Dendritic pathology in mental retardation: from molecular genetics to neurobiology


  • M. Dierssen,

    Corresponding author
    1. Neurobehavioral Analysis Laboratory, Genes and Disease Program, Center for Genomic Regulation, (CRG-UPF) PRBB, 08003 Barcelona, Spain,
    Search for more papers by this author
  • G. J. A. Ramakers

    1. Neurons and Networks, Netherlands Institute for Brain Research, Graduate School Neurosciences Amsterdam, Meibergdreef 33, 1105 AZ Amsterdam ZO, the Netherlands
    Search for more papers by this author

*M. Dierssen, Genes and Disease Program, Center for Genomic Regulation, Passeig Marítim 37-49, 08003 Barcelona, Spain. E-mail: mara.dierssen@crg.es


Mental retardation (MR) is a developmental brain disorder characterized by impaired cognitive performance and adaptive skills that affects 1–2% of the population. During the last decade, a large number of genes have been cloned that cause MR upon mutation in humans. The causal role of these genes provides an excellent starting point to investigate the cellular, neurobiological and behavioral alterations and mechanisms responsible for the cognitive impairment in mentally retarded persons. However, studies on Down syndrome (DS) reveal that overexpression of a cluster of genes and various forms of MR that are caused by single-gene mutations, such as fragile X (FraX), Rett, Coffin-Lowry, Rubinstein–Taybi syndrome and non-syndromic forms of MR, causes similar phenotypes. In spite of the many differences in the manifestation of these forms of MR, evidence converges on the proposal that MR is primarily due to deficiencies in neuronal network connectivity in the major cognitive centers in the brain, which secondarily results in impaired information processing. Although MR has been largely regarded as a brain disorder that cannot be cured, our increased understanding of the abnormalities and mechanisms underlying MR may provide an avenue for the development of therapies for MR. In this review, we discuss the neurobiology underlying MR, with a focus on FraX and DS

Mental retardation: multiple causes but common mechanisms

Mental retardation (MR) is a developmental disability characterized by significant impairment of intellectual functioning and adaptive skills MR has an onset during childhood and ranks first among chronic conditions that cause major limitations to live a normal, independent life (Pope & Tarlov 1991). Its occurrence in a child puts a heavy burden on the affected families. In view of the high prevalence of MR, there is a high demand on the educational and health care system, resulting in a considerable economic cost. In practice, the limitations in intellectual functioning are defined by an intelligence coefficient (IQ) below 70 [two standard deviations below the average IQ of 100], which corresponds to about 2.2% of the general population. In USA, an average prevalence of 1.14% was found among children between 6 and 17 years of age (Massey & McDermott 1995).

Mental retardation is a highly diverse disorder in terms of the severity of the cognitive disability, as well as the manifestation of additional (non-cognitive) symptoms, which can be partly related to the heterogeneity in the underlying causes. With regard to the severity of the cognitive disability, MR is commonly divided into IQ ranges between 70 and 55 (mild MR), 55 and 40 (moderate MR), 40 and 25 (severe MR) and below 25 (profound MR). In addition, a borderline group is distinguished with IQs between 85 and 70. Mental retardation is often associated with other clinical symptoms, which has led to the distinction of syndromic and non-syndromic forms of MR. In syndromic MR, the cognitive disability is associated with a fixed constellation of other manifestations, such as body and brain malformations, neurological or psychiatric symptoms or metabolic defects. In non-syndromic MR, no additional abnormalities are observed (including alterations in brain anatomy), and a subnormal intelligence is thus the only detectable deficit. The division between syndromic and non-syndromic forms of MR has become blurred, as (different) mutations in the same gene can produce both syndromic and non-syndromic MR (Frints et al. 2002).

Mental retardation can be caused by both environmental and genetic factors (Chiurazzi & Oostra 2000). In about 30–50% of the cases, the cause is not known (Curry et al. 1997). Genetic causes of MR include cytogenetically detectable chromosome abnormalities (duplications, translocations, deletions and full and partial trisomies), single-gene mutations and presumed combinations of polymorphisms in several genes. The most common genetic cause of MR is trisomy of chromosome 21, which causes Down syndrome (DS), whereas the most frequent single-gene deficit is the fragile X (FraX) syndrome. Environmental factors include prenatal exposure of the fetus to toxic agents, alcohol and drugs, intrauterine infections by cytomegalovirus, rubella, herpes simplex and maternal malnutrition, premature birth, perinatal trauma, hypoxia and hypothyroidy. In addition, there is an important contribution by socio-economic factors, in particular, a low education level of the mother (McDermott 1994).

Mental retardation as a deficit in neuronal network connectivity

Mental retardation is a global concept that can probably be categorized by temporal and spatial alteration of functional domains, not by mutated genes or by phenotypes. These different functional aspects may stem from both abnormal postnatal development and correct remodeling of the brain circuitry. Neural mechanisms underlying MR may include defects in the formation of neuronal networks and/or defects in properties of brain plasticity that are believed to be important for information processing. Neuropathological studies of post-mortem brains of persons with MR have shown that MR is usually associated with detectable alterations in the structure of the cerebral cortex, hippocampus and/or various other brain areas. Severe forms of MR are often associated with gross brain malformations and/or microcephaly. In these and other cases with severe prenatal toxicity, infection or perinatal trauma, the total number of neurons is reduced, because of either impaired production or pathological loss.

Aberrant migration of neurons in the cerebral cortex is also frequently linked with MR. Brain malformation, neuronal loss and migration disorders will obviously result in deficient network connectivity and could thereby impact on information processing. On the other hand, many (often milder) forms of MR (and non-specific MR by definition) show little, if any, consistent changes in brain macroanatomy, including relevant areas like cerebral cortex and hippocampus. In fact, dendritic pathology appears to be a consistent feature of MR across multiple conditions (reviewed in Kaufmann & Moser 2000). Investigation of neocortical and hippocampal microanatomy using Golgi staining has demonstrated that also only neuronal connectivity is often altered.

Huttenlocher (1974), Marin-Padilla (1972) and Purpura (1974) were the first to use the Golgi method to compare dendritic and spine morphology of post-mortem human cerebral cortex neurons in persons with MR and controls. A study by Purpura (1974) described the phenomenon of ‘spine dysgenesis’ in children with non-specific MR, which showed a reduced density of dendritic spines and a predominance of very long, thin spines at the expense of stubby and mushroom spines. As the long, thin spine morphology resembled immature spines, whereas the stubby and mushroom spines are typically adult type spines, the observations indicated a disturbance of the development of dendritic spines. Reductions in dendritic complexity (mainly reflecting the degree of branching) in MR were described by Huttenlocher (1974). Subsequent Golgi studies on pyramidal neurons in the cerebral cortex and hippocampus comparing Down, Rett and FraX syndrome confirmed an association between MR and abnormalities in the morphology and density of dendritic spines, and/or reduced dendritic branching (reviewed in Huttenlocher 1990; Kaufmann & Moser 2000). Whereas, in most cases, a reduction in spine densities and altered spine shapes were observed, FraX syndrome was characterized by an increase in dendritic spines with an immature morphology, suggesting a deficiency in developmental pruning of the spines in FraX syndrome. These Golgi studies suggested that, in many cases, MR is due to abnormal development of connectivity within the cerebral cortex. Similar abnormalities were found in adults with untreated phenylketonuria (Bauman & Kemper 1982) and in infants with severe malnutrition (Benitez-Bribiesca et al. 1999). Animal models in which the effects of environmental causes of MR on brain development were investigated, such as prenatal alcohol exposure (Galofre et al. 1987; Stoltenburg-Didinger & Spohr 1983), hypothyroidism (Thompson & Potter 2000), experimental hyperphenynalaninemia (Hogan & Coleman 1981), fetal hypoxemia (Rees et al. 1999) and protein deprivation (Diaz-Cintra et al. 1990), all displayed reduced dendritic arborization, spine deficits or both. More recently, mouse models for MR in which specific MR genes have been mutated or deleted reveal similar alterations in spines and dendrites as seen in MR in humans, often in conjunction with impaired cognitive behavior.

Taken together, these studies have demonstrated that MR is associated with structural abnormalities in the cognitive centers of the brain, such as a reduction in neuron numbers, disturbed neuronal migration and alterations in dendritic arborization and in spine densities and morphology. These changes may be quantitative and reflected in a suboptimal number of synaptic connections, or qualitative, expressed in an increased number of inappropriate connections. Based on these observations, we proposed a structural network hypothesis that assumes that MR is primarily due to abnormal development and/or plasticity of structural neuronal network connectivity (Ramakers 2000, 2002). As detailed below, the abnormal network connectivity results in deficient information processing within the network, manifested as intellectual discapacity. Additional support for the network hypothesis comes from the finding that an increasing number of genes, that give rise to MR upon mutation, encode proteins that play an important role in structural network development and/or plasticity (Chelly & Mandel 2001; Ramakers 2000, 2002). Thus, different etiological factors converge in a common pathogenetic mechanism affecting the communication systems in the brain. However, dendritic spines are heterogeneous with regard to their structure, stability and function (see Kasai et al. 2003) being critical for the integration of synaptic inputs and for the neuronal ‘computation’. Whereas the growth of dendritic spines initially follows genetic dictates, it later becomes modified by levels and patterns of activity (Bartlett & Banker 1984) in both developing (Kossel et al. 1995; Segal 1995) and mature systems in response to numerous manipulations, including long-term potentiation (LTP) (Buchs & Muller 1996; Sorra & Harris 1998), and environmental stimulation (Comery et al. 1997; Jones et al. 1997). While most spines in adult mouse cortex appear to have a half-life of up to 13 months (Grutzendler et al. 2002), a fair proportion of the spines has lifetimes of less than a day (Trachtenberg et al. 2002). Moreover, the more mature large (stubby) spine has longer lifetime than the typically immature long, thin spine. In spite of their considerable half-life, many spines are motile and can change in shape and size in vivo (Bonhoeffer & Yuste 2002) and in vitro (Fischer et al. 1998). This motility is dependent on the dynamics of the actin cytoskeleton, which, in the adult nervous system, is highly concentrated in spines (Fischer et al. 1998). The motility of spines is furthermore under control of glutamatergic neurotransmission (Fischer et al. 2000) and stimuli that induce LTP promote the formation of novel spines (Engert & Bonhoeffer 1999; Maletic-Savatic et al. 1999; Toni et al. 1999). During development, dendritic protrusions start out as filopodia, which search out contacts with synaptic terminals and then mature into adult spines (Ziv & Smith 1996). In adults, modulation of the number and shape of spines is associated with synaptic plasticity (Engert & Bonhoeffer 1999; Lendvai et al. 2000; Maletic-Savatic et al. 1999; Toni et al. 1999) and learning (Comery et al. 1997).

Changes in synaptic function and remodeling of synaptic networks induced by specific patterns of activity are believed to represent key mechanisms for modulating information processing (Stepanyants et al. 2002). Dendritic spines are the postsynaptic compartment for most excitatory glutamatergic synapses and are investigated as focal points of synaptic plasticity, so that abnormal spines may disrupt the cellular mechanisms underlying memory consolidation, severely affecting learning and memory (Matus 2000). Alterations in the shape, in particular, the length of the spine neck have been shown to be important in the dynamics of calcium mobilization in the spine (Volfovsky et al. 1999), with potential significance for neurotransmission and synaptic plasticity. The size of the spine head, but not the diameter and length of the spine neck, correlates well with the area of the postsynaptic density and the number of vesicles in the presynaptic terminal (Harris & Stevens 1989). Spines with large heads are stable, express large numbers of alpha-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA)-type glutamate receptors and contribute to strong synaptic connections. In contrast, spines with small heads are motile and unstable and contribute to weak or silent synaptic connections. These structure–stability–function relationships suggest that large and small spines are ‘memory spines’ and ‘learning spines’, respectively. Moreover, mushroom spines contain many more AMPA receptors than thin spines or filopodia (Matsuzaki et al. 2001). These observations indicate a clear correlation between the dimensions of the spine head and the synaptic efficacy. Thus, the morphological properties of the spines are critical determinants of neurotransmission, information processing and synaptic plasticity. Changes in spine morphology, as seen in MR, are thus likely to have a strong impact on these functions.

Finally, dendritic branching complexity is crucial to synaptic integration (Hausser et al. 2000; Yuste & Tank 1996). For instance, computational studies of Mainen and Sejnowski (1996) have shown that dendritic geometry in itself is an important determinant of action potential-firing patterns in single neurons, because it determines the nature of a neuron's inputs and its role in cortical circuitry (Bacon & Murphey 1984; Friedlander et al. 1982). What is the functional importance of (modest) alterations in dendritic complexity and/or spine numbers or morphology? Although this area has hardly been investigated experimentally, it is clear that alterations in dendritic structure will have a major impact on the processing of afferent information by single neurons. At the level of the neuronal network, even modest alterations in dendritic structure and organization of many neurons, as seen in MR, will lead to considerable changes in overall information processing.

Genetic causes of MR

Genes causing MR have been cloned at an increasing pace during the last decade. The most recent count (updated regularly at the internet site maintained by P. Chiurazzi: was 14 genes involved in non-syndromic MR and 39 genes involved in syndromic forms of single-gene MR (including metabolic and neuromuscular disorders). This raises the question of how deregulation/dysfunction of a myriad of different molecules may affect the cellular mechanisms underlying neuronal development (neurite outgrowth and neuronal morphogenesis, shaping the three-dimensional architecture of neurons, synapse formation, network formation), and synapse rearrangement, allowing long-term memory formation and adaptation to the environment.

Based on their functions, genes involved in MR can be roughly divided in two groups. One group is clearly involved in the regulation of neuronal morphology and structural connectivity. The core of this group is formed by genes that encode proteins which interact directly with Rho GTPases, such as OPHN1, ARHGEF6 (which encodes αPix), PAK3 and FGD1. Other genes that may belong to this group encode proteins that may act upstream or downstream of Rho GTPase signaling, such as L1CAM, TM4SF2 (tetraspanin), DLG3 (synapse-associated protein 102 or SAP102), neuroligins, LIM kinase 1 and possibly even FMR1. An important function of Rho GTPases is the regulation of neuronal outgrowth, path finding and spine formation in response to many extracellular and intracellular factors (Luo 2000; Ramakers 2002). Rho GTPases integrate these signals in growth cones and spines to subsequently co-ordinate the dynamics of the actin cytoskeleton, which is crucial in neurite path finding and branching. However, multiple molecular cascades are likely to be involved in normal dendritic and spine maturation (see Elston & DeFelipe 2002; Harris 1999; Shepherd 1996; for reviews), and this fact may explain how dysregulation of the expression of many genes, as occurs in multigenic MR disorders such as DS, leads to similar phenotypes.

The other group of genes encodes proteins that are somehow involved in the regulation of gene expression. The functions of these genes provide less insight into the neurobiology of MR than the Rho-associated MR genes, as the question is now shifted to the establishment of the major target gene(s) responsible for the cognitive impairment. This group consists of genes that encode proteins involved in signaling to transcription factors (RSK2), transcription (ARX, FMR2, ZNF41, ZNF81, SOX3), chromatin remodeling (MECP2, ATRX) or regulation of translation (FTSJ1, FMR1).

In DS brains, several candidate genes within the Down syndrome critical region (DSCR) of chromosome 21 are involved in brain development and neural functions related to synaptic plasticity with particular impact on dendritic function and spine motility and plasticity have been proposed. DSCR1, MNB/DYRK1A or ITSN1 may be candidates to explain the dendritic spine functional and structural alterations. Intersectin (ITSN1 and ITSN2) (Pucharcos et al. 1999, 2000) is a multidomain scaffold protein that interacts with several signaling proteins. A long splice variant of intersectin, ITSN1, that contains a Dbl domain with guanine nucleotide exchange factor (GEF) activity for Cdc42 is expressed specifically in neurons. Intersectin controls local formation and branching of actin filaments. Other proteins, such as Dyrk1A that phosphorylates actin-binding proteins, might also have a role in shaping the interaction of the spine membrane with the actin cytoskeleton. Yang et al. (2001) showed that overexpression of a kinase-deficient form of DYRK1A attenuates the neurite outgrowth induced by a neurogenic factor in immortalized hippocampal cells. The reduction in brain size and behavioral defects observed in mice lacking one copy of the murine homolog Dyrk1A (Dyrk1A+/–) supports the idea that this kinase may be involved in monosomy 21-associated MR (Benavides-Piccione et al. 2005). DSCR1 is overexpressed in fetal and adult DS brains, and a number of experimental data would be consistent with calcineurin inhibition by DSCR1 overexpression. Calcineurin is involved in synaptic plasticity and has a role in the transition from short- to long-term memory through perturbation in LTP and long-term depression (LTD), and calcineurin-dependent induction of DSCR1.4 product may represent an important auto-regulatory mechanism for the homeostatic control of Nuclear factor of activated t cells (NFAT) signaling in neural cells (Cano et al. 2005). Another interesting example is drebrin, an actin-binding protein, thought to regulate assembly and disassembly of actin filaments (Asada et al. 1994; Hayashi & Shirao 1999), thereby changing the shape of spines and the synaptosomal associated proteins-α SNAP and SNAP 25. In post-mortem DS brains, levels of drebrin are reduced in the early second trimester. Finally, another interesting example is S100B, a calcium-binding protein found within astroglial cells. When released, S100B has extracellular neurotrophic effects involving the neuronal cytoskeleton. The gene for S100B is located on chromosome 21, and levels of the protein are elevated in DS and Alzheimer's disease (AD). Thus, overexpression of S100B may be related to the cytoskeletal abnormalities seen in these disorders (Shapiro & Whitaker-Azmitia 2004).

Changes in levels of expression of these genes may lead to changes in the timing and synaptic interaction between neurons during development which can lead to suboptimal functioning of neural circuitry and signaling at that time and in later life.

Now that many MR genes have been cloned, new major questions arise regarding the functions of these genes in neuronal development, plasticity and function and the neurobiological alterations by which mutations in these genes affect information processing.

Fragile X syndrome

Fragile X syndrome is the most prevalent form of MR caused by mutation in a single gene, with a prevalence of about 1:4000. Fragile X syndrome is characterized by overall mild to moderate MR, macro-orchidism (enlarged testicles; most notable after puberty), a long face with a large jaw and protruding ears, hyperflexible joints and aortic dilatation. Verbal IQ is superior to performance IQ, but speech and language development are also impaired. Cognitive impairment may be largely dependent on deficient executive control (Cornish et al. 2004). Many boys with FraX suffer from attention deficit hyperactivity disorder (ADHD), with concentration problems and hyperactivity. Autistic features are seen in a fair proportion of FraX patients, with evasion of eye contact, social anxiety, communication problems and ritualistic behavior. About 20% of the patients show epileptic seizures (Musumeci et al. 1999).

Reiss et al. (1995) investigated neuroanatomical differences between 51 individuals with FraX and matched controls using magnetic resonance imaging (MRI). The volume of the caudate nucleus was increased, and in males also the volume of the lateral ventricle. The volumes of both structures were found to correlate with IQ, while the volume of the caudate nucleus correlated with the methylation status of the FMR-1 gene. In FraX post-mortem cerebral cortex, Hinton et al. (1991) and Irwin et al. (2001) found increased densities of dendritic spines with an immature morphology. Hinton et al. (1991) investigated the post-mortem brains of three FraX males with mild to moderate MR and did not find gross neuropathological alterations. No differences were found in the density of neurons in layers II–VI of cingulate and temporal association areas. In one brain, dendritic spine morphology was investigated, which confirmed a previous report that showed the presence of immature, long, tortuous spines in adult cerebral cortex in FraX. Using the Golgi-Kopsch method, Irwin et al. (2002) compared the length, morphology and number of dendritic spines on layer V apical dendrites of visual and temporal cortex of three persons with FraX syndrome with the temporal cortex of three age-matched controls. Fragile X brains showed more long and immature spines and fewer short and mature dendritic spines in both areas as compared with the temporal cortex of controls. In addition, FraX patients showed a higher density of dendritic spines on distal segments of apical and basal dendrites. The alterations in spine density and morphology in FraX syndrome could reflect a deficiency in synaptic pruning.

In almost all cases, FraX syndrome is caused by expansion of a CGG repeat in the 5′ untranslated region of the FMR1 gene. When this repeat expands beyond 200 triplets, the promoter region of the FMR1 gene becomes methylated, which blocks transcription of the gene and prevents expression of the FraX MR protein (FMRP). A missense mutation (I304N), found in one male, causes more unusually severe phenotype with extreme cognitive decline (IQ < 20) and macroorchidism (De Boulle et al. 1993).

Fragile X MR protein is a member of a small family of RNA-binding proteins that include the FMR1 homologs FXR1 and FXR2. The FMR1 protein (FMRP) contains from N- to C-terminal a nuclear localization signal (NLS), a coiled-coil domain, two RNA-binding KH domains, a nuclear export signal, another coiled-coil domain and an RGG box, which also binds RNA. Fragile X MR protein and its homolog are associated with polyribosome as cytoplasmic messenger ribonucleoprotein (mRNP) particles. Immunocytochemical staining of mouse brains showed that all three are expressed in brain, where they are colocalized in the cytoplasm of neurons (Bakker et al. 2000). In Purkinje cells and brain stem motor, FMRP was also found in the nucleolus. A minor proportion was observed in the nucleus, consistent with a role of FMRP and its homolog in nucleocytoplasmic shuttling (Bakker et al. 2000). Feng et al. (1997a) also demonstrated FMRP in dendrites and dendritic spines, as well as a lower presence in axons and presynaptic terminals, using immunoelectron microscopy. The association of FMRP with polyribosomes is probably of great functional importance, as the I304N mutation, which causes a severe phenotype in humans, may sequester mRNAs into non-translatable mRNP particles (Feng et al. 1997b). The functioning of FMRP may be regulated by phosphorylation on the highly conserved serine 499, close to the RGG box (Ceman et al. 2003). Unphosphorylated FMRP is associated with actively translating polyribosomes, while a fraction of phosphorylated FMRP is associated with apparently stalled polyribosomes, indicating that phosphorylation of FMRP may regulate translational suppression. In Drosophila, phosphorylation of the ortholog of FMRP influences its association with mRNAs (Siomi et al. 2002).

A current hypothesis for the role of FMRP is that FMRP is a key regulator in the targeting of specific mRNAs to dendrites and dendritic spines, where it controls the translation of these mRNAs, possibly in response to plasticity-inducing signals. This local translation of mRNAs has been proposed to modulate spine morphology and synaptic function. Important remaining questions concern the identity of the mRNAs that are targeted to the spines and the mode of translational control that FMRP exerts over these mRNAs. The RGG box in FMRP mediates preferential binding of RNAs that contain a G-quartet structure. Several groups have used different approaches to identify some of these mRNAs (Brown et al. 2001; Darnell et al. 2001; Miyashiro et al. 2003; Schaeffer et al. 2001; Zhang et al. 2001), while most studies indicate a role for FMRP in repressing translation. Fragile X MR protein has also been shown to be associated with the RNA-induced silencing complex (RISC), which mediates post-transcriptional gene silencing in response to double-stranded RNA and endogenous microRNA genes (reviewed in Carthew 2002; Siomi et al. 2004). Although FMRP does not appear to be required for RISC-mediated RNA degradation, it may play a role in the selection of RNAs that are silenced by the RISC. More recently, it was found that the second KH (KH2) domain in FMRP binds to a specific sequence element within the ‘kissing complex’, present in certain mRNAs (Darnell et al. 2005). Interestingly, unlike the G-quartet-containing mRNAs, RNA containing a kissing complex can elicit competitive dissociation of FMRP and polyribosomes. Because the I304N mutation is located within the KH2 domain and interferes with the binding of kissing complex RNA, this finding points to an important mechanism by which mutated FMR1 may cause FraX syndrome, as well as to a new class of mRNA targets of FMRP. The expression of FMRP is increased in vivo by environmental enrichment and whisker stimulation, conditions which are known to induce structural rearrangement of neocortical connectivity (reviewed in Greenough et al. 2001). Moreover, stimulation of synaptoneurosomes in vitro also promotes translation of FMRP, indicating that the levels of FMRP in synapses/spines are modulated by synaptic activity (Weiler et al. 2004).

A mouse knockout (KO) model for FraX syndrome has been generated by Bakker et al. (Dutch-Belgian Fragile X Consortium 1994). In contrast with patients with FraX, the FMR1 KO mouse showed a rather mild phenotype, characterized by a limited increase in activity, abnormalities in auditory sensory gating, mild spatial memory deficits and little if any change in exploratory and emotional behavior (Kooy 2003). In addition, object recognition might be impaired. Several tests have yielded contrasting observations in some cases, due to genetic background, which might also explain part of the phenotypic variability in humans. In addition, the highly homologous FXR1 and FXR2 proteins might compensate loss of FMRP. A mouse overexpressing FMRP showed decreased locomotor activity and increased anxiety and no effects in hippocampus-dependent behavior, opposite to effects seen in the KO mouse (Peier et al. 2000). In addition, the combined KO/overexpressing mouse showed impaired motor co-ordination, which was not seen in the KO mouse.

The mouse model did reproduce the main morphological alteration observed in FraX patients, namely a shift from mature-type dendritic spines of visual cortex pyramidal neurons to immature, long, thin spines (Comery et al. 1997; Irwin et al. 2002). In barrel cortex, spine length and density was only increased during the first postnatal week in the KO mouse, indicating that, at least in this cortex area, some alterations in the spines may occur only transiently early on (Nimchinsky et al. 2001). On the other hand, a lack of dendritic rearrangement in KO mouse barrel cortex supports the notion that developmental pruning of dendrites/spines is impaired (Galvez et al. 2003). This was further substantiated in Drosophila, in which only one ortholog of FMR1 exists. For instance, in the mushroom body, an important area of plasticity in DrosophilaPan et al. (2004) found that loss of DFMR1 resulted in increased process formation, elongated axons and increased axonal and dendritic branching, while overexpression decreased outgrowth and branching.

Loss of FMRP in mice did not affect LTP in hippocampal area CA1, but LTP in cerebral cortex was impaired and associated with a decrease in the GluRI subunit (Godfraind et al. 1996; Li et al. 2002; Paradee et al. 1999). Area CA1 did show enhanced non-NMDA-dependent LTD (Huber et al. (2002). Interestingly, this form of synaptic plasticity is mainly dependent on the mGluR5 receptor, and its maintenance requires postsynaptic translational activity. Because activation of mGluR5 stimulates translation of FMR1 mRNA at pre/postsynaptic sites, Huber et al. (2002) proposed that, in wild-type mice, FMRP represses translational activity that is necessary for the prolongation of LTD. Loss of FMRP would decrease this repression and promote prolongation of LTD. A consequence of this model could be that type I mGluRs may be targeted for therapeutic interventions in FraX syndrome. In support of this idea, in Drosophila, antagonists of mGlu receptors can rescue certain neuroanatomical and behavioral deficits caused by loss of DFMRP (McBride et al. 2005). Although the nature of the receptors targeted by the antagonists is not clear, these findings will stimulate attempts to develop pharmacotherapy for FraX (Dolen & Bear 2005).

Down syndrome MR

Down syndrome (DS) is a public health problem with more than 500 000 patients in Europe, most of them without a real autonomous life. It is the most frequent cause of MR of genetic origin and accounts for nearly 30% of moderate to severe MR. DS results from trisomy of chromosome 21, 95% of which are free trisomies, with the remaining 5% the result of translocations (4%) or mosaicism (1%) (Anneren & Edman 1993; Jyothy et al. 2002). Thus, it is a complex and multigenic disorder due to chromosome imbalance. However, based on rare cases of partial trisomies, a minimal chromosomal region situated in the long arm of chromosome 21, extending from APP to the distal end of HSA21, was defined, that when in triple copy is enough to produce most of the symptoms of DS. During many years, DS research focused on the identification of HSA21 markers, which facilitated the construction of the genetic and physical map of the chromosome, the analysis of partial trisomies and their relationship with the phenotypic maps and the development of the transcriptional map of HSA21. Genes on this DS critical region have been defined as the most probable candidates to produce the phenotypes (Antonarakis et al. 2004). The sequencing of HSA21 was a turning point for the understanding of DS and has allowed huge advances in comparative genomics.

Trisomy 21 reduces the IQ to ranges from 20 to 80 (Anneren & Edman 1993). In contrast to normally developing children and to other cases of MR, there is a progressive IQ decline in DS beginning in the first year of life, so that the ratio of mental age to chronological age is not constant. By adulthood, IQ is usually in the moderate-to-severe retarded level (IQ 25–55) with an upper limit on mental age of approximately 7–8 years, although a few individuals have IQs in the lower normal range (70–80) (see Nadel 2003 for review). In this context, several areas of speech and language development are more importantly delayed: articulation, phonology, vocal imitation and expressive syntax (Stoel-Gammon 2001); in contrast, language comprehension is relatively spared (Miolo et al. 2005). Similarly, hippocampal-dependent processes, such as visuo-spatial memory, and cerebellar function, are also more deeply affected (Nelson et al. 2005). It has been suggested that the rate of cognitive development tends to slow, as these children get older, so that a progressive postnatal degenerative process is burdening DS brains (see Nadel 2003 for review).

Neither the neuropathological features nor the molecular basis and the genes involved in the alterations in the cognition processes and in their early decline across development are known. Brains of patients with DS present specific features: brain size is reduced, mainly affecting specific brain areas, such as the cerebellum and the hippocampus; the number and depth of sulci in the cerebral cortex is decreased; observed are also neuronal heterotopias, abnormal neuronal migration/differentiation, decreased neuronal densities, affecting in particular specific cell populations such as granule cells in cerebral cortices and dendritic anomalies (Becker et al. 1991; Marin-Padilla 1972, 1976; Takashima et al. 1989, 1994; Vuksic et al. 2002). Beyond 35 years, the neuropathological changes seen in DS brains are identical to those seen in sporadic AD. These neurodegenerative changes are characterized by progressive accumulation of senile plaques and neurofibrillary tangles and occur with a similar regional distribution as in AD (Wisniewski & Kida 1994).

At the microscopic level, neuronal density is decreased in distinct regions of DS brains, including the cochlear nuclei, cerebellum, hippocampus, basal forebrain, the granular layers of the neocortex and areas of the brainstem. In the neocortex, neuronal morphology is abnormal, and neuronal orientation is aberrant (reviewed in Coyle et al. 1986; Flórez 1992). Recently, it has been reported (Raz et al. 1995) using MRI that young adult DS subjects have reduced volumes of cerebral and cerebellar hemispheres, ventral pons, mammillary bodies and hippocampal formation, even after adjustment for body size. In contrast, the parahippocampal gyrus is enlarged. However, it is difficult to attribute the cognitive alterations in DS to a simple reduction of brain size or neuronal density. Instead, different aspects of the three-dimensional neuronal cell structure such as the size of the basal dendritic arbors, their branching structure and spine density may have a specific pathogenesis and evolution in DS.

In this regard, several studies (Becker et al. 1986, 1991; Marin-Padilla 1976; Suetsugu & Mehraein 1980) have reported reductions in cortical width, abnormal cortical lamination patterns, altered dendritic arbors and dendritic spines, aberrant electrophysiological properties of membranes, reduced synaptic density and abnormal synaptic morphology in DS patients (Reviewed in Benavides-Piccione et al. 2004). Some of these alterations are probably due to an abnormal development of the nervous system during postnatal life.

Quantitative analysis of dendrites in layer IIIc pyramidal neurons of prefrontal cortex (prospective area 9) of the brains of 2.5-month-old infants revealed no significant differences in dendritic differentiation between normal and DS cases (Vuksic et al. 2002). In contrast, basilar dendrites of cortical pyramidal neurons were shorter than normal in DS subjects older than 4 months (Takashima et al. 1981). These findings suggest that children with DS begin their lives with morphologically normal layer III pyramidal neurons and that pathologic changes occur after 2.5 months of postnatal age. Subsequent to this age, there is a steady decrease, so that, in subjects with DS older than 2 years, these parameters are reduced relative to controls especially in apical dendrites (Becker et al. 1986).

In adults with DS cross-sectional studies demonstrate marked reductions in dendritic branching and length, and in spine density (Becker et al. 1991; Takashima et al. 1989). Degenerative neuronal changes are also associated with dendritic abnormalities. Two other reports on pyramidal neurons in parietal cortex (Schulz & Scholz 1992) and on non-pyramidal neurons in motor cortex (Prinz et al. 1997) support these findings. In cases of aged DS brains that do not show AD-like pathology, the number of spines in the middle and distal segments of apical dendrites of pyramidal neurons in cingulate cortex and in hippocampus is reduced significantly. These changes are more severe in area CA1 of those DS cases with AD (Ferrer & Guilotta 1990; Suetsugu & Mehraein 1980).

Taken together, these results suggest that Down syndrome MR is most probably another example of deficient neuronal network connectivity that results in deficient information processing and MR. In support of this hypothesis, functional imaging with positron emission tomography showed reduced functional connections between brain circuit elements (Azari et al. 1994; Balogh et al. 2002). This abnormally connected brain could thus be generally affected or could have disturbances in specific pathways or neurotransmitter systems.

A remaining question is whether the neurons in DS do not achieve the structural complexity during postnatal development, or whether they undergo greater dendritic retraction than those of euploid brains during maturation. It is probable that both abnormal postnatal development in DS brain and the correct remodeling of the brain circuitry are responsible for the dendritic abnormalities observed in DS. These alterations may be responsible for maladaptative behaviors and for the inability to acquire and stabilize the information in DS (Clark & Wilson 2003; Thiel & Fowkes 2005). However, from a mechanistic point of view, neural network formation and neural plasticity are complex processes that raise some important biological questions in DS: what are the signaling pathways that trigger altered neuroplasticity? and, what are the molecules that provide the information for this process?

Because the abnormal development and maturation of the DS brain is most probably regulated by specific genes in HSA21, the identification of the catalogue of human genes in the context of the Human Genome Project has been the first step toward the understanding of the molecular mechanisms leading to neuronal abnormalities. Dendritic pathology in DS may be the consequence of higher levels of proteins coded by three copies of genes on HSA21. The challenge is now to decipher the role of the relevant proteins encoded, information that should facilitate a rapid and global understanding of DS pathology. The study of specific genes from within the DSCR, specially those that have been demonstrated to be dosage-sensitive, with cognitive impairment may bring to light some interesting and unsuspected culprits that will give us entry points into novel biological pathways.

Transgenic mouse models that carry single candidate genes and chromosomal segments homologous to the DSCR to excess, such as Ts1Cje and Ts65Dn, two partial trisomy models of murine chromosome 16 (MMU16) show some but not all the pathological, biochemical and transcriptional changes seen in DS (reviewed in Cairns 2001). Ts65Dn mice phenotype is characterized by cognitive deficits, hyperactivity, behavioral disruption and reduced attention levels similar to those observed in DS (Escorihuela et al. 1995, 1998). The cortical pyramidal neurons of these mice show reduced dendritic tree with significantly less spines, suggesting that behavioral deficits could be attributed to abnormal circuit development (Fig 1.). Moreover, in the hippocampus (fascia dentata, CA1) as well as the motor and somatosensory cortex, entorhinal cortex, and medial septum confocal microscopy showed that both presynaptic and postsynaptic elements were significantly enlarged and showed abnormal internal membranes in Ts65Dn. In addition, spine density was decreased on the dendrites of dentate granule cells, and there was reorganization of inhibitory inputs, with a relative decrease in inputs to dendrite shafts and an increase in inputs to the necks of spines (Belichenko et al. 2004). On the contrary, environmental enrichment had a marked effect on pyramidal cell structure in control animals, so that pyramidal cells in environmentally enriched control animals were significantly more branched and more spinous than non-enriched controls. However, environmental enrichment had little effect on pyramidal cell structure in Ts65Dn mice (Dierssen et al. 2003). For these reasons, Ts65Dn mouse has been proposed as a good model for abnormal synapse structure and function in DS and point to the importance of studies to elucidate the mechanisms responsible for synapse enlargement. Recently, a new trisomic model, designated as Ts{Rb[12.17(16)]}2Cje, carrying a chromosomal rearrangement of the Ts65Dn genome, whereby the marker chromosome has been translocated to Chromosome 12 (MMU12) forming a Robertsonian chromosome, has been created. This model retains a dosage imbalance of human Chromosome 21 homologous genes from App to the telomere and expression levels similar to Ts65Dn within the triplicated region. Quantitative confocal microscopy of granule cells in the fascia dentata revealed decreased spine density on the dendrites of dentate granule cells and significantly enlarged dendritic spines affecting the entire population in Ts{Rb[12.17(16)]}2Cje as compared with 2N controls (Villar et al. 2005). These findings document that the structural dendritic spine abnormalities are similar to those previously observed in Ts65Dn mice. There are also some single-gene transgenic models that stress the importance of specific dosage-sensitive genes in the dendritic pathology associated to MR. Transgenic mice overexpressing human S100B shows changes in cytoskeletal markers such as the dendritic associated protein, MAP-2, the growth-associated protein-43 and the dendritic spine marker, drebrin (Shapiro & Whitaker-Azmitia 2004), leading to an increased density of dendrites within the hippocampus of transgenic animals. More specifically, alterations in the phenotype of neocortical pyramidal cells are observed in the Dyrk1A+/– mouse. Basal dendritic arbors of layer III pyramidal cells of Dyrk1A+/– mice are smaller, less branched and less spinous than those in their control littermates. It needs to be determined whether pyramidal cells in juvenile Dyrk1A+/– mice do not achieve the structural complexity of those in wild-type mice or whether they undergo more intense dendrite retraction during maturation. DYRK1A might be implicated in both processes, because it has been shown that overexpression of a kinase-deficient DYRK1A impedes neurite outgrowth (Yang et al. 2001) and that a putative substrate of DYRK1A, dynamin 1, has also been implicated in neurite outgrowth (Chen-Hwang et al. 2002). As pyramidal cells comprise more than 70% of the neurons in the cortex, it is reasonable to think that Dyrk1A is involved in the determination of complexity of circuits in the mature cerebral cortex (Dierssen et al. 2003). However, as occurs with other MR syndromes, the disrupted pyramidal phenotype is probably due to the convergence of effects of various dosage sensitive genes.

Figure 1.

An example of neuronal network structure in adult mouse primary visual cortex as revealed by Golgi staining. On the right are non-pyramidal and pyramidal neurons at a higher magnification, as well as part of a dendritic segment with spines. Alterations in neuronal morphology are usually not directly visible in brains of persons with mental retardation (MR) or mouse models of MR, and will only be revealed by quantitative analysis. Essential in this respect is a blind and random sampling procedure, to prevent various sources of bias.

Comparison between FraX/DS brain phenotypes

Down syndrome and FraX syndrome are probably the most thoroughly investigated forms of MR, in terms of both the human studies and the underlying neurobiology. Because FraX syndrome is a single-gene disorder, studies have been more directed than in DS, a polygenic disorder, in which some of the major contributing genes are beginning to emerge. In FraX syndrome, the ability to focus on a single gene has led to a reasonable understanding of the role of the encoded FMRP at the molecular, cellular and neuronal network level, as well as the regulation of its expression and activity. Consequently, a model of FMR1 as a regulator/repressor of the translation of specific mRNAs has emerged. Importantly, this regulatory activity may be partly under environmental control and possibly result in site-specific (repression of) translation. On the other hand, the function of FMRP in the control of translation has led to a search for the relevant target mRNAs, and in an indirect way, FraX has become a polygenic disorder like DS. Assuming that FMRP functions as a repressor of translation, FraX and DS have in common that they cause MR by overexpressing different ranges of proteins. Although none of these genes/proteins are shared, the majority is signal transduction proteins, some of which are involved in cytoskeletal regulation. Both FraX and DS are characterized by alterations in spine morphology, indicating an increased proportion of immature spines, and consequently abnormal development or maturation of spine synapses in both syndromes. However, whereas DS also shows reduced dendritic complexity, this has not been observed in FraX. Similarly, signs of AD neuropathology are seen early in DS, as well as a very high incidence of Alzheimer dementia, but not in FraX. The Alzheimer pathology can be ascribed to overexpression of APP in DS although not exclusively. Whether common factors/genes lead to the spine changes is not known, but it is likely that, in both cases, dysregulation of the (actin) cytoskeleton is involved. Cognitive phenotypes also differ. In both syndromes, a comparable spread in the severity of the cognitive disability is seen, whereas in DS verbal abilities are weaker than performal abilities (Pennington et al. 2003), the reverse is seen in FraX syndrome. Also, executive functions and attention appear to be especially weak in FraX, indicative of prefrontal dysfunction, while in DS hippocampal functions appear to be more affected than prefrontal functions (Pennington et al. 2003). Altogether, FraX and DS mainly share the decline in global cognitive abilities and adaptive skills and the abnormalities in spine morphology. This correlation supports the view that the abnormal spines and associated abnormal neuronal network may be the primary cause of the cognitive deficit, in line with our network hypothesis of MR.

Final remarks

Investigation of the causes and the underlying (neuro) biological mechanisms of MR are justified by itself, given the high prevalence and the emotional, social and economic strain MR causes to families and society. Identification of the genetic and environmental causes is essential for the prevention of MR. Study of the underlying cellular mechanisms is required to understand the neuropathology responsible for MR and essential for the development of therapies for MR.

The identification of many of the causative genes has lead to the investigation of the neurobiological mechanisms that give rise to MR. Both in FraX and in DS, as well as in other forms of MR, increasing evidence supports the view that deficient structural neuronal network connectivity is a major, if not, primary cause of MR. As exemplified by the findings in FraX research, our increasingly detailed understanding of the molecular and environmental mechanisms that underlie the development and/or activity-dependent plasticity of neuronal circuitry is for the first time pointing to possible means of intervention.


This study was supported by FIS PI041559, CIEN Foundation, MCYT (SAF2002-00799), EU QLGI-CT-2002-00816, Jerôme Lejeune Foundation. GJA Ramakers is supported by EU grant QLG3-CT-2002-01810 and a grant of the NWO-Cognition program.