PROGRESS TOWARDS A CELLULAR NEUROBIOLOGY OF READING DISABILITY

Abstract

Reading Disability (RD) is a significant impairment in reading accuracy, speed and/or comprehension despite adequate intelligence and educational opportunity. RD affects 5–12% of readers, has a well-established genetic risk, and is of unknown neurobiological cause or causes. In this review we discuss recent findings that revealed neuroanatomic anomalies in RD, studies that identified 3 candidate genes (KIAA0319, DYX1C1, and DCDC2), and compelling evidence that potentially link the function of candidate genes to the neuroanatomic anomalies. A hypothesis has emerged in which impaired neuronal migration is a cellular neurobiological antecedent to RD. We critically evaluate the evidence for this hypothesis, highlight missing evidence, and outline future research efforts that will be required to develop a more complete cellular neurobiology of RD.

INTRODUCTION

Reading Disability (RD) involves significant impairment of reading accuracy, speed and/or comprehension despite adequate intelligence and educational background (Katzir et al., 2006). Dyslexia presents with similar cognitive, neuroanatomical and genetic traits despite additional spelling and writing impairments associated with the disorder, therefore for the purpose of this review dyslexia is considered synonymous to RD. RD,, is a phenotypically complex developmental disorder with a significant genetic component. As the most common learning disability (Lerner JW, 1989), affecting 5–12% of school aged children, RD has far-reaching social and economic consequences. Several cognitive and perceptual changes appear to associate with RD including changes in short-term memory (Kibby MY et al., 2004; Swanson HL et al., 2006), occulomotor skills(Frith C and U Frith, 1996; Rayner K, 1998; Swanson HL et al., 2006), visuo-spatial abilities(Facoetti A et al., 2009); sensory processing(Tallal P, 1980; Tallal P et al., 1993; Tallal P et al., 1980; Wright CM and EG Conlon, 2009), semantic encoding (Booth JR et al., 2007), integration of letter and speech sounds(Blau V et al., 2009), and phonological processing (Habib M, 2000; Ramus F et al., 2003; Shaywitz SE et al., 1998). It remains controversial as to whether all of these features are central to the core RD phenotype, but the behavioral findings do suggest that any underlying cellular neurobiological cause of RD should have the capacity to affect multiple neural systems to varying degrees.

Over the past several years, increased evidence for neuroanatomic changes in RD, identification of candidate genes, and elucidation of the functions of three candidate genes (KIAA0319, DCDC2 and DYX1C1) in neuronal migration have strengthened a hypothesis which states that impaired neuronal migration in development causes a predisposition to RD. We review the evidence for this hypothesis and highlight both the missing pieces and future research efforts that will be needed for a more complete cellular understanding of RD.

NEUROANATOMY OF RD

Neuroanatomical studies have revealed neurostructural correlates of RD. Postmortem studies were the first to support an association between cortical migration anomalies and RD, and animal models have supported a causal connection between cortical migrational anomalies and specific deficits in perception and learning. More recent imaging studies have found evidence of changes in grey and white matter that correlate with RD. Evidence, however, is still lacking from the absence of a large scale anatomical study, and it remains unclear whether the identified neuronal migration anomalies in postmortem studies are directly related to changes in neocortical structure or function that have been revealed in MRI studies.

Ectopia

Postmortem neuroanatomical examination of a relatively small sample of individuals with RD and controls revealed increases in the incidence of a focal neocortical abnormality known as neocortical “ectopia” in RD brains (Galaburda AM, 1988; Galaburda AM and TL Kemper, 1979; Galaburda AM et al., 1985). These layer 1 ectopia are generally too small for detection by MRI, and are present in most control brains studied although at a much lower rate of occurrence than in RD brains. Ectopia occur as a result of disrupted migration caused by either abnormal interactions between migrating neuroblasts and radial glial fibers and/or disruptions in the pia and layer I (Caviness VS, Jr. et al., 1978; McBride MC and TL Kemper, 1982). The discovery of increased ectopia occurrence in RD was the first finding to suggest a connection between neuronal migration in the neocortex and RD.

Periventricular nodular heterotopia

Periventricular nodular heterotopia (PNH) are characterized by clusters of immature neurons partially embedded within the white matter near the surface of the lateral ventricles. This rare condition is most often caused by mutations in the x-linked gene filamin (Barkovich AJ and RI Kuzniecky, 2000; Dubeau F et al., 1995; Raymond AA et al., 1994), and loss of filamin function results in a failure of initial neuronal migration from the precursor population that lines the ventricles in fetal development. Studies combining in vivo imaging and behavioral assessments have shown an association between this cortical malformation and RD (Chang BS et al., 2007; Chang BS et al., 2005). More specifically, PNH patients show processing deficits in real-word and non-word reading tasks (Chang BS et al., 2007). Moreover, the reading deficits in PNH patients were independent of intelligence, and the severity of reading disability amongst this group was found to correlate strongly with the amount of white matter disruption proximal to the PNH malformations (Chang BS et al., 2007). The white matter changes found in studies relating PNH to RD are consistent with a growing body of evidence for changes in white matter tracts reported in several studies comparing RD and control cases with diffusion tensor imaging (DTI) (Beaulieu C et al., 2005; Deutsch GK et al., 2005; Klingberg T et al., 2000; Niogi SN and BD McCandliss, 2006; Odegard TN et al., 2009; Steinbrink C et al., 2008).

Malformation models and behavioral disruption

Animal models of the malformations have been used to test whether focal neocortical abnormalities can create behavioral and sensory deficits similar to some non-language based deficits correlated with RD. Ectopias virtually identical to those described in humans have been found in three strains of autoimmune mice (i.e. NZB/BlNJ, BXSB/MPJ and NXSM-D/Ei). Mice with neocortical ectopias are impaired in spatial and non-spatial working memory (Balogh SA et al., 1998; Boehm GW et al., 1996; Denenberg VH et al., 2001; Hoplight BJ et al., 2001; Hyde LA et al., 2002), and in processing rapid auditory stimuli (Clark MG et al., 2000; Frenkel M et al., 2000; Peiffer AM et al., 2001). Similarly, male rats with induced-microgyria in parietal cortex, a disruption in cortical lamination with similarities to and often associated with ectopias, display rapid auditory processing deficits (Clark MG et al., 2000; Fitch RH, SW Threlkeld et al., 2008; Herman AE et al., 1997) and working memory deficits (Fitch RH, H Breslawski et al., 2008). In all, the animal studies indicate importantly that even relatively small malformations in neocortical structure can have very specific effects on sensory and learning tasks without having large scale effects on general learning ability.

GENETICS OF RD

The current status of RD genetic association efforts, reviewed more thoroughly elsewhere (Gibson CJ and JR Gruen, 2008; Paracchini S et al., 2007), underscores the complexity of RD genetics. The composite evidence clearly shows a strong genetic risk; however, even the most consistently associated genetic loci and genes have not been significantly associated with RD in all sample populations. Moreover, there is no reported extended pedigree or large multigenerational family showing a specific mutation in the coding region of one of the three most replicated candidate genes. Similarly, RD risk haplotypes in all candidate genes to date are found in a significant percentage of individuals without RD. In spite of these caveats, the identification of three candidate genes (KIAA0319, DYX1C1 and DCDC2) has ushered in a new era of RD research that can begin to focus on molecular and cellular mechanisms.

A brief history of RD genetics

Familial cases of RD was noted as early as 1896 by W.P. Morgan, and since then the overall risk of RD within a family with an affected individual has been estimated at between 34% and 48% (Finucci JM et al., 1976; Gilger JW et al., 1991; Hallgren B, 1950; Klasen, 1968; Zahalkova M et al., 1972). A strong genetic component to this familial association was demonstrated by twin studies. The first twin studies carried out in the 1950s showed a monozygotic concordance for RD of 100% (Hallgren B, 1950; Hermann K, 1956; Hermann K and E Norrie, 1958), however this was revised in later studies (Bakwin H, 1973; Decker SN and SG Vandenberg, 1985; Stevenson J et al., 1987), to a heritability between 29% and 82% (Alarcón M and JC DeFries, 1997; DeFries JC et al., 1987; DeFries JC and JJ Gillis, 1991; Harlaar N et al., 2005; Hawke JL et al., 2006; Hohnen B and J Stevenson, 1999; LaBuda MC and JC DeFries, 1988; Pennington BF and JW Gilger, 1996). The large range of estimated heritability likely indicates that the genetic risk for RD is complex and modifiable by a variety of environmental influences operating within different sample populations.

Genetic linkage disequilibrium analyses have now led to the identification of nine chromosomal loci across the genome that significantly associate with RD risk: DYX1-DYX9. Of these nine loci those located on chromosome 1p34-p36 (DYX8), 2p (DYX3), 6p21.3 (DYX2), and 15q21 (DYX1) have been frequently replicated, while those located at 3p12-q12 (DYX5), 6q13-q16 (DYX4), 11p15 (DYX7), 18p11 (DYX6), and Xq27 (DYX9) have been replicated once or not at all. Candidate susceptibility genes for DYX1 and DYX2, as well as other candidate genes are discussed below (see Table 1 for summary of recent findings).

Table 1.

	Significant association	N	Nationality	No significant association	N	Nationality
DYX1C1	Taipale et al (2003)*	109	FIN	Bellini G et al (2005)*	57	IT
	Wigg et al. (2004)**	148	CAN	Meng et al. (2005)**	150	US
	Brkanac Z et al (2007)**	191	US	Scerri et al (2004)**	264	UK
	Dahdouh F. et al (2009)**	366	GDR	Marino et al (2004)**	158	IT
KIAA0319	Francks et al (2004)**	223	UK	Brkanac Z et al (2007)**	191	US
	Cope et al. (2004)**	223	UK
	Harold et al (2006)**	148	UK
DCDC2	Meng et al (2005)**	150	US	Brkanac Z et al (2007)**	191	US
	Schumacher et al (2006)**	239	GDR
	Wilcke et al. (2009)*	72	GDR
	Ludwig et al (2008)**	396	GDR

^*Case control linkage disequillibrium

^**Family based transmission disequillibrium test

Candidate gene for DYX1: DYX1C1

DYX1C1 was the first gene to be linked to RD when it was reported in 2000 that a chromosomal translocation involving 15q in two Finnish families with a history of RD caused a breakpoint within the DYX1 locus (Nopola-Hemmi J et al., 2000; Nothen MM et al., 1999; Schulte-Korne G et al., 1998; Smith SD et al., 1983). Closer examination of the breakpoint showed that it disrupted exons of the gene EKN1, which was subsequently renamed Dyslexia Susceptibility 1 Candidate 1 (DYX1C1). Although this initial study showed a significant association in two relatively small Finnish samples (Taipale M et al., 2003), several subsequent studies in populations in the UK, US, and Italy did not find a significant association (Cope NA et al., 2005; Marino C et al., 2005; Meng H, K Hager et al., 2005). More recent studies of a German and US sample, however, have shown that one of the RD-related single nucleotide polymorphisms (SNPs) in DYX1C1 does associate with RD (Brkanac Z et al., 2007; Dahdouh F et al., 2009). In addition, a study of the same Italian sample in which a family based association between DYX1C1 risk alleles and RD was not found, did find a significant association with verbal short term memory (Marino C et al., 2007). Thus, DYX1C1 variants can associate with reading impairment in some population samples, and with component features often associated with RD in other samples (Marino C et al., 2007). Finally, the DYX1C1 risk allele associated with RD appears to be functional in that there is a change in the 5′ promoter region that affects DNA interaction with a complex of proteins (TFII-I, PARP1, and SFPQ) that regulate gene expression (Tapia-Paez I et al., 2008).

Candidate genes forDYX2: KIAA0319 and DCDC2

DYX2, on chromosome 6p, is the most replicated of DYX loci, and has been linked with both global and component RD phenotypes, particularly orthographic sub-phenotypes (Cardon LR et al., 1994; Cardon LR et al., 1995; Fisher SE et al., 1999; Grigorenko EL et al., 2003; Grigorenko EL et al., 1997; Kaplan D et al., 2002; Smith SD et al., 1991). Two peaks of genetic association have been identified within DYX2 that include two candidate genes, KIAA0319 and DCDC2 (Kaplan DE et al., 2002; Meng H, K Hager et al., 2005).

In 2002, Kaplan et al. showed a peak of association at a marker in the 5′ untranslated region of KIAA0319 (Kaplan DE et al., 2002). In 2004 a study by Francks et al. showed a peak of association in a 77 kb region containing the first four exons of KIAA0319, and this was replicated by Cope et al. in 2005 using a dense set of SNPs to further identify a risk haplotype in the same region (Cope N et al., 2005). The risk haplotype was later shown to be related to a selective decrease in the expression of KIAA0319 but not other genes in the locus (Harold D et al., 2006). The risk haplotype of KIAA0319 that includes the promoter region has more recently been shown to confer reduced promoter activity and an aberrant binding site for the transcriptional silencer OCT-1 (Dennis MY et al., 2009).

Meng et al. (2005) identified a deletion and compound short tandem repeat (STR) in intron 2 of DCDC2, a gene located 500 kb from KIAA0319. The STR in DCDC2 showed a significant association with RD in a cohort of 153 American dyslexic families (Meng H, K Hager et al., 2005), and this association was independently confirmed in a German population (Schumacher J et al., 2006). In this same association study in the German sample a haplotype within DCDC2 was identifiedthat showed strength of association directly proportional to the severity of RD (Schumacher J et al., 2006). More recently, the association between DCDC2 and RD has been independently confirmed in an Italian cohort, and the risk haplotype of DCDC2 associated with a decrease in enhancer activity (Meng et al. un published).

Other candidates genes

In addition to the three genes discussed above, candidate genes have also been identified in several other studies (Anthoni H et al., 2007; Hannula-Jouppi K et al., 2005; Poelmans G et al., 2009); however, the genes identified in these studies have not yet been reported to associate with RD in larger populations. In 2005, FISH analysis of a translocation (t(3;8)(p12;q11)) in a Finnish RD individual revealed disruption between exons 1 and 2 of ROBO1 on 3p (Hannula-Jouppi K et al., 2005). This gene lies within DYX5. Further linkage and association analysis of the proband’s extended family revealed a SNP haplotype spanning the gene that showed marginal association with RD. ROBO1 and its ligand SLIT are well known to play roles in axonal targeting and also in cell migration. As such, ROBO1 is a compelling developmental candidate gene that may have direct effects on the development of axonal connections.

FUNCTIONS OF CANDIDATE GENES

The identification of candidate genes DYX1C1, KIAA0319 and DCDC2, presented a new opportunity to begin to test hypotheses with respect to potential cellular causes of RD. The association of RD with neuronal migration impairment discussed above led to a series of experiments to determine whether these candidate genes play a role in neuronal migration, and if so, to determine the types of malformations created by decreased expression of these genes (Burbridge TJ et al., 2008; Meng H, SD Smith et al., 2005; Paracchini S et al., 2006; Rosen GD et al., 2007; Threlkeld SW et al., 2007; Wang Y et al., 2006). Figure 1 shows a summary of the results from these experiments. In this section we discuss the molecular features of three candidate genes as well as the studies that demonstrated their role in neuronal migration in developing neocortex. One limitation of the neuronal migration studies carried out so far is that they have directly tested for an involvement in migration in neocortex, and were not designed to test for changes in development of other structures or other developmental processes. Mouse knockout experiments are currently underway to test for a more general developmental role of these genes. The current evidence for a role in migration in neocortex should therefore not be viewed as evidence for a single specific function of the candidate genes in neuronal migration.

Figure 1.

Summary of RNAi studies demonstrating a role for Kiaa0319, Dcdc2, and Dyx1c1 in neuronal migration in the developing neocortex. A) Example images of eGFP (green) labeled neurons in a patch of embryonic neocortex 4 days after four different manipulations. “Control” shows the normal dispersion pattern of neurons after they have migrated.” Kiaa0319”, “Dcdc2”, and “Dyx1c1” panels show the effects on cell dispersion and migration following RNAi knockdown of the indicated candidate gene. Note that in each condition the vast majority of cells fail to disperse by migration, and that each RNAi condition creates a somewhat distinct pattern. B) A summary of RNAi results for both short-term and long-term effects of RNAi targeted against candidate genes (Burbridge TJ et al., 2008; Meng H, SD Smith et al., 2005; Paracchini S et al., 2006; Rosen GD et al., 2007; Wang Y et al., 2006). After 4 days of migration cells targeted with RNAi are largely stalled, but when examined in the adult brain the RNAi treatments caused a final phenotype that included normally positioned neurons, a mixture of heterotopia and ectopia, as well as scattered neuronal displacement. C) Example of three types of malformations resulting from Dyx1c1 RNAi and examined in the mature rat brain: ectopia (small arrows), white matter heterotopia (open arrows) and hipocampal dysplasia (arrow points). In the lower panel transfected cells are labeled brown (Rosen GD et al., 2007).

DYX1C1

The protein domains of DYX1C1 include an N-terminal p23 and three C-terminal tetratricopeptide repeat (TPR) domains. The N-terminal p23 domain of DYX1C1 protein, when overexpressed in cell lines, can interact with Hsp70, Hsp90 and an E-3 ubiquitin ligase, CHIP, suggesting that the protein may be involved in degradation of unfolded proteins (Hatakeyama S et al., 2004). Recently, DYX1C1 has been shown to be involved in the degradation of the estrogen receptor potentially through its interaction with CHIP (Massinen S et al., 2009).

In vivo RNAi studies indicate that Dyx1c1 plays a role in neuronal migration in developing neocortex. Soon after transfection of a cohort of newly produced neurons with plasmids that induce RNAi or knockdown of Dyx1c1 expression, neurons become arrested in their normal migration path through the intermediate zone (Wang Y et al., 2006). Wang et al. 2006, went on to show that the TPR domains of Dyx1c1 were critical to migration in that mutations missing the TPR domains failed to rescue migration while expression of the TPR domains alone was sufficient to restore normal migration. Moreover, a 3 amino acid deletion in the last TPR domain which results from a SNP that was initially shown to associate with RD, but later not replicated in any other population, was found to be dispensable for Dyx1c1 function in neuronal migration in neocortex (Wang Y et al., 2006).

In a follow-up study to the initial RNAi study, Rosen and colleagues (Rosen et al. 2007) examined what the final malformation profile would be for brains in which Dyx1c1 was knocked down. Although the embryo knockdown created a nearly uniform arrest in migration (Figure 1A), most neurons restarted their migration and attained position similar to control treated neurons by juvenile postnatal ages (Figure 1B). However disruption of the laminar organization of the cortex was still evident in the Dyx1c1 knockdown condition compared to RNAi control group treated at same developmental stage. In addition, distinct malformation types were found to occur with some variety in different animals. These malformations included heterotopia in white mater, ectopia in layer one of neocortex, and hippocampal heterotopia with dysplasia. Behavioral assays of animals with Dyx1c1 RNAi induced malformations have also shown impairments in auditory processing and maze learning, and the changes in learning were correlated with the presence of the hippocampal heterotopia (Threlkeld SW et al., 2007). Co-occurrence of deep malformations such as white matter heterotopia with superficial malformations such as ectopia and microgyria have also been reported in certain human malformation syndromes (Wieck G et al., 2005). Overall the experiments with Dyx1c1 knockdown have shown a surprising agreement with the neuronal migration hypothesis of RD. In particular the occurrence of diverse malformation types, including importantly ectopia (Figure 1C), and complex behavioral outcomes following a genetic disruption, are reminiscent of the RD phenotype.

KIAA0319

The KIAA0319 gene encodes an integral membrane protein with a large extracellular domain, a single transmembrane domain, and a small intracellular C-terminus. There are several splice variants of KIAA0319 (Velayos-Baeza A et al., 2007), all of which are glycosylated, and one form is secreted (Velayos-Baeza A et al., 2008). The extracellular domain is characterized by a consensus signal peptide, and 5 PKD domains. PKD domains in the polycytsin 1 protein have been shown to be involved in adhesion between kidney cells and so has been suggested as a cell adhesion domain (Silberberg M et al., 2005). To date, there is only one defined protein interactor of KIAA0319 protein, adaptor protein-2 (AP-2), which is part of the endosomal pathway (Levecque C et al., 2009). The structure, membrane localization, and emerging cell biology of KIAA0319 protein are consistent with it being in a relatively unstudied new class of neural cell adhesion molecule.

KIAA0319 expression was targeted with RNAi in migrating neocortical neurons to test for a potential role in migration. Similar to Dyx1c1, RNAi knockdown of Kiaa0319 interrupted migration 4 days after transfection (Figure 1) (Paracchini S et al., 2006). Kiaa0319 knockdown, in contrast to knockdown of Dcdc2 and Kiaa0319, created a distinct cellular phenotype. Namely, disrupted neurons appeared to loose their normal radial association with radial glial fibers and migrating neurons were often found orthogonal to the radial glia scaffold that they typically migrate along (Paracchini S et al., 2006). This phenotype, the emerging cell biology, and molecular structure of Kiaa0319 protein, suggest a possible role in neuron to radial glia adhesion; however, additional direct experimentation will be required to prove this role.

DCDC2

DCDC2 is one of an eleven-member group of proteins distinguished by the presence of dcx or doublecortin domains (Coquelle FM et al., 2006; Reiner O et al., 2006). The first characterized gene of this family, DCX, was identified by the discovery that mutations in DCX cause double cortex syndrome in females and lissencephaly in males (Barkovich AJ and RI Kuzniecky, 2000; Gleeson JG et al., 1999). The dcx domain is critical for binding to and stabilizing microtubules and is regulated by phosphorylation (Gleeson JG et al., 1999). At least two members of this family, Dcx and Dclk, have now been found to interact genetically in mice in terms of growth of axons across the corpus callosum and in neuronal migration in cerebral cortex (Deuel TA et al., 2006; Koizumi H et al., 2006). In a study comparing the biochemical and cellular functions of proteins in the Dcx family it was found that Dcdc2 exhibits the same functional features displayed by Dclk and Dcx protein (Coquelle FM et al., 2006).

Based on the similarity of structure between Dcdc2 and Dcx proteins Meng et al. (2005) hypothesized that DCDC2 may play a role in neuronal migration. Using an RNAi approach targeting Dcdc2 expression in migrating neocortical neurons in the embryonic rat neocortex it was found that knockdown of Dcdc2 interrupted neuronal migration (Meng H, SD Smith et al., 2005) (Figure 1). More recently, Burbridge et al. (2008) has shown that the migration disruptions caused by knockdown of Dcdc2 results in diverse disruptions similar to but not identical to those created by Dyx1c1 knockdown (Burbridge TJ et al., 2008). Knockdown of Dcdc2 creates both scattered heterotopia within the white matter similar to PNH, and also causes a population of neurons to over-migrate to ectopic positions in neocortex, although they do not form ectopia. The over-migration may be a secondary effect of migration delay because unlike the PNH “add back” or rescue experiments in which Dcdc2 is re-expressed resolved the PNH malformations, but did not eliminate the over-migration.

Conclusions

Accumulating evidence is highly suggestive of a connection between neuronal migration disruption and genetic susceptibility to RD. It would be premature however at this point to conclude that RD is a disorder of neuronal migration. The changes in migration may be correlates of another function of these initially defined candidate genes, and the majority of the genetic risk for RD is not carried by KIAA0319, DCDC2, or DYX1C1. In addition, other potential functions of these three genes have not yet been thoroughly tested in genetic knockout experiments. In fact, all three of these candidate genes are expressed in mature neurons after migration, and may therefore have important functions in such processes as synaptic plasticity that would also affect learning. Enhanced genetic techniques and larger genome wide studies will be needed in the future to identify a larger fraction of the genetic risk and to determine more definitively the contribution of KIAA0319, DCDC2, and DYX1C1 risk alleles. Such studies will likely include association of copy number variations (CNV) in RD as has recently been shown successful in identifying candidate genes for autism and schizophrenia (Cantor RM and DH Geschwind, 2008; Glessner JT et al., 2009; Sebat J et al., 2007). If RD does have a common underlying cause of neuronal migration impairment, then genes identified within such expanded searches would be expected to be genes that code for proteins involved in neuronal migration.

Enhanced structural and functional imaging studies are needed to more clearly define the set of anatomical changes that associate most frequently with RD. Future studies should also include correlation between imaging and genetics. In particular it would be important to determine whether the candidate gene alleles correspond to specific morphological alterations. A recent preliminary study indicates a potentially promising beginning to this approach (Meda SA et al., 2008).

Finally, the identification of a common neurodevelopmental disruption would only be the beginning of a mechanistic understanding of RD. Functional neurophysiological studies will be needed to connect any developmental disruption in neuronal positioning to changes in connectivity or function within cortical circuits. Work with animal genetic models will most likely be required to define in detail the specific cellular neurophysiological disruptions following interruption of RD related genes. There are already very effective educational interventions for RD that can greatly improve reading ability. Understanding the cellular basis of RD in detail in animal models could help to define critical periods of intervention, and may identify physiological mechanisms or changes that could be targeted pharmacologically to enhance early education-based remediation of RD.