Introduction
Developmental dyslexia is defined as a specific deficit in reading acquisition that cannot be accounted for by low intelligence, poor educational opportunities, or any obvious sensory or neurological damage. Numerous theories were proposed trying to identify potential causes of dyslexia, however, no consensus was reached yet with regard to the neurological and cognitive basis of the disorder. It can be partly explained by the fact that the large body of data on cognitive deficits in dyslexia fails to fit into a single coherent theoretical framework and partly by the fact that the disorder is heterogeneous (Ramus and Ahissar
2012).
There is great variability in how dyslexia can be expressed in an individual relative to another depending on particular cognitive deficits that are present or not. Attempts were made to find distinguishable subtypes of the disorder. Several studies examined different dyslexia theories in multiple case studies of either adults (Ramus et al.
2003) or children (White et al.
2006) trying to assess the prevalence of each of the studied cognitive deficit adopting a criterion of deviance = 1.65 SD. In both aforementioned studies, performed on British native speakers, the phonological deficit was the most common, while the sensory deficits (magnocellular, auditory, motor/cerebellar) were less represented. Another study using a different criterion for deviance (below the 10th percentile) tried to determine if a proportion of dyslexic children could be characterized as suffering mainly from a selective visual span deficit or a phonological impairment (Bosse et al.
2007). The research showed that both French and British dyslexic children could be divided into four subgroups—having selective phonological, selective visual span deficit, both or none. Finally, using cluster analysis, Heim et al. (
2008) divided German dyslexic children into three clusters with different cognitive deficits. Specifically, cluster no. 1 compared to age-matched controls had worse phonological awareness, cluster no. 2 had an attention deficit, whereas cluster no. 3 performed worse on phonological, auditory, and magnocellular tasks.
These studies show that distinguishable phenotypes of dyslexia exist on the cognitive level. There is, however, much more limited understanding of the potential neural markers of these specific subtypes. At the neurofunctional level, first attempts were made to relate cognitive profiles of dyslexia to brain activation during reading (Heim et al.
2010b), phonological awareness, visual-spatial attention, visual processing, and auditory processing (Heim et al.
2010a; for a review see Heim and Grande
2012). However, we are not aware of any study successfully defining subtypes of dyslexia and providing evidence for distinct cognitive and neuroanatomical profiles associated with each subtype.
Previous voxel-based morphometry (VBM) studies revealed differences in brain structure between children and adults with developmental dyslexia and controls. When summarized in a review (Richardson and Price
2009), increases and decreases of grey matter volume (GMV) in group comparisons of developmental dyslexics and good readers, constitute a widely distributed set of regions in both left and right hemispheres. The most frequently reported areas include: posterior temporal/temporo-parietal regions with increases and decreases of GMV (Brambati et al.
2004; Hoeft et al.
2007; Silani et al.
2005; Steinbrink et al.
2008), decreases of GMV in the left inferior frontal (Brown et al.
2001; Eckert et al.
2005), decreases of GMV in the occipito-temporal regions bilaterally (Eckert et al.
2005; Brambati et al.
2004; Kronbichler et al.
2008), and decreases of GMV in the cerebellum bilaterally (Brown et al.
2001; Brambati et al.
2004; Eckert et al.
2005; Kronbichler et al.
2008). Finally, there are also studies where no differences in GMV between dyslexic and control groups were revealed (Pernet et al.
2009). Taking into account the heterogeneity of behavioral deficits, it is not surprising that VBM studies of dyslexia do not always agree one with the other, depending on the sample at hand, its age and the behavioral profile.
Here, we aimed at distinguishing specific dyslexic subtypes based on examining four cognitive domains: phonological awareness, rapid automatized naming, visual magnocellular-dorsal processing, and auditory attention shifting. We hypothesize that dyslexic subtypes can be characterized by a specific pattern of GMV.
Discussion
In the present study, we identified three subtypes of dyslexia with distinct cognitive and neurobiological profiles. All had severe reading and spelling impairments compared to controls but did not differ in these measures with each other. On the neuronal level, they showed reduced GMV in the left inferior frontal gyrus relative to age-matched good readers consistent with previous studies (Brown et al.
2001; Eckert et al.
2005; Vinckenbosch et al.
2005). Importantly, this effect was common across all dyslexic subtypes. The left inferior frontal gyrus contributed mostly to the third discriminant function, which was associated with single word and pseudoword reading, spelling, and rapid naming. In line, the activity in this area increased with reading ability and was related to rapid naming (Turkeltaub et al.
2003). It was also shown that the size of inferior frontal gyrus, particularly pars triangularis can predict rapid naming speed in a group of children with predominating double deficit (Eckert et al.
2005).
The first subtype of dyslexic children had worse phonological awareness and magnocellular-dorsal skills compared to all other groups. A similar cluster was previously described by Heim et al. (
2008) though in that study children besides phonological and magnocellular had also an auditory deficit. VBM revealed that compared to other groups this subtype was characterized of increased GMV in the left cerebellum, lingual gyrus and right putamen together with a decrease of GMV in the left parietal (mainly somatosensory) and right dorsal premotor cortices. However, only the differences in the left cerebellum, the right putamen and right dorsal premotor cortex were unique for this dyslexic subtype. The second dyslexic subtype had a reverse cognitive profile compared to first subtype, i.e., while phonological and magnocellular-dorsal skills were comparable to controls, the children showed impairments in rapid naming and auditory attention shifting. This was nicely reflected in GMV profiles since in the second subtype a decrease in GMV in the left cerebellum, lingual gyrus and an increase of GMV in the left parietal (somatosensory) cortex were observed, in regions overlapping with the ones showing a reverse pattern in subtype 1, together with an increase of GMV in the medial part of the right superior frontal gyrus. However, no unique effects were revealed for this subtype.
The structures differentiating subtypes 1 and 2 contributed to the first discriminant function, which was associated with phonological awareness. Previous anatomical studies yielded inconsistent results for the cerebellum of dyslexic subjects showing either decreased GMV compared to controls (Brown et al.
2001; Brambati et al.
2004; Eckert et al.
2005; Kronbichler et al.
2008) or no differences between the groups followed by a negative association between cerebellar GMV and phonological skills in controls (Kibby et al.
2008; Pernet et al.
2009). Nicolson et al. (
2001) have proposed two mechanisms by which cerebellum may play a role in dyslexia. The first is associated with so-called motor-articulatory feedback hypothesis (Heilman et al.
1996), which suggests that recognition of phonemes is dependent on awareness of the positions and movements of the articulatory system. Poor quality articulatory representations lead to impaired sensitivity to the phonemic structure of language and to reduced phonological awareness. The second is related to decreased processing speed, which is reflected in difficulties with rapid naming.
Left somatosensory and right premotor cortices GMV reduction in the first subtype might suggest problems with articulatory feedback, which produces a severe phonological deficit. This in turn might cause greater reliance on silent articulatory processes (Wimmer et al.
2010) when dealing with decoding resulting in increased GMV in cerebellum and putamen. The role of the latter in reading, mainly silent articulation (Hernandez and Fiebach
2006) and phonology (Tettamanti et al.
2005) has been shown previously although it was not specifically linked with dyslexia.
Interestingly, one structure, the right anterior/middle cingulate gyrus showed a common decrease of GMV for both the first and second dyslexic subtype. This region is widely believed to play a role in cognitive control, helping to resolve conflict from distracting events by focusing attention towards task-relevant stimuli (Weissman et al.
2005). Several studies have suggested a role of anterior cingulate during anticipation (Murtha et al.
1996) and voluntary attentional orienting (Hopfinger et al.
2000; Weissman et al.
2002). It was also shown that activity in this region was correlated with the level of attention dedicated to learning events (Bryden et al.
2011). Both the first and the second dyslexic subtype, although at first glance seemingly different with regard to the behavioral profile, showed deficits in two different tasks requiring high amounts of attentional control—coherent dot motion and stream segregation. It was shown that performance in the formed is largely modulated by attention orienting (Liu et al.
2006), whereas the latter is regarded as a measure of attention shifting (Lallier et al.
2009). It seems possible that the reduced GMV in the right anterior cingulate might lead to poorer performance on these two different tasks involving attention focusing. In line, anterior cingulate was consistently revealed as having decreased GMV in attention deficit disorder (Amico et al.
2011; Seidman et al.
2006). It remains, however, unclear why these two dyslexic subtypes are impaired either in coherent motion or SST and not in both. Most probably, the whole pattern of GMV changes and not only differences the right anterior cingulate influence the behavioral outcome.
The third subtype had a double deficit [described before by Wolf and Bowers (
1999)], whereas it had preserved magnocellular-dorsal and attentional shifting skills. Compared to other groups it was characterized of lower GMV in the right parietal cortex, however, no unique effects were found for this subtype. The right parietal cortex contributed to the second discriminant function significantly correlated with both perceptual thresholds—magnocellular-dorsal and auditory attention shifting. The role of right inferior parietal cortex for attention was well documented (Behrmann et al.
2004). It is also known that visual input the right parietal cortex projects from the magnocellular layers of the lateral geniculate nucleus (Eden and Zeffiro
1998). We found that the lower the perceptual thresholds the lower the GMV in the right inferior parietal cortex, in agreement with previous studies showing significantly larger GMV in inferior parietal lobule in adults with ADHD (Seidman et al.
2011). Additionally the third subtype showed common with the first subtype increase of GMV in the left cerebellum cluster, possibly reflecting deficient phonological awareness skills.
On the basis of the GMV in the significant clusters described above using a discriminant analysis, 79 % of cross-validated cases were correctly re-classified into four groups (controls versus three dyslexic subtypes).
In conclusion, our results are in line with the hypothesis that dyslexic subtypes can be characterized by specific patterns of GMV. We have dissociated three different groups of dyslexic behavior and identified brain areas with local GMV that differs from controls and between dyslexic groups. However, it seems that the relationship between brain structure and behavior is more complicated than anticipated, i.e., there is no clear, unique patterning of GMV differences. The obtained results form an intricate pattern of differences and not a plain 1:1 association between one subgroup and one region. Thus, based on a GMV difference in a specific brain area it is not possible to univocally predict the behavioral phenotype.
On the other hand, taking into account the complex aetiology of dyslexia it is not surprising that the array of GMV differences is also complex. Besides, having only four different cognitive tests, one cannot expect to fully describe the behavioral and neural phenotypes of dyslexia. Nevertheless, our study shows that it is important to look for potential subtyping of this disorder both on the behavioral and brain level. Further studies with a larger battery of cognitive tests and bigger sample size are needed to verify current findings. Lastly, it would be important to examine whether the revealed profiles (subtypes) are stable in development and whether they can be differentiated in younger children at the pre-reading stage.