Convergent and divergent structural and functional brain abnormalities associated with developmental dyslexia: a cross-linguistic meta-analysis of neuroimaging studies

Brain abnormalities in the reading network have been repeatedly reported in individuals with developmental dyslexia (DD); however, it is still not totally understood where and why the structural and functional abnormalities are consistent/inconsistent across languages. In the current multimodal meta-analysis, we found convergent structural and functional alterations in the left superior temporal gyrus across languages, suggesting a neural signature of DD. We found greater reduction in grey matter volume and brain activation in the left inferior frontal gyrus in morpho-syllabic languages (e.g. Chinese) than in alphabetic languages, and greater reduction in brain activation in the left middle temporal gyrus and fusiform gyrus in alphabetic languages than in morpho-syllabic languages. These language differences are explained as consequences of being DD while learning a specific language. In addition, we also found brain regions that showed increased grey matter volume and brain activation, presumably suggesting compensations and brain regions that showed inconsistent alterations in brain structure and function. Our study provides important insights about the etiology of DD from a cross-linguistic perspective with considerations of consistency/inconsistency between structural and functional alterations.

. In Chinese, phonological deficit 53 is associated with a higher rate of semantic errors during character reading (Shu et 54 al., 2005), because children with DD over-rely on the semantic cue in the character 55 during reading due to the inability to use the phonological cue. According to 56 research, 80% of Chinese characters have a semantic radical and a phonetic radical 57 Paulesu et al., 2000). It has also been found that the left TP is more involved in 139 English reading and the left dorsal IFG is more involved in Chinese reading (Bolger  170 brain structure might be due to the lack of differentiation in participants' language. It 171 is important to differentiate language-universal structural alterations as a core deficit 172 which might be related to the cause of DD and language-specific structural 173 alterations as a consequence of being DD in a specific language. While brain 174 structural deficits may cause reading disability, learning experience may also shape 175 brain development. Learning a specific language with DD may affect brain 176 development in language-specific regions (Mechelli et al., 2004). 177 DD is associated with altered brain structure and function, but very few studies 178 have investigated whether brain structural alterations and brain functional alterations  190 In this meta-analysis study, we aimed to explore how structural and functional 191 impairment of DD converge or diverge and whether this pattern is similar or 192 different across writing systems. We expected to find brain regions that show 193 decreased brain structure and function, indicating insufficient neuronal resources for 194 certain cognitive computations. For regions that show increased brain structure and 195 function, we believe they develop to an unusually high degree for compensation. 196 For brain regions with increased structure but decreased function or decreased 197 structure and increased function, it may be due to brain structures receiving 198 inhibitory input from other regions. We also expected to find language-universal as 199 well as language-specific neurological abnormalities. For language-universal 200 deficits, we tend to believe that they are related to the cause of DD, while the 201 language-specific deficits tend to be consequences of DD in different languages.

203
Literature retrieval and data extraction 204 We searched in "PubMed" (http://www.pubmed.org) and "Web of science" 205 using a combination of "dyslexia", "reading disorder", "reading impairment" or 206 "reading disability" and "fMRI", "PET", "voxel-based morphometry", "VBM" or 207 "neuroimaging" as key words for neuroimaging studies published from January 1986 208 to January 2020. Additionally, we manually added studies by checking the references 209 of the selected papers that were missed in the search. The inclusion criteria were: (1) 210 PET, fMRI, voxel-based morphometry (VBM) studies or structural studies using a 211 volumetric FreeSurfer pipeline, (2) whole-brain results were reported, (3) direct 212 group comparisons between readers with DD and age control readers were reported,

213
(4) coordinates were reported in Talairach or MNI stereotactic space, and (5) studies 214 on DD in the first language. The exclusion criteria were (1) studies with only ROI 215 analysis, (2) resting-state studies, (3) studies that only included readers with DD or 216 did not report group differences, (4) studies with direct group comparisons only 217 between readers with DD and reading level control readers, (5) studies on children at   (from 23 papers) (see Table 1, Table 2 and Figure 1 for detail). From the original 225 publications, we extracted peak coordinates, where there is a significant difference 226 between controls and individuals with DD either in brain activation or regional 227 GMV. We also extracted effect sizes and other information from the publications.

228
Voxel-wise meta-analysis 229 After data acquisition, we conducted a voxel-wise meta-analysis using the  To test the stability of the meta-analysis results, we conducted a whole-brain 248 jack-knife sensitivity analysis. The standard meta-analysis was repeated n times 249 (n=92 for functional studies, n=27 for structural studies) but leaving out one study 250 each time, to determine whether the results remained significant.

251
Multimodal meta-analysis 252 Because we were interested in the convergence between functional deficits and 253 structural deficits, a multimodal meta-analysis was conducted, which provided an  Subgroup meta-analysis 261 In order to explore the language effect, we conducted a subgroup meta-262 analysis. According to the native language of the participants, the functional and 263 structural studies were subdivided into an alphabetic language group in which 264 writing symbols represent phonemes, and a morpho-syllabic language group in 265 which each writing symbol represents a morpheme with a syllable, which resulted in 266 79 functional and 21 structural studies for alphabetic languages and 12 functional 267 (including a Japanese study focused on Kanji) and 6 structural studies for morpho-268 syllabic languages. First, a standard functional and structural meta-analysis was 269 conducted for alphabetic languages and morpho-syllabic languages, separately. Then 270 multimodal meta-analysis was conducted to find the convergence between functional 271 and structural deficits separately for each language group. The threshold for the 272 separate structural and functional maps and multimodal analysis was the same as 273 mentioned above. Next, we conducted a direct comparison between the alphabetic 274 language and morpho-syllabic language group for functional studies and structural 275 studies separately to find the difference between the two language groups. The  Confirmation study 279 Because there were many more functional studies included in the alphabetic 280 group than in the morpho-syllabic group, and most of the morpho-syllabic studies 281 focused on children, the language difference may be due to these differences in the 282 two groups of studies. To avoid influences of these factors, we selected ten English

301
Functional impairment associated with DD across all languages 302 In the meta-analysis of all functional studies, hypoactivation in DD was found 303 in a large cluster peaked at the left IPL which extended to the inferior frontal cortex, 304 occipitotemporal cortex and cerebellum, and a cluster peaked at the right MOG 305 (Table S1and Figure S1). Hyperactivation in DD was found in the right cerebellum, 306 right precentral/postcentral gyrus and bilateral caudate nucleus. The jack-knife 307 sensitivity analysis showed that all results reported above were replicable (Table S1).

308
Structural impairment associated with DD across all languages 309 In the meta-analysis of all structural studies, readers with DD showed a 310 decrease in GMV in the left inferior frontal cortex and right STG (Table S2 and 311 Figure S2). In contrast, readers with DD showed an increase in GMV in the right 312 MTG and left IPL. The jack-knife sensitivity analysis showed that all results 313 reported were replicable (Table S2).

314
Multimodal analysis across all languages 315 As shown in Table 3 and Figure 2, decreased GMV and hypoactivation in DD

327
We also found brain regions that had normal structure but altered function. We 328 found normal brain structure but reduced brain activation at the right MOG and a 329 large cluster peaked at left supramarginal gyrus, which extended to the left ITG, 330 MTG and IFG. We found normal brain structure but increased activation at the right 331 cerebellum and right precentral gyrus ( Figure S7 and Table S7). There are also 332 regions that had altered structure but normal function. We found normal brain 333 activation but reduced GMV at the left IFG and right STG, as well as normal brain 334 function but increased GMV at the right MTG ( Figure S8 and Table S8).    We also found greater increase of GMV in DD in morpho-syllabic languages than in 359 alphabetic languages in the right STG and left ITG (Table 5, Figure 3). We found no 360 regions that showed greater GMV changes in alphabetic languages than in morpho-361 syllabic languages. In this meta-analysis study, we examined the relationship between structural 417 and functional deficits associated with DD as well as whether the deficits are 418 consistent across languages. We found that readers with DD showed both GMV which is consistent with our finding. Taken together, the lack of structural deficits 479 with only hypoactivation at the OT area appears to suggest that the visuo-480 orthographic deficits at the OT might be a consequence of being DD. In contrast, the 481 left STG which was discussed above, appears to be associated with the cause of DD.

482
Our results provide further support for the phonological deficit hypothesis that

501
Our study adds to the literature that not only the brain activation but also the GMV is Chinese and English showed similar brain activation in these two regions. Therefore, 515 readers with DD fail to show language specialization due to their limited reading 516 experience and skills. In summary, this language-specific deficit is believed to be a 517 consequence of being DD in learning morpho-syllabic languages, indicating their 518 inability to accommodate to their own writing system.

519
In the direct comparison between alphabetic and morpho-syllabic languages, 520 we also found greater hypoactivation in DD in alphabetic languages than in morpho- In the multi-modal meta-analysis, we found increased GMV and 554 hyperactivation in participants with DD in the right MTG which was mainly driven 555 by the morpho-syllabic languages. For the functional studies, we also found greater 556 hyperactivation in the right precentral gyrus in morpho-syllabic languages than in 557 alphabetic languages. We believe that these findings are related to the compensation 558 mechanism of the right hemisphere. As precentral gyri play an important role in  The cerebella 615 Increased GMV and hypoactivation in DD were also found in the left 616 cerebellum; however, in the right cerebellum, for the alphabetic language group, we 617 found decreased GMV and hyperactivation. Previously, it was found that the right   The caudate 656 We also found decreased GMV and hyperactivation in readers with DD in the  In addition, there are also brain regions that only showed alteration in structure 679 but not in function or vice versa. It suggests that brain activation is only partially                                        Figure S1. Functional deficits in individuals with DD across all languages (red-yellow: decreased activation in individuals with DD than in controls, blue-green: increased activation in individuals with DD than in controls). Figure S2. Structural deficits in individuals with DD across all languages (red-yellow: decreased GMV in individuals with DD than in controls, blue-green: increased GMV in individuals with DD than in controls). Figure S3. Functional deficits in individuals with DD in alphabetic languages (red-yellow: decreased activation in individuals with DD than in controls, blue-green: increased activation in individuals with DD than in controls). Figure S4. Structural deficits in individuals with DD in alphabetic languages (red-yellow: decreased GMV in individuals with DD than in controls, blue-green: increased GMV in individuals with DD than in controls). Figure S5. Functional deficits in individuals with DD in morpho-syllabic languages (red-yellow: decreased activation in individuals with DD than in controls, blue-green: increased activation in individuals with DD than in controls). Figure S6. Structural deficits in individuals with DD in morpho-syllabic languages (red-yellow: decreased GMV in individuals with DD than in controls, blue-green: increased GMV in individuals with DD than in controls).    Figure S9A. Direct comparison between alphabetic languages and morpho-syllabic languages. Figure  S9B. Confirmation study results when matched language, age and number of studies. Figure S9C. Overlap of the two difference maps. (red-yellow: greater decreases in DD in alphabetic languages/English than in morpho-syllabic languages/Chinese; blue-green: greater decreases in DD in morpho-syllabic languages/Chinese than in alphabetic languages/English; purple: greater increases in DD in morpho-syllabic languages/Chinese than in alphabetic languages/English; cyan: greater increases in DD in alphabetic languages/English than in morpho-syllabic languages/Chinese).