A, et al. , 2012. The role of gray and white matter segmentation in quantitative proton MR spectroscopic imaging. NMR Biomed. 25, 1392–1400. 10.1002/nbm.2812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  45. Torgesen JK, et al. , 1999. Test of word reading efficiency (TOWRE). Vol., Pro-Ed, Austin, TX. [Google Scholar]
  46. Traber F, et al. , 2004. 1H metabolite relaxation times at 3.0 tesla: Measurements of T1 and T2 values in normal brain and determination of regional differences in transverse relaxation. J Magn Reson Imaging. 19, 537–545. 10.1002/jmri.20053. [DOI] [PubMed] [Google Scholar]
  47. Wagner RK, et al. , 1999. Comprehensive Test of Phonological Processing (CTOPP). Pro-Ed, Austin, TX. [Google Scholar]
  48. Wansapura JP, et al. , 1999. NMR relaxation times in the human brain at 3.0 tesla. J Magn Reson Imaging. 9, 531–538. . [DOI] [PubMed] [Google Scholar]
  49. Woodcock RW, Johnson MB, 1989. Woodcock-Johnson Psycho-Educational Battery-Revised (WJ-R). Developmental Learning Materials, Allen, TX. [Google Scholar]
  50. Woolrich MW, et al. , 2009. Bayesian analysis of neuroimaging data in FSL. NeuroImage. 45, S173–186. 10.1016/j.neuroimage.2008.10.055. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1681149-supplement-1.docx (29.8KB, docx)

RESOURCES