Aneuploidy: From a Physiological Mechanism of Variance to Down Syndrome

Mara Dierssen, Yann Herault, Xavier Estivill


Quantitative differences in gene expression emerge as a significant source of variation in natural populations, representing an important substrate for evolution and accounting for a considerable fraction of phenotypic diversity. However, perturbation of gene expression is also the main factor in determining the molecular pathogenesis of numerous aneuploid disorders. In this review, we focus on Down syndrome (DS) as the prototype of “genomic disorder” induced by copy number change. The understanding of the pathogenicity of the extra genomic material in trisomy 21 has accelerated in the last years due to the recent advances in genome sequencing, comparative genome analysis, functional genome exploration, and the use of model organisms. We present recent data on the role of genome-altering processes in the generation of diversity in DS neural phenotypes focusing on the impact of trisomy on brain structure and mental retardation and on biological pathways and cell types in target brain regions (including prefrontal cortex, hippocampus, cerebellum, and basal ganglia). We also review the potential that genetically engineered mouse models of DS bring into the understanding of the molecular biology of human learning disorders.


Fifty years after the discovery of the molecular basis of Down syndrome by Lejeune et al. (214), aneuploidy, defined as an abnormal number of copies of a genomic region, is recognized as a common mechanism of human genetic disease, often leading to abnormal gene expression patterns with over- or underexpression of specific genes (59, 156). Classically, the term was restricted to the presence of supernumerary copies of whole chromosomes (trisomy) or the absence of chromosomes (monosomy). However, with the advent of high-resolution molecular tools to scan the genome, novel aneuploidy syndromes are being identified, and the definition has been extended to include deletions or duplications of subchromosomal regions. Good examples are segmental duplications (SDs) and copy number variants (CNVs) that can contribute to phenotypic diversity and to disease (6, 112, 137, 166, 324, 333335), influence gene expression indirectly through positional effects, predispose to deleterious genetic changes, or provide substrates for chromosomal changes (115, 116, 124, 228). The recently published catalog of CNVs and SDs of the human genome in different populations has demonstrated the ubiquity and complexity of these forms of genetic variation (290; see below). Thus, to include all these possibilities, the term aneuploidy had to be redefined. Suggested alternatives are microaneuploidy or segmental aneuploidy, implying that the pathogenesis of the disorder is the result of inappropriate dosage of critical genes within a discrete genomic segment or of changes in the expression pattern of neighboring genes. In support of this assumption, high-throughput expression profiling in a number of organisms has revealed that variation in gene expression levels within and among populations is abundant, with a large proportion of genes showing significant patterns of interindividual variation (244, 349). These new data suggest that quantitative differences in gene expression are responsible for a significant amount of variation in natural populations that can lead to physiological or pathological differences among individuals (83; see next section).

The most common example of neurogenetic aneuploid disorder is Down syndrome (DS; OMIM 190685), a neurodevelopmental disorder caused by the presence of an extra copy of all or part of human chromosome 21 (HSA21 for Homo sapiens chromosome 21). Trisomy 21 can be divided into four categories according to the size of triplicated genomic region: complete trisomy, partial trisomy, microtrisomy, and single-gene duplication. The majority of DS patients (>95%) have three complete copies of human chromosome 21, and others have mosaics or translocations. The trisomic condition in DS is of particular interest for understanding how deregulated gene expression in the brain will lead to altered brain function and specifically to subnormal intellectual functioning. In trisomy 21, the effects of the gene-dosage imbalance on brain phenotypes have been explained by two hypotheses: 1) the “gene-dosage effect” hypothesis claims that DS phenotypes are determined by increased dosage of a subset of dosage-sensitive genes, and of their encoded proteins, especially during development (5, 79, 202, 229), and 2) the “amplified developmental instability” hypothesis holds that HSA21 trisomy determines general alteration in developmental homeostasis (for review, see Refs. 13, 303).

The gene-dosage hypothesis proposes that the 1.5-fold increase in the expression of some, if not all, of the around 430 genes that have been identified on HSA21 (; Refs. 74, 131, 157, 297, 298, 302, 338) is responsible for the brain anomalies and moderate mental retardation of individuals with DS (12, 100). According to this hypothesis, genes encoded by HSA21 in the trisomic state are overexpressed, thus leading to an imbalance of critical genes (79). It has been proposed that the so-called DS-critical region (DSCR; part of 21q22.1 to 21q22.3), identified by studying human patients with partial segmental duplication of HSA21, is sufficient when triplicated, to give rise to most of the DS phenotypic traits. However, several evidences have questioned this hypothesis. First, analysis or a large panel of human patients with segmental duplications revealed that genes located in regions of HSA21 outside the DSCR also contribute to the DS phenotypes (202, 229), and work on mouse models that will be described in section iv, led to a similar conclusion showing that triplication of the region homologous to the DSCR in mice does not recapitulate the DS phenotypes (265).

One reason may be that increased copy number does not always correspond with increased gene expression level or even less with increased gene function. Even though most of the expression studies comparing trisomic to euploid tissues support the hypothesis of increased mRNA levels due to gene dosage (56, 130, 185, 230), several reports could not find increased protein levels in trisomic brains for some specific HSA21 genes (198), and others have shown up- or downregulation of genes located on disomic chromosomes. One possible explanation for these controversial results on expression levels could be that regulatory mechanisms are also deregulated. One fascinating example was recently given showing that trisomy 21 gives rise to overexpression of HSA21 miRNAs that could result in turn in decreased expression of specific target proteins and thereby contribute to features of the DS phenotype. This interesting finding was obtained by Kuhn et al. (205) that detected overexpression of several HSA21-encoded miRNAs, namely, miR-99a, let-7c, miR-125b-2, miR-155, and miR-802, in fetal brain and heart specimens from individuals with DS when compared with age- and sex-matched controls. This makes clear that to understand the impact of gene expression alteration in DS, we will have to consider new and previously overlooked elements such as transcription factors, noncoding RNAs (microRNAs, associated small interfering RNAs, other unclassified non-translated RNAs; 205, 239, 329), epigenetic processes (e.g., CpG dinucleotide methylation, histone acetylation and deacetylation), and mRNA transcript variation (e.g., alternative promoters, alternative polyadenylation sites, differentially spliced exons and introns). In particular, the unexpected abundance and functional significance of noncoding RNAs has to be explored in DS. These RNAs control a remarkable range of biological pathways and processes, all with obvious consequences, such as initiation of translation, mRNA abundance, transposon jumping, chromosome architecture, stem cell maintenance, and the development of brain.

An additional problem in DS is the temporal pattern of change and regulation of overexpressed genes along the entire life-span, and the impossibility of separating the effects of trisomy that disturb development from those that alter function of mature cells. Upregulation of trisomic genes in a specific cell, in a particular time frame in the adult period, may lead to altered function but could also produce a developmental error that has later functional consequences (developmental versus functional effects). Thus important goals will be to determine the relationship between microscale transcriptional temporal-spatial profiles and specific cellular phenotypes. Moreover, both cell autonomous and nonautonomous effects may also contribute to specific DS phenotypes so that genotypically trisomic cells can cause other cells (regardless of their genotype) to exhibit a mutant phenotype (187). An example of cell-autonomous deficits has been observed in response to Sonic hedgehog (SHH) in trisomic neuronal precursor cells in the cerebellum of mouse models of DS (304 and see sect. iv).

Other differences between trisomic and euploid individuals may include presence/absence of an extra centromere, a potential trisomic maternal environment during development, and the presence of trisomy in all or in part of the cells (mosaicism). The centromere plays a fundamental role in accurate chromosome segregation during mitosis and meiosis in eukaryotes. Its functions include sister chromatid adhesion and separation, microtubule attachment, chromosome movement, formation of the heterochromatin structure, and mitotic control (which correct functioning is essential for cell growth and mitotic progression). Thus it could be proposed that an extra centromere could have deleterious consequences and could contribute to DS-associated phenotypes, including the altered growth rates and decreased mitotic index and proliferation rates described in DS patients and mouse models.

Taken together, these data demonstrate that the elucidation of the exact nature of the genetic mechanisms and molecular pathways that underlie the cognitive alterations in DS is still far from clear. One additional problem in establishing a genotype-phenotype correlation in DS is the presence of a broad phenotypic variability among DS individuals. This directly relates to the fact that trisomy 21 can have differential pathogenicity on individual genomes. One example is syndrome-specific facial abnormalities: we can easily recognize a DS patient across different races, but each DS individual is different and one can see their resemblance with their family. Thus, even though they share some morphogenetic characteristics, these coexist with family- and ethnic/race-specific traits (44, 362).

Studies of expression variation of HSA21 genes in lymphoblastoid cells (282) suggested that allelic differences are likely to be important determinants of the phenotypic variability of DS. These studies have shown considerable overlap in total expression levels between cells from normal and trisomy 21 individuals due to allelic variation, leading to the hypothesis that overexpressed genes whose expression varies little between individuals would be predicted to lead to the more penetrant phenotypes, whereas genes with high variation in expression would contribute to incompletely penetrant or variable DS-related phenotypes. Considering the importance of allelic variation to the outcome of the DS phenotype, the “critical region” (79) could be interpreted as a region encompassing genes with less allelic variation, thus contributing more importantly to DS phenotypes (265267).

Finally, the assumption that several HSA21 genes may be better candidates for mental retardation is tempting because it may open new therapeutic possibilities. Should one specific gene in trisomy give rise to a specific phenotype, its normalization would be of therapeutic value. To understand the genotype/phenotype correlation completely, functional information on the impact of specific subchromosomal aneuploidies is necessary because this will give information on the molecular networks in which candidate genes operate and on how changes in these networks lead to changes in disease traits. However, annotation of the chromosome continues and the functions of many genes on HSA21 remain uncertain, making it difficult to identify good candidates. The major challenge still remains to not only confirm that over- or underexpressed genes are truly those involved in specific disease traits, but to elucidate the functional roles that these genes play with respect to disease. Probably genes with a broad impact, affecting genetic regulatory circuits, or influencing feedback loops, buffers, and amplifiers within these circuits will be of special interest for intervention (14).


According to classical cytogenetics definition, aneuploidy corresponds to the loss or gain of a complete chromosome leading to an abnormal karyotype. When a single chromosome is lost (2n-1), it is called a monosomy. In these cases the daughter cell(s) with the defect will have one chromosome missing from one of its pairs. When a chromosome is gained, it is called trisomy, in which the daughter cell(s) with the defect will have one chromosome in addition to its pair. The biological consequences of loss or gain of whole chromosomes usually are devastating and are the hallmark of numerous pathological conditions in humans. Now, with refinement of the DNA analysis technology, the term is used to describe changes in copy number that affect only portions, small or large, of a chromosome.

A. Nondisjunction and Aneusomies

Nondisjunction refers to the failure of chromosome pairs to separate properly during cell division. It can be produced by a failure of homologous chromosomes to separate in meiosis I, or the failure of sister chromatids to separate during meiosis II or mitosis. The result of this error is a cell with an imbalance of chromosomes. Abnormalities arising from nondisjunction events cause cells with aneusomy (additions or deletions of entire chromosomes). Most aneuploidy cases derive from errors in maternal meiosis I, with maternal age being a risk factor for most human trisomies (155). It has been shown that alterations in recombination are an important contributor to meiotic nondisjunction. In DS, the parental origin of the extra HSA21 and the meiotic versus mitotic origin are known. Errors in meiosis that lead to trisomy 21 are mainly of maternal origin with an association with advanced maternal age, while ∼5% occur during spermatogenesis. Most errors in maternal meiosis occur in meiosis I, while maternal meiosis II errors constitute 20% of all cases of trisomy 21. In 4.5% of the cases, the supernumerary chromosome results from an error in mitosis (11, 264, 336).

B. Genomic Rearrangements and Genomic Disorders

Genomic rearrangement is the general term for alterations of the architecture of chromosomes, which can lead to duplications, deletions, and inversions of a given genomic segment; to translocations, additional chromosomal material (marker chromosomes), and isochromosomes (metacentric chromosomes produced during mitosis or meiosis); or to other complex rearrangements (347). Chromosome rearrangements have been described since the very early use of cytogenetic and molecular tools to study human disorders (67, 68, 106, 138, 383). The identification of recurrent genomic changes in some diseases led to coinage of the term genomic disorders to refer to clinical phenotypes that are due to genetic alterations that affect a relatively large segment of the genome (227). Genomic disorders often involve genomic rearrangements that span large deletions or duplications and affect several genes.

Genomic rearrangements can be sporadic or recurrent (Figs. 1 and 2). Although most originate in the germ cells, they can also be of somatic origin. Independently of their origin, rearrangements often are produced of specific sequences or regions of the genome. Recurrent rearrangements tend to have breakpoints in specific sequences, while most sporadic changes have different breakpoints. Nonallelic homologous recombination (NAHR) between sequences of high level of identity (>90%), known as low copy repeats (LCRs, also defined as duplicons and segmental duplications), facilitates the occurrence of rearrangement (227), whereas rearrangements that do not occur in the same specific genomic region are the result of nonhomologous end-joining (NHEJ) rather than NAHR (335).

FIG. 1.

Types of structural variation of the human genome. Representation of a human chromosome with the reference order of four segments (A, B, C, and D) with respect to the annotated sequence of the human genome. Deletions, duplications, or inversions could occur at the sequence of reference. The finding of such events in the population can depend on the viability of the rearrangements or the selective advantage that they might have. Some segments can be amplified and become a segmental duplication, which may lead to a change in copy number. Some copy number variants (CNVs) have a complex organization with different types of changes within the same cluster.

FIG. 2.

Main mechanisms that lead to CNV changes. Nonhomologous recombination between sequences with a high level of identity (segmental duplications, low copy repeats, or duplicons) (over 90%) might cause the duplication or deletion of genetic material. Depending on the peculiarities of genes involved in the rearrangements, a clinical phenotype could be observed. Inverted duplicons might cause an inversion of the genetic material, but other types of changes could occur depending on the complexity of the duplicons, which often contain sequences that are in a parallel orientation, in which case deletions or duplications can result.

Over 50 genomic disorders have been identified (Table 1). With the wide use of genomic arrays (49), we are now able to explore the genome in a systematic way and with exhaustive coverage. This has led to the identification of new genomic changes associated with clinical disorders of unknown molecular basis (374), and to the definition of clinical syndromes that were previously not recognized (331, 334).

View this table:

Syndromes and disorders associated with copy number variants

1. Segmental duplications and copy number variants

Segmental duplications were initially defined based on comparison of segments sharing at least 90% identity with known or reference genome sequence duplications (20, 55, 332). Over 5% of the human genome is composed of segmental duplications (20, 55, 105). These can be divided into intra- and interchromosomal segmental duplications, although a subset of duplications occurs both intra- and interchromosomally. These dual cases often reflect a complex organization for some genomic sequences. Intrachromosomal segmental duplications that are organized in tandem along the chromosome often lead to deletion, duplication, or inversion rearrangements (228).

Comparison of sequences from two individuals indicates that the genomic structure of the human genome varies largely (194). Two reports in 2004 described that the human genome varies in its sequence at several regions (177, 325). This opened the concept of CNVs, which were originally described in 1980 in the α-globin gene (138). These reports were followed by others that, either directly (221, 332) or indirectly (59, 241), explored structural variation of the human genome (115, 194). Most of these initial studies were not complete, as the type of tools employed and samples that were analyzed only provided a partial view of the structural variability of the human genome. Systematic analyses of the samples used for the International HapMap Consortium (180) have shown that possibly up to 12% of the human genome varies at the structural level (394). Redon et al. (290) showed that the human genome contains at least 1,400 CNV regions, which correspond to a much larger number of loci. The most recent version of the Genome Diversity web page ( includes over 15,500 CNV loci (177).

The number of CNVs currently known is far from complete, as the efforts made to elucidate their localization and characterization at the structural level are very rudimentary for the majority of CNVs, in particular for those that have a complex organization or that are multiple alleles. An estimate of the number and type of CNVs that remain to be discovered indicates that the majority of CNVs are smaller than those described so far.

2. Copy number variants, genomic variability, and disease

Variability of the human genome sequence has been used to explore diversity, to construct genetic maps, and to identify the genes responsible for more than 2,000 human monogenic disorders ( Single nucleotide polymorphisms (SNPs) are the most abundant type of genetic markers with over 12 million entries deposited in public databases ( The HapMap Consortium has genotyped nearly four million SNPs ( and has reported blocks of linkage disequilibrium (180), which should help to define the location of genetic changes responsible for different phenotypes in association studies of common disorders and complex traits.

Although genetic variability was initially thought to be mainly restricted to SNPs and to small repeat amplifications, CNVs have emerged as a common and useful source of variability. CNVs and structural variants include deletions, duplications, inversions, and complex rearrangements (Figs. 1 and 2). They are observed variables during genetic transmission and sometimes also during somatic replication (279). CNVs are now considered a common source of genomic variability in several organisms (143, 274).

The phenotypic consequences of these submicroscopic variations have only been defined for few loci. Thus, while a typical SNP implies one single nucleotide change, CNVs can affect between one kilobase and several megabases of DNA per event, corresponding to a significant fraction of the genome sequence. The extensive number of CNVs in the genomes of individuals has provided new hypotheses about phenotypic variability in inherited disorders, aneuploidy syndromes, and common disorders (26).

3. Aneuploidy syndromes and phenotypic variability

Phenotypic variability associated with aneuploidy has mainly been attributed to variability in gene expression of the genes located in the aneuploid regions or to the contribution of other genes of the genome (13). However, genomic imbalances involving genes of the aneuploid chromosome could also contribute to the phenotype (26). Some genes located in the aneuploid chromosome or the aneuploid chromosomal region could be within a CNV region. If these genes are sensitive to gene dosage, their copy number could have functional consequences, either boosting or diminishing expression levels, with consequences to the phenotype. Thus CNVs should be considered as potential mediators of the phenotypic variability of aneuploidies (26). Through altering gene dosage, disrupting genes, or perturbing regulation of their expression, even at long-range distances, CNVs exert their actions on numerous genes, therefore potentially having many phenotypic consequences (244, 349). Genome-wide efforts to uncover genomic changes involved in patients with clinical alterations are underway (

CNVs have been found as events of common genetic disorders in human and other mammalian systems (148, 212, 275, 382). Common CNVs have been detected in subjects affected by several inherited disease. Cases of hereditary pancreatitis are caused by duplications or triplications of the cationic trypsinogen gene (211). Somehow the susceptibility to human immunodeficiency virus (HIV)-1 infection may vary depending on copy numbers of the CCL3L1 chemokine gene (137). Individuals with low copy numbers of CCL3L1 related to their ethnic background have markedly enhanced HIV-1/acquired immunodeficiency syndrome (AIDS) susceptibility. Increase in CCL3L1 gene dosage also influences predisposition to autoimmune disease such as rheumatoid arthritis (242). Similarly, a copy number polymorphism in FCGR3 has shown to predispose to several types of systemic autoimmune disorders, such as systemic lupus erythematosus (SLE), microscopic polyangitis, and Wegener's granulomatosis (6, 109, 386). CNVs in complement component C4 (C4A and C4B) lead to different susceptibilities to SLE (398). Compared with healthy subjects, patients with SLE have lower copy numbers of C4 and C4A. Interestingly, variability in copy number for the C4 genes and the genetic association with markers of the major histocompatibility complex (MHC) on chromosome 6p21.32 has been known to be variable for several years, but their complex organization and their relationship with SLE has not been revealed until now. Variability within the β-defensin 2 gene (DEFB4) shows that low DEFB4 gene copy number predisposes to Crohn's disease, with diminished β-defensin expression in colon of patients with the disease (112). Finally, the defensin locus has been found to be associated with susceptibility to psoriasis (166).

Several studies point clearly to CNV as genetic cause of cognitive and neuropsychiatric disorders (63, 375). Multiple cases of patients with Parkinson's disease due to genomic duplication or triplication of the α-synuclein gene (SNCA) have been reported, suggesting that SNCA may contribute to hereditary early-onset parkinsonism with dementia (344). Several cases of duplication of the amyloid precursor protein (APP) have been described in families with early-onset Alzheimer's dementia with cerebral amyloid angiopathy (308). On the contrary, decrease in CNTNAP2 dosage is associated with epilepsy and schizophrenia in Caucasian population (126), and a few CNV were shown to disrupt candidate genes in schizophrenia patients (375). CNV may also act as a susceptibility factor as has been described at the 7q11.23 segmental duplications for the Williams-Beuren syndrome deletion (30, 70). Microdeletions and microduplications in genes have been associated with autism (30, 108, 116, 186, 237, 384).

The use of technologies such as sequencing or deeper array methods, which are able to explore the genome at higher resolution in many individuals with the development of new tools for statistical and epidemiologic analysis, should provide a more comprehensive picture of the structural variation of the human genome (98, 201) and will lead to a better understanding of human disease (181).


Aneuploidy usually manifests as a highly variable constellation of phenotypes including malformations of different systems and organs (e.g., cardiac and gastrointestinal), morphogenetic abnormalities affecting craniofacial structure, limbs, and digits, and hematological and metabolic abnormalities. Most aneuploidy syndromes present a high incidence of brain phenotypes that affect learning, language, and behavior and show increased vulnerability for psychiatric disorders. The susceptibility of the brain to gene dosage changes probably depends on the unprecedented degree of cellular diversity, anatomical asymmetries (laterality) and specialized regional micro domains, regional interconnections, structural and functional plasticity, and exquisite environmental responsiveness that characterizes mammalian and in particular primate nervous system. An increasing number of reports of rearranged and aneuploid chromosomes in brain cells suggest a link between developmental chromosomal instability and brain complexity (132). Aneuploidy may especially affect brain regions, such as the cerebral cortex, in which the phenotypic properties of neuronal and nonneuronal cells are largely the result of unique combinations of gene products so that the cell pattern of gene expression is crucial for achieving cell phenotypes and cellular diversification. Interestingly, the assessment of chromosome variations in normal mouse and in normal and diseased human brain has indicated that aneuploidy is present in a high proportion of neural cells (188, 293, 294), thus giving rise to a mosaicism of intermixed aneuploid and euploid cells. Approximately 33% of neural progenitor cells display genetic variability, manifested as chromosomal aneuploidy that encompasses both loss and gain of chromosomes. In the mature brain, aneuploid cells that express neuronal markers are generated at least in part by chromosomal missegregation mechanisms during mitosis. In mouse brain, aneuploid neurons have been shown to participate in the development of cortical circuitry and to become functionally integrated into that circuitry (188, 293, 294, 399, 400). If we consider that the phenotypic properties of different neuronal and nonneuronal cells in the cortex are largely the product of unique combinations of expressed gene products, gene expression variation profiles are crucial to define cellular diversity. Thus aneuploidy with physiological consequences exists serving to neural cell diversity or as a mechanism of selection of specific cells. Conversely, pathogenetic aneuploidy can affect morphogenesis and function in the brain (187) and correlates with the severity of several neurogenetic disorders (248).

Genes, brain structure, and behavior are strongly related so that differences in genetic brain maps may be fundamental for determining individual physiological or pathological differences in cognition. Abilities such as speech, consciousness, tool use, symbolic thought, cultural learning, and self-awareness are highly dependent on the interplay between different genetic and environmental factors in spatial-temporal dimensions, leading to individual differences in brain organization and function. In this regard, recent studies of neural tissue from equivalent regions of different primate species suggest a general upregulation of gene expression in the human cortex compared with the chimpanzee cortex (42, 99, 145, 365). These upregulated genes tend to be enriched in genomic regions that have been recently duplicated in human evolution (192, 193) and that can give rise to new genes (93).

But, what could be the mechanisms by which the anatomical, chemical, and neurophysiological brain abnormalities underlying subnormal intelligence arise from deregulation of gene expression? Efforts to explain the major features of mental retardation are still mostly based on superficial gross neuroanatomical features (e.g., size, sulcal patterns) and on the analysis of high-level functions that lack precise neurobiological predictions (e.g., general intelligence, innate grammar) (for review, see Ref. 257). In DS, high-resolution magnetic resonance imaging (MRI) has revealed neuroanatomic abnormalities in the cerebral cortex that correlate with the cognitive profile of DS patients (135, 277, 278, 289), thus leading to the idea that DS cognitive dysfunction is mainly cortical. In support of this hypothesis, functional data exploring intellectual functioning, impaired language ability, and impaired adaptative behavior in DS indicate possible deficits in specific corticohippocampal circuits, also involving regions of the prefrontal and temporal cortices and the cerebellum (see next section) in the phenotypes. Interestingly, in the last few years, imaging studies have suggested that these brain structures, especially the frontal and temporal cortices, are under significant genetic control (45, 134, 358, 359), a fact that may explain the damaging effects of gene deregulation on the integrity of the cerebral cortex. Although this structure-function relationship at the macroscopic level is obviously important, we cannot overlook the fact that the mammalian central nervous system, and particularly the neocortex, contains an enormous variety of cell types, each with unique morphology, connectivity, physiology, and function so that each cortical region comprises a highly complicated mix of distinctive cell types. Since functionally discrete brain regions are delineated by gene expression patterns, probably genes with regionalized expression patterns will provide potential substrates for functional differences across brain regions and may be potential candidates for specific dysfunctional cognitive patterns (34, 339).

For these reasons, to get real insight into cognitive dysfunction, it will be necessary to reveal the contributions of genes with highly specific cortical expression patterns restricted to discrete neuronal populations in different cortical layers related to higher-level brain function and disease-related dysfunction. Thus the characterization of the full range of neural cell types is essential to understand the impact of gene expression deregulation on functional neural circuit properties and their relation to higher cognitive functions and behaviors. To date, neural expression profiling has lacked the power to detect effects that involve subsets of cells, since the techniques used for expression analysis have typically been applied to large brain regions. Such data are difficult to interpret because they do not resolve specific brain subregions or even at a lower level, cellular diversity within those structures (184, 198, 234). To address this problem, some projects (The Allen Brain Atlas project:; EurExpress:; 213) have recently taken a global approach to understanding the genetic structural and cellular architecture of the mouse brain by generating a genome-scale collection of cellular resolution gene expression profiles using in situ hybridization across 20,000 genes.

A. Neurocognitive Profile in Down Syndrome

Although the clinical presentation of DS is variable, 100% of people with trisomy 21 present with mental retardation. Mental retardation is defined as a developmental disability with onset during childhood, characterized by significant impairment of intellectual functioning and adaptive skills causing major limitations to living a normal, independent life [Diagnostic and Statistical Manual of Mental Disorders (DSM-IV); Ref. 280]. In clinical practice, the limitations in intellectual functioning are defined by an intelligence coefficient (IQ) below 70 (two standard deviations below the average IQ of 100). In DS patients, mental retardation is highly diverse in terms of the severity of the cognitive disability as well as the manifestation of additional (noncognitive) symptoms. IQ in people with DS is usually in the severely to moderately retarded range (IQ = 25–55) (133, 313, 370). IQ in DS is not constant across the life (35), but it progressively decreases with age (273, 370), although this scenario has changed in the last years thanks to the early intervention programs and the social integration of these persons (284). Speech and language deficits, well documented in DS, probably contribute to the IQ decline with age. In DS adults, IQ may be also influenced by the increased risk of early-onset dementia of the Alzheimer type, which is highly prevalent among DS individuals (39). The intellectual disability in DS is associated with a more or less fixed constellation of morphogenetic and functional central nervous system manifestations, such as craniofacial and brain malformations, and neurological or psychiatric symptoms.

Although extensively used in clinical practice, the IQ measure is too vague to be useful for establishing a genotype/phenotype correlation, since it does not inform about the specific functional domains affected, and thus it does not allow predicting the structures involved in the phenotypes or their underlying pathogenetic mechanisms. Establishing specific diagnostic criteria and using a battery of evaluative instruments to document, for example, verbal fluency, abstract reasoning, short- and long-term memory, skill learning, etc. is crucial, but also an extensive and costly process, and few, if any, psychometric studies actually attain this. The diagnosis of the specific DS cognitive deficits is even more challenging since it requires a combination of psychometric tests and neuropsychological paradigms, since low baseline levels of cognitive ability, impaired communication, and attention can affect many psychometric tests (38, 77).

By definition, mental retardation implies a delay in typical development, but in DS, this is an oversimplification of the problem. As in typically developing children, IQ in DS is influenced by genetic and environmental factors (272) so that a positive correlation is found between parental and DS children IQs (102). However, the behavioral phenotype is a potentially critical tool for shaping IQ during the first years of life, and for planning education and cognitive intervention in this population (117). In the DS adults, the neuropsychological profile is not homogeneous, with some abilities more impaired than others. In particular, motor function, language (specifically morpho-syntax), verbal short-term memory, and explicit long-term memory are usually more impaired, while visual-spatial short-term memory, associative learning, and implicit long-term memory are relatively preserved (37, 50, 57, 92, 163, 257, 289) and DS children have specific impairments in receptive language that exceed their impairment in other cognitive abilities (53). We still need to address the question of which traits are specific for DS (94, 163, 164) and which psychometric and functional domains are more impaired in each DS individual (see Ref. 1). The fact that not all skills are affected to the same extent in all persons with DS suggests that either the impact of aneuploidy on brain structures is different among DS individuals or that the functional consequences of relatively homogeneous structural changes may not affect all DS patients equally (37, 57, 69, 204, 209, 273, 371). To increase the complexity of this scenario, individuals with DS may have additional behavioral and psychiatric problems, such as disruptive behavior, attention deficit, hyperactivity disorder (6.1%), conduct/oppositional disorder (5.4%), or aggressive behavior (6.5%); 25.6% of adults with DS have a psychiatric disorder, most frequently a major depressive disorder or aggressive behavior (see Ref. 303 for review).

B. Genetic Shaping of the Down Syndrome Brain

There are still unanswered fundamental questions about the genetic shaping of the DS brain. How are specific structural pattern alterations under genetic control linked to measurable differences in cognitive function? Does aneuploidy affect cognitive phenotypes via disruptions of embryonic development or by altering function during cognitive processing (or both)? How does genetic variability relate to variations in cognitive profiles in people with DS? If specific genes have specific, separable effects on specific cognitive processes, what are the longitudinal effects on cognitive profiles for DS people carrying particular risk alleles? Can the genetic information be correlated with data from structural or functional neuroimaging? Does the genetic profile influence the response of an individual to particular behavioral interventions, and could this help to target therapies based on an individual's genetic information?

The main problem to answer these questions is that during many years, it has been assumed that mental retardation was similar across different neurogenetic syndromes, with different morphometric alterations in brain. The grounds for this generalization rest partially on the fact that for many years the assumption was that we could group mentally retarded people on the basis of nonspecific psychometric measurements, such as IQ (see previous section), and thus the intellectual disability of each syndrome was considered by many researchers to be similar. It is reasonable to expect that specific, well-defined neurobehavioral phenotypes as well as the definition of intermediate phenotypes will be good tools for gene discovery because they will reduce phenotypic heterogeneity and help to determine the mechanistic dependence of phenotypes. At present, we have to mainly rely on studies in DS mouse models (see sect. iv), that have provided some cues on the molecular and cellular mechanisms causing developmental and/or functional alterations in neural patterning that may lead to defective neuronal circuits responsible for the pathogenesis of mental retardation in DS. These studies have provided evidence that overexpression of specific genes or gene networks leads to the disturbance of specific biological processes, thus supporting the idea of dosage-sensitive phenotypes. However, increased gene expression varies for each gene and in each cell type, and cooperative functional interaction among HSA21 genes in a signaling pathway may result in additive or synergistic effects, leading to signal amplification. On the other hand, the effect of overexpression of one gene may counteract the detrimental effects of others acting in the same pathway, thus compensating the phenotype (76). It has been proposed that genes having a functional target that affects information processing, such as synapses and synaptic function, will lead to similar alterations in cognition and adaptation (402).

Finally, an important factor may be the genetic context in which an individual's mental retardation is taking place and that can affect the final outcome. This is well known in experimental models, as shown for example by the differences in the performance on learning tasks or in the amount of cell proliferation and neurogenesis among different mouse strains such as C57BL/6, BALB/c, CD1 (ICR), and 129X1 mice (320, 377). In contrast, some phenotypic traits (such as motor phenotypes, or memory) are maintained even after backcrossing to obtain a homogeneous background genotype in DS mouse models (M. Dierssen, M. Martinez de Lagrán, and G. Argué, unpublished observations). For decades, researchers and practitioners have attempted to find evidence for a personality stereotype in individuals with DS that includes a pleasant, affectionate, and passive behavior style. In fact, individuals affected with DS present some features that are common regardless of their particular genetic background, but other features that may be dependent on their unique genetic background.

C. Brain Topology of the Cognitive Impairment in Down Syndrome

An intimate relationship exists between domains of human cognition and functional brain architecture; thus the neuropsychological profiles described in DS patients are assumed to be due to altered brain structure, presumably derived from anomalous brain but also cranial development (Fig. 3). This relationship became clear when fMRI studies enabled direct visualization of the functional architecture of the brain networks during the performance of cognitive tasks, thus illuminating the neurophysiological underpinnings of the alteration of cognitive functions in DS.

FIG. 3.

Morphometrical characteristics of Down syndrome (DS) skull and brain. A: lateral views of euploid and DS skulls. Left: the face of an euploid young adult presents downward growth of the maxillae; therefore, the distance between the inferior orbital ridge, the nasal spine, and the alveolar crest is considerably increased. The mandible is angled. Right: although the DS skull grows to nearly the same size as the normal adult, it presents brachycephaly, which means “short headed,” and occurs when the right and left coronal sutures close prematurely. Brachycephaly results in an abnormally broad head with a high forehead. It is often associated with other craniofacial abnormalities. In DS, the face is small, with underdeveloped maxillae, and the mandible is still relatively straight. (Redrawn from Benda CE. Down's Syndrome: Mongolism and Its Management. New York: Grune & Stratton, 1960.)

From the strictly structural point of view, postmortem observations and volumetric MRI studies in people with DS have documented reduced brain weight, with particularly small cerebellum as well as frontal and temporal lobes (390), reduced overall brain volumes and brachycephaly, with disproportionately smaller volumes in frontal and temporal areas (including uncus, amygdala, hippocampus, and parahippocampal gyrus) and cerebellar regions (277, 278). In contrast, DS brains usually show relatively preserved volumes in subcortical areas (183, 277, 278), such as lenticular nuclei (260) or the posterior (parietal and occipital) cortical gray matter (278). The relatively large size or preservation of these structures in subjects with DS, in the context of significantly smaller overall cerebral volumes, suggests that there is a dissociation of the development of cortical versus subcortical regions.

To what extent will these structural alterations produce the functional consequences of DS? Neural network models of the different brain structures have identified cognitive operations that are unique to each structure (268), thus allowing predictions on structural-functional maps in DS. For example, the lower performances of DS in linguistic tasks could be partially explained in terms of impairment of the connectivity of frontocerebellar structures involved in articulation and verbal working memory (107), whereas the reduced long-term memory capacities may be related to the temporal lobes and, specifically, to an hippocampal dysfunction (273). Similarly, the functional dissociation between implicit and explicit memory in DS may be due to the severe cerebellar hypoplasia along with normal morphology of basal ganglia that is observed in these patients (183). Here we will focus our review on only some of the structures that have been shown to participate in the cognitive alterations of DS patients.

1. Hippocampus

One important aspect of mental retardation in DS is the specific alteration of memory patterns. The hippocampus and related structures of the medial temporal lobe have long been linked to the ability to learn and retain new memories for facts and events. This ability is termed declarative memory in humans (254, 346) and relational memory in animals (95). Evidence of the participation of the hippocampal formation in the mnemonic process derives from studies demonstrating that damage to these structures in adult humans and animals causes a profound loss of declarative memory function without other sensory, motor, or cognitive impairments (246). Indeed, episodic memory is sensitive to hippocampal damage, and the retrieval of autobiographical memories activates the hippocampus in adults (231, 403).

The capacity for this basic, intrinsic ability to form new memories is ultimately genetic in origin, and human disorders that are associated with mental retardation usually present functional alterations in the basic mechanisms involved in the formation of memories. This is also the case in DS, although there have been relatively few studies addressing hippocampal functioning in these patients. In the context of the overall cognitive dysfunction, a common finding in DS is abnormalities in some forms of visual-spatial memory and contextual learning that depend on the functional integrity of the hippocampus (for review, see Refs. 128, 366). In a recent study, Carlessimo et al. (48) described a hippocampally mediated deficit in episodic memory, that particularly affected encoding and retrieval, although there was a discrepancy in the performance level achieved by DS individuals in the visual and spatial domains. Specifically, they had reduced visual but relatively preserved spatial abilities (372). The most extensive study to date showed clear deficits in all hippocampally mediated tasks in DS patients, most strikingly those related to long-term storage of explicit memories (273). Similar to these findings, DS preschool children showed deficits in a place learning and recall task, being more severely impaired in delayed recall. This deficit in delayed recall in DS may be explained by a specific dysfunction of the hippocampus disrupting remote information, since hippocampal memory systems show a clear temporal gradient with a gradual strengthening of memories over time (e.g., Refs. 195, 258, 385, 403). Presumably, once information has been processed by the hippocampus, it is transferred to neocortical association areas for further processing and permanent storage (e.g., Refs. 258, 356). The hippocampal abnormalities described in DS patients are suggested to be syndrome specific, since they are more severe in individuals with DS than in other mental retardation conditions (48, 273, 373).

Evidence for neuroanatomical abnormalities involving the hippocampus in DS derives from neuroimaging studies. The hippocampal formation is comprised of a group of cortical regions located in the medial temporal lobe that includes the dentate gyrus, hippocampus, subiculum, presubiculum, parasubiculum, and entorhinal cortex. Moreover, the regional specialization of the hippocampus in the CA1, CA2, CA3, and dentate gyrus areas reflects differential gene expression (10, 182, 223), where a substantial proportion of neural genes (213) show region-specific expression. In individuals with DS, the volume of the hippocampus, adjusted for brain size, is smaller than in controls, whereas amygdala volume reductions do not exceed the overall reduction of brain size (16, 203, 277). Interestingly, the volume of the parahippocampal gyrus (i.e., perirhinal and entorhinal cortices) that appears to serve distinct and important memory functions (e.g., Refs. 96, 97, 153), such as nonrelational memory (i.e., stimulus-stimulus associations and stimulus recognition; Ref. 159), is relatively larger in DS (191, 289). Although there is no relationship between total brain size and the cognitive deficits, general intelligence and mastery of linguistic concepts correlated negatively with the volume of the parahippocampal gyrus in DS patients (289).

The reduced volumes of these regions may depend on the reduced neuronal densities and reduced dendritic branching, length, and spine densities that are found in the hippocampus and parahippocampal gyrus of DS (reviewed in Refs. 29, 146, 257). Analysis of cell phenotype showed that DS fetuses had a higher percentage of cells with astrocytic phenotype but a smaller percentage of cells with neuronal phenotype, along with less proliferating cells in the germinal zones of the hippocampus and parahippocampal gyrus and higher incidence of apoptotic cell death. Similar findings have also been reported in DS mouse models (for a review, see Ref. 129 and sect. iv).

2. Cerebral cortex

The analysis of genetic cortical expression maps has revealed a strong relationship between gene expression, cortical structure, and behavior, suggesting that heritable factors affecting the development of this brain region may be fundamental in determining individual differences in cognition (45, 359). As mentioned above, DS is categorized as a disease that affects the integrity of the cerebral cortex, since trisomy of HSA21 affects more the neocortical areas, with decreased neuronal density within all layers of cortex (178, 235, 236, 287, 353, 354, 379, 389, 392) and functional differences in the cortices of individuals with DS compared with euploids widely reported (84, 101, 121, 168, 216, 270, 319, 323).

The neuropsychological profiles described in DS do reflect these neocortical alterations, but might also result from the disrupted balance in cortical and subcortical structures. One of the classical neocortical domains of function is executive function, that includes planning ability, cognitive fluency, and judgment, and is attracting attention as a basic mechanism underlying pervasive developmental disorders. Some studies have found impairment on some tests of executive function in DS individuals as shown by the greater weakness in a dual-task processing component of executive function (197, 309). With the use of the Complex Figure Drawing Test (296), visual-motor (hand-eye) coordination difficulties and perceptual problems are frequently found in individuals with DS (Fig. 4). This drawing and visual memory test examines the ability to construct a complex figure and remember it at later recall. It measures memory as well as visual-motor organization. The results obtained using this test supported the possibility of a different pattern to drawing development in DS rather than a delayed version of typical development (209, 210). In these studies, many of the children with DS showed poor understanding of spatial concepts, which was related to grammar comprehension. Those individuals with very poor understanding adopted inconsistent strategies to represent the different spatial arrays, and children using a developmentally more advanced strategy demonstrated superior pencil control.

FIG. 4.

Executive cognitive function in DS: the Complex Figure Drawing Test. Visual-motor, hand-eye coordination difficulties, and perceptual problems are frequently found in individuals with DS. Examples are shown of drawings of a 12-yr-old (A) and a 16-yr-old (B) DS girl of the Complex Figure Drawing Test (296) that examines the capacity to integrate stimuli on the levels of perception, spatial organization, and planning, in a visual-graphic modality. The subject is first instructed to copy the figure, which has been so set out that its length runs along the subject's horizontal plane. The examiner watches the subject's performance closely. Each time the subject completes a section of the drawing, the examiner hands him a different colored pencil and notes the order of colors. Time to completion is recorded, and both test figure and the subject's drawings are removed. This is usually followed by one or more recall trials. Children with DS show perceptual and fine motor difficulties and impaired capacity for integration in the spontaneous first copy. The vertical drawing (A) reflects a more concrete way of perceiving the complex figure that has no realistic meaning, similar to the way most young children draw. The poor first memory drawing shows that this tracing, copying did not produce any learning. However, the learning phase resulted in quite accurate copy and drawing from memory. C: complete figure with configural elements, clusters, and details. (Adapted from∼altonweb/cs/downsyndrome/i).

Frontal and parietal brain cortical regions are the primary areas involved both in working memory and visual attention, functions that allow us to respond efficiently to the varying environmental demands of everyday life (17, 18, 276). Several recent observations suggest that the dorsal areas of the parietal cortex are involved in spatial processing, whereas the ventral temporal (and perhaps frontal) areas participate in working memory for objects and faces and, more generally, in the processing of visual material (65, 259). On the basis of the results of psychometric studies, which show that persons with DS perform better on visual-object than on visual-spatial memory tests, it appears that the dorsal component of the visual system in DS individuals functions closer to normal levels than does the ventral component (371). Impairment in verbal short-term and working memory, measured by digit or word span, has also been documented in individuals with DS compared with mental age-matched controls (for a review, see Ref. 373). This impairment can be categorized as a type of working memory, defined as a limited capacity system for the temporary storage of information held for further manipulation (17, 18).

Pinter et al. (278) showed that among individuals with DS cerebral lobe volumes, reductions were predominant in frontal and occipital lobes. However, these effects were not significant after adjustment for overall brain size, suggesting that the relative preservation of temporal lobe volume was related to a significantly larger white matter volume. A note of caution should be taken in interpreting these studies, since most of them used a normal brain template that might have introduced a bias in defining the lobe borders on the differently proportioned brains of the DS subjects.

Another important function that relates to the cortical function and is altered in DS is language. The processing of language information relies on a cortical network that comprises inferior frontolateral and anterior as well as posterior temporal lobe structures in both hemispheres, and it has been described that the grey matter density in the anatomical region that includes frontal and language-related cortices is more similar in genetically related individual. Of particular significance for language processing is the finding that the superior temporal gyrus appears to be severely narrowed or underdeveloped in DS brains (136, 139, 390). In this brain region, the planum temporale occupies the superior temporal plane posterior to Heschl's gyrus, which is generally agreed to represent auditory association cortex. In the left hemisphere, Wernicke's area includes part of planum temporale and has traditionally been viewed as a language processor. The planum temporale has been implicated in the representation and updating of auditory speech traces that are necessary for phonological working memory and speech production. It is also activated by natural speech but not by acoustically similar non-speech sounds, by deviant or unpredictable verbal and nonverbal events, or by verbal self-monitoring (for review, see Ref. 144). In people with DS, the planum temporale has been shown to be smaller than normal, although it has not been possible to establish a correlation across individuals between this deficiency and degree of language deficit (123).

Language deficits in individuals with DS are evident at the onset of language use and continue through adolescence and into adulthood with some indication of changes in language strengths throughout their life span. Potential explanations for the specific difficulties in language include deficits in phonological working memory, hearing sensitivity, and visual short-term memory, which are related to comprehension and production of both syntax and vocabulary (51, 52, 54, 210). Difficulties in oral language skills and auditory working memory such as those reported in DS individuals are related to phonological awareness and decoding and comprehension of reading material (41, 210, 369). Recent brain imaging studies on language processing (with auditory stimulation) have shown that temporal and frontal areas of both hemispheres are involved in the processing of connected speech, with the left hemisphere responsible for the bulk of on-line processing of syntactic features and prosody (125, 245). Thus the lower performances of DS individuals in linguistic tasks may be also explained in terms of impairment of the frontocerebellar structures involved in articulation and verbal working memory (107).

Some regions involved in language, such as the temporal lobe, may be potentially more susceptible to subtle gene-dosage effects. Early histological analyses of DS described a decrease in neuronal cell packing density particularly in layer III of the cerebral cortex (58, 352, 393). Moreover, the histogenetic development of the cerebral cortex is affected, and difficulties in recognizing cortical layers and cytoarchitectonic areas in DS brain samples have been reported (190, 401). Wisniewski et al. (391), in a study of the visual cortex of 60 children with DS from birth to the age of 14 yr, reported that DS patients had 20–50% fewer neurons than normal children during this period. They suggested that the neurons were rearranged before birth, particularly in layer IV. Golden and Hyman (136) plotted a set of reference curves for control fetuses, reflecting cell density in the cerebral cortex throughout cortical development. They suggested that neuronal migration is normal, but that the second phase, after 20–21 wk and corresponding to lamination, is both delayed and disorganized in patients with trisomy 21. The observed pattern of cortical maturation may also reflect an abnormality in axonal and dendritic arborization. Normal children display expanding dendritic arborization during childhood, whereas patients with DS display greater dendritic branching and total dendritic length than controls in the infantile period (up to the age of 6 mo), decreasing steadily thereafter to significantly below normal levels in the juvenile group (over the age of 2 yr; Refs. 23, 25). An interesting finding is the abnormal rate of development of minicolumns in children with DS. Anatomical minicolumns represent a basic functional unit of the cortex that have been closely associated with physiology (110, 111, 255, 360) and are related to information processing capacity. Minicolumns are among the structural features that are genetically regulated they may underlie the differences in cortical function and IQ observed among DS individuals. Buxhoeveden et al. (40) found that DS patients had larger columns with lower cell density than normal individuals, indicating a reduction of complexity due to abnormal development.

Even though numerous genes have been identified with highly specific patterns of gene expression restricted to discrete neuronal populations in different cortical layers, their specific function has not yet been elucidated. Moreover, DS-associated behaviors are not easy to trace to specific molecular elements or interactions in the brain, and it is even more difficult to dissect a complex cortically dependent cognitive function such as language and to relate it to a particular genetic cause.

3. Cerebellum

The cerebellum has classically been related to motor control, specifically motor learning and coordination. However, there is increasing evidence that the cerebellum has a role in cognition, although the observed role with respect to changes in working memory, affect, and language production is mostly modest (see Ref. 200 for review). Motor behavior disturbances are part of the clinical features of DS and have classically been attributed to hypotonia in the early developmental stages. Shumway-Cook and Woollacott (341) contested this hypothesis and proved that balance difficulties were not caused by altered motoneuron function and stretch reflex mechanisms, but rather by defects within higher level postural mechanisms mediated by the cerebellum. From the neuroanatomical point of view, reductions in cerebellar volume are a constant finding in DS individuals (183), and the growth of the DS cerebellar field is smaller in the sagittal and vertical directions than in normal fetuses (222). However, although cerebellar volumes are disproportionately small in individuals with DS, they do not further diminish significantly with age and do not undergo age-related atrophy that is different from that of normal controls. Volume reduction in the cerebellum does not appear to be specifically responsible for the age-related decline in fine-motor control that is observed in adults with DS (16). However, both neuropsychological (250) and functional neuroimaging (368) data assign a critical role to basal ganglia and cerebellum in the implicit learning of visual-motor skills. High-resolution three-dimensional MRI revealed that the reduced cerebellar volume is a completely penetrant phenotype that is accurately recapitulated in a mouse model (22) bearing segmental trisomy for MMU16 (Ts65Dn; see sect. iv).

D. Nature and Nurture in Down Syndrome: The Building of a Trisomic Brain

DS is a neurodevelopmental disorder that combines genetic effects on different cells, structures, and functions throughout development and in the adult. Normal development of the central nervous system depends on complex, dynamic mechanisms with multiple spatial and temporal components during gestation. Neurodevelopmental disorders originate during fetal life from genetic as well as intrauterine and extrauterine factors that affect the fetal-maternal environment, and the later can also affect the final outcome of genetically driven disorders. Fetal neurodevelopment depends on cell genetic programs, developmental trajectories, synaptic plasticity, and maturation, which are variously modifiable by factors such as stress, exposure to teratogenic drugs, maternal teratogenic alleles, or premature birth. There is a clear need to expand our understanding of how, when altered, prenatal factors may lead to disordered development, the signs of which may not appear until long after birth, and this is especially important in DS, since it will define the best moment for therapeutic intervention. The postnatal gene-environment interaction also modifies the final structural and functional outcome so that the functional profile observed in individuals with DS emerges as a result of the cross-domain relations between both. Moreover, during the first years of age, more primary (cognitive, social-emotional) aspects of the DS behavioral phenotype interact with a general profile of delays in the development of instrumental thinking coupled with emerging relative strengths in social-emotional functioning.

The brain is an extremely complex structure in which different parts serve different functions and arise from distinct cell populations across different temporal windows of development. The coordination of neurogenesis, growth, and development of each brain component is controlled by precisely localized patterns of gene expression (e.g., Refs. 2, 225). Many primary deficits may have cascading effects that produce clinically observed phenotypic traits. Brains of patients with DS present specific features that may be dependent on morphogenetic alterations, but also on a reduced remodeling potential and impaired neural plasticity. The first include neuronal heterotopias, abnormal neuronal migration/differentiation, and decreased neuronal density, that mainly affect specific cell populations such as granule cells in cerebral cortices and dendritic anomalies that affect pyramidal neurons (23, 235, 236, 355; see Ref. 120 for review). Thus a first step in DS research is to separate developmental effects from functional perturbations to the adult brain, since altered functions in adults may be the consequences of a developmental error caused by trisomy or to the functional effects of upregulation of trisomic genes.

Obviously, the developmental aspects are very important in defining the most important part of the functional consequences of the genetic imbalance in DS at the cognitive level. Individuals with DS have a wide range of dysfunction in all areas of psychomotor development. As highlighted before, in addition to the underdevelopment typical of early infancy, the IQ of DS patients decreases in the first decade of life, thus indicating that the activity-dependent maturation of the central nervous system is compromised. DS children present asynchronic development in functional areas different from typically developing children, as they develop with a totally different biological background, so that the organic impairments and pathologies associated with DS can be determinant in limiting development (257). As an example, DS children do not appear to follow the same sequence of stages as typically developing children with respect to certain domains, such as language development or object concept development. This constitutes a crucial aspect that must be taken into account when determining the developmental versus functional causes of the adult dysfunction. For example, a deficiency in language production relative to other areas of development often causes substantial impairments that affect other domains. Moreover, motivational aspects, derived from increased failure rates for learning experiences, clearly have an impact on further learning through the establishment of counterproductive learning strategies, which include avoidance strategies when faced with cognitive challenges. Finally, the morphogenetic alteration affecting different brain regions may have a convergent impact so that difficulties experienced when acquiring new skills will have an impact on the success of learning in different areas (162, 165). As an example, delayed recall deficits are mostly caused by hippocampally related functions such as memory acquisition or storage, but are also affected by cortical functions, such as reduced cognitive flexibility manifested as the persistent use of old strategies to solve new problems (162, 165, 289, 388).

Analysis of the developmental process is even more complicated if we consider the structural aspects. In DS fetuses, differences in brain size become apparent as early as 15 wk of gestation, but increase significantly after birth. The frontothalamic distance on ultrasound scans is 5–10% lower than that of normal fetuses, and more than 50% of DS fetuses have frontothalamic distances below the 10th percentile (19, 387). DS has a stronger negative effect on cerebellum growth in the last part of gestation and after birth, resulting in 30% size reduction in children with DS compared with euploid subjects (183, 278). Studies of the weight of the infratentorial part of the brain (including the brain stem and cerebellum) have shown that the growth of this region is less restricted than that of the supratentorial part of the brain, between 15 and 38 wk of gestation (147). However, the only part of the cerebrum showing a clear decrease of magnitude is the hippocampus (27% decrease). Since the size and shape of brain components depend on the production of different neural and nonneuronal cells, changes in early morphogenesis of a specific region can create a localized change in volume, resulting in associated displacement of nearby structures, producing both localized and global differences in shape. Still, changes in overall brain volumes and brain shape are not necessarily correlated across individuals, suggesting that size and shape phenotypes provide complementary information about the brain's developmental history (7).

One important aspect when considering the developmental process is the proliferation of neuronal cells. In DS, as discussed previously, the number of neurons is significantly reduced in adult brains. This has received special attention, since for many years the general assumption was that number of neurons was correlated with cognitive proficiency, based on the fact that mental retardation was frequently associated with microcephaly.

Even though we now are aware of the wide constellation of factors involved in the definition of cognitive abilities, neuronal proliferation still deserves much attention, since proliferation of neuronal progenitor cells occurs at a significantly lower rate both in human DS fetuses and in trisomic Ts65Dn mice than in healthy controls (62, 312; see sect. iv), suggesting an impairment of neural precursor proliferation during early phases of brain development. The number of surviving cells is also smaller compared with euploids, and a higher proportion of the surviving cells in DS individuals had an astrocyte phenotype. Several studies have pointed to a defect in cell cycle progression as a likely determinant of impaired proliferation and have highlighted some genes as possible contributors to cell cycle and replication deregulation. For example, DS fetuses and Ts65Dn mice show a greater proportion of cells in the G2 phase of the cell cycle and a smaller number in the M phase, suggesting that the longer time spent by proliferating cells in the G2 phase may underlie their reduced proliferation rate. In Ts65Dn mice, significant downregulation of cyclin B1 and the cell cycle kinase Skp2, involved in the regulation of G2/M progression, and of Bmi1, a gene involved in maintenance of the proliferative capacity of progenitors cells, has been reported. In addition, an upregulation of genes involved in the exit from the replicative state and induction of differentiation towards either a neuronal (Mash1, NeuroD, Neurogenin2) or a glial (Olig1, S100b) fate was found (60, 62). These observations suggest that neurogenesis impairment starting from the earliest stages of development may underlie the widespread brain atrophy of DS and the delayed and disorganized emergence of lamination in the fetal DS cortex (136). The transition of neuronal precursors from a proliferating state to a completely differentiated phenotype represents one of the most crucial events of the central nervous system developmental program. In a relatively short time, neuroblasts permanently lose their ability to reenter the cell cycle, and they migrate to their definitive positions and undergo major morphological changes that establish the neuronal network. Candidate genes such as the kinase encoding Dyrk1A gene have been related to this particular developmental stage (see Ref. 86 for review) as shown by its transient expression in neural progenitor cells during the transition from proliferating to neurogenic divisions in early vertebrate embryos (149, 150).

With regard to the neuronal differentiation phase, postmortem studies show that DS persons start their lives with an apparently normal neuronal architecture that progressively degenerates. During the peak period of dendritic growth and differentiation (2.5-mo-old infants), no significant differences were detected in dendritic differentiation between euploid and DS cases in layer IIIc pyramidal neurons of prefrontal cortex (prospective area 9, 376). Similarly, DS infants younger than 6 mo showed greater dendritic branching and length than normal infants in both apical and basilar dendrites (24, 353). Thus normal or even increased dendritic branching in the DS fetus and newborn contrasts with the reduced number of (especially apical) dendrites and degenerative changes in DS children older than 2 yr (21). Similar findings of decreased values of dendritic parameters could also be observed only after the fourth postnatal month in the visual cortex of DS brains. Finally, abnormalities of synaptic density and length, fewer contact zones, a reduction of dendritic spines, and shorter basilar dendrites are observed in DS (23, 123, 232, 283, 350, 353355). When the mechanism of these alterations is determined, the dynamic reorganization of the actin cytoskeleton turns out to be a very good possibility, since it plays a crucial role during every stage of neuritogenesis and structural plasticity (71, 226). Particularly, in differentiating neuroblasts, actin rearrangements drive the extension and the path finding of axons and dendrites and contribute to the establishment of synaptic contacts (71, 90, 226). Molecules affecting actin cytoskeleton structural or functional properties are thus excellent candidates to explain the neuronal architecture abnormalities observed in DS. Several candidate genes located in the HSA21 DS critical region have been proposed to be involved in synaptic plasticity and in particular in dendritic function and spine motility and plasticity. The DSCR1, DYRK1A, or ITSN1 genes may be candidates to explain the dendritic spine functional and structural alterations. Intersectins (ITSN1 and ITSN2) (285, 286) are multidomain scaffold proteins that interact 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. Dyrk1A phosphorylates actin-binding proteins and may also have a role in shaping the interaction of the spine membrane with the actin cytoskeleton (see Ref. 86). Yang et al. (397) showed that overexpression of a kinase-deficient form of Dyrk1A attenuates the neurite outgrowth induced by a neurogenic factor in immortalized hippocampal cells. Moreover, the reduction in brain size and dendritic defects observed in mice lacking one copy of the murine homolog Dyrk1A (Dyrk1A+/−) supports the idea that this kinase may have a physiological role in neural differentiation (29, 122) and that deregulation of its expression may inhibit neuritogenesis. In this regard, overexpression of Dyrk1A in several DS models, from segmental trisomic mice to BAC models (for review, see Refs. 4, 86), correlates with changes in size of specific regions of the brain and with microstructural alterations at the dendritic level (34; Dierssen et al., unpublished observations). The DSCR1 gene is overexpressed in fetal and adult DS brains, and many experimental data are consistent with calcineurin activity inhibition by DSCR1 overexpression (46). Calcineurin is involved in synaptic plasticity and has a role in the transition from short- to long-term memory through perturbation of long-term potentiation (LTP) and long-term depression (LTD). In addition, calcineurin-dependent induction of the gene product of DSCR1.4 may represent an important autoregulatory mechanism for the homeostatic control of nuclear factor of activated T cells (NFAT) signaling in neural cells (46). Another interesting example of synaptic-related proteins involved in DS is drebrin, an actin-binding protein, thought to regulate assembly and disassembly of actin filaments, thereby changing the shape of spines and the synaptosomal associated proteins SNAP and SNAP 25 (15, 158). In postmortem DS brains, levels of drebrin are reduced in the early second trimester. Also, overexpression of Dyrk1A is associated with altered synaptic plasticity and learning and memory deficits, that could also deregulate the NFAT pathway (14). Finally, S100β, a calcium binding protein found in astroglial cells, has extracellular neurotrophic effects involving the neuronal cytoskeleton, and thus its overexpression may be related to the cytoskeletal abnormalities seen in mental retardation (330). 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.

After the first genetically driven developmental period, neuronal circuits are continuously modified and shaped by experience (epigenetic development) to ensure that the neural control of behavior is flexible in the face of a varying environment. To this end, morphological and physiological changes are possible at many levels, including that of the individual cells. According to current concepts, long-term memory is based on structural-functional changes in particular synaptic connections between neurons in the brain (synapse-specific plasticity), which depend on local translation and transcription of specific genes, so that stimulation of a synapse will lead to activation of intracellular second messengers in the synapse as well as to “synaptic tagging,” involving the recognition of specific transcription products. In the neuronal body, second messengers induce the synthesis of RNA and protein molecules, which are widely distributed in neuron processes and which are inserted selectively only into stimulation-tagged synapses, causing long-term changes in their functional and morphological characteristics (reviewed in Ref. 262) that will affect neural plasticity and thus adaptation to the environment.

In this regard, research addressing early intervention in DS showed positive effects in different developmental domains but with unclear evidence of long-term effects, probably reflecting an impaired structural plasticity in DS persons, and also DS mouse models (85). These data support the emergent recognition of the importance of the context of child development and is already having consequences in the development of a contextualized approach to child development, particularly in relation to the impact of early intervention on families and the long-term goals of early intervention.


To better understand the molecular and cellular origin of DS as well as the defects associated with the disease, several groups developed a large panel of mouse models. The advantages of modeling DS in mouse are clear. First, restricting the comparison of individuals either carrying a trisomy or not, but with a similar genetic background and that are bred in a controlled environment. Second, the impact of gene-dosage can be studied further during embryonic and postnatal development or in the adult, allowing easier discrimination of the effect of gene imbalance during organogenesis from that on the function of the adult organ, even though both events are interconnected (78, 305). Several models developed through transgenesis for single or multiple genes using BAC or YAC have been already extensively described elsewhere (13, 87). In this review, we will concentrate on the mouse models of DS that correspond to genetically engineered trisomy.

The laboratory mouse is a powerful model system, with well-known biology and genetics. In addition, mouse and human share many homologies at the genetic and the physiological levels so that conclusions drawn in one species can be used to approach similar issues in the other. The sequencing of both genomes by the beginning of the 21st century confirmed their syntenic conservation, and ∼80% of murine genes are homologous to human (381). This similarity is lower for the homologous region to the HSA21 (131, 299, 337). More than 430 genes, including RNA coding genes, are identified on HSA21, and 293 homologs are found in the mouse genome, spread across three chromosomes (; see Fig. 5). For example, a large fragment from Lipi to Zfp295 encompassing 37 Mb and 224 genes is located on the telomeric part of mouse chromosome 16 (MMU16; MMU for Mus musculus). The most telomeric part of HSA21 is split between MMU17, with 22 genes found between Umodl1 and Hsf2bp, and MMU10, where 47 genes lie in the interval between Cstb and Prmt2. Interestingly, the order and the relative orientation of the genes known to be shared between mice and humans have been conserved together with 2,261 conserved nongene sequences, of which only 60% is predicted to have a function (81, 82, 357). On the basis of this knowledge, several groups have focused on mouse models to decode the cellular and molecular mechanisms linked to the neuropathological, cognitive, physiological, and morphological alterations found in human DS patients.

FIG. 5.

Genetic regions triplicated in DS mouse models. The relative position of the mouse homologous regions on MMU16, -17, and -10 to HSA21 are shown with the genes located at the borders of the genetic interval. The human transchromosomic, Tc1, lines carrying a complete HSA21, described by O'Doherty et al. (263), are indicated with two internal deletions shown by the broken line. The other mouse models are shown with the corresponding region that is either segmentally duplicated [Ts1Yu (215); Ts1Rhr (265)] or translocated [Ts65Dn (292); Ts1Cje (314)].

A. The Early Phase

In 1980, the first publication describing the linkage of the superoxide dismutase (SOD) gene to MMU16 concluded by suggesting the opportunity of using mice trisomic for MMU16 as a model for DS (66), although this trisomic model encompasses a large set of genes not homologous to HSA21, and it leads to death just after birth. Consequently, the description in the 1990s of Ts(1716)65Dn (noted Ts65Dn) mice opened new perspectives and increased our understanding of the mechanisms underlying DS (75, 292). Ts65Dn mice carry a translocation of the distal part of MMU16 on a small centromeric fragment of the MMU17, which corresponds to a trisomy of 167 genes between Mrpl39 and Zfp295 (Fig. 5; Table 2). This model has been studied extensively, and it shares several phenotypes with human patients, including memory impairments indicative of hippocampal and cortical dysfunction (103, 104). Learning and memory deficits have been seen in trisomic mice using various paradigms requiring hippocampal function such as the Morris water maze and the radial arm maze (80, 103, 104, 140, 167, 170, 173, 176, 252, 292, 348). Accordingly, their hippocampi, but also their cortical pyramidal layer, displayed changes in circuitry development and structure, in synaptic plasticity, in growth, in response to signaling pathways, and in LTP (7, 28, 32, 62, 85, 151, 342). Reduced neuronal density and reduced excitatory synapses have been demonstrated in specific subregions of the hippocampus (151, 179, 206, 207). In addition, LTP is not induced in their dentate gyrus, and this defect has been linked to an enhancement in inhibitory synaptic transmission (32, 89, 103, 199, 342). Developmental and adult abnormalities in neurogenesis or in plasticity have been documented in DS mouse models (220, 224, 312). Indicative of neurodegenerative disorders, these models also show the degeneration of cholinergic neurons from the basal forebrain, similar to the defect found in Alzheimer and in trisomic patients. This alteration was described in 12-mo-old Ts65Dn mice (61, 64, 167, 315, 328), even though when the degeneration starts is controversial (141, 142, 169, 171).

View this table:

Mouse models of Down syndrome with their major phenotypes

In a way similar to human patients, Ts65Dn mice present motor dysfunction (173), although controversial data have been published using the rota-rod test (reviewed in Ref. 87). However, the cerebellum of Ts65Dn mice shows clear structural and functional alterations. The basic structure of the cerebellar cortex exhibits a bilaterally symmetric series of sagitally oriented bands attributable to a number of genes that are heterogeneously expressed within cerebellar granule and Purkinje cell populations (22). Similar to DS persons, Ts65Dn show reduced cerebellar size, and area measurements of histological sections of trisomic mice demonstrated reduced thickness of the internal granule and molecular layers. Moreover, the density of granule cells and Purkinje cells is significantly lower in Ts65Dn mice (22). Defects in the granule cell number is clearly dependent on their low response to the SHH mitotic signal (249, 304). But the impact of such defect on locomotor activity has so far been detected only in a few studies (175). Nevertheless, these cerebellar phenotypes provide quantitative end points for future studies of gene-dosage imbalance in mouse models with smaller, defined segmental trisomies (see Ref. 249 for review).

Several other phenotypes that originate during development are conserved in Ts65Dn mice, including short body length and defects in craniofacial structure, including smaller rostrocaudal length, brachycephaly, and shorter mandibles (161, 300, 301). In addition, a recent study reveals that Ts65Dn displayed heart abnormalities (251), causing the death of affected newborns at birth and the subsequent lower survival of trisomic animals at weaning (251, 306). Similar work carried out on the craniofacial defect and on the size and cell reductions in the cerebellum and hippocampus clearly demonstrates the developmental origins of the defects (161, 224, 271). For instance, alterations in cellular proliferation during postnatal development induces a 13% decrease in granule cell numbers in the dentate gyrus that persists until adulthood, during the period when the cholinergic neurons from the basal forebrain have not yet begun to degenerate (224). Such analyses highlight the value of mouse models for detailed characterization of the DS phenotype.

A second model arose in 1998 with the description of the Ts(1216)1Cje mouse (Ts1Cje; Fig. 5; Table 2), which carries a translocation of the Sod1-Zfp295 genetic interval to the telomeric end of the MMU12, but with disomic Sod1 (314). Comparative analysis of these two models allows studies of phenotype/genotype relationships in the mouse. A first consequence of this comparison was the reinforcement of the concept of a polygenic contribution to the DS phenotypes. For example, Ts1Cje mice are less severely impaired in learning and memory than are Ts65Dn (314). Four- to five-month-old Ts1Cje mice perform normally in the novel object recognition test (110), while 3- to 4-mo-old Ts65Dn mice show impaired performance (113) even though this defect is absent in older Ts65Dn mice (174) . These results suggest that at least two distinct loci, located upstream and downstream of Sod1, are interfering with object recognition memory in trisomic individuals. Similarly, craniofacial alterations and cerebellar phenotypes are attenuated in Ts1Cje compared with Ts65Dn (266), while heart defect was only found in Ts65Dn (251). In addition, the novel chromosome of the Ts1Cje is essentially transmitted in a Mendelian ratio, suggesting that at least one crucial locus for the heart phenotype is located in the Mrpl39-Sod1 genetic interval. Thus phenotypic analyses of these two models of trisomy begin to lead us toward phenotype/genotype correlation using the mouse as a model system with the aim of deciphering which genes are playing key roles in the phenotypes associated with DS.

Both models have been extensively studied to characterize the effects of dosage imbalance on gene expression (43, 56, 72, 185, 230, 281, 316). Not surprisingly, most of the genes with three copies were overexpressed in various organs studied at a ratio of ∼1.5 times disomic controls (9, 72, 185, 230, 316). This is similar to the human condition (119, 233, 234), even though in humans, dosage compensation exists with a tissue-dependent specificity (185, 230). This is referred to as the primary dosage effect of trisomic genes, while the secondary dosage effect refers to the consequence of the overexpression of trisomic genes on the rest of the genome (118, 249, 305). Further transcriptome analysis looked at the consequence of trisomic genes' overexpression at the level of the whole genome. In a detailed series of experiments, the transcriptome analysis of Ts1Cje mice at three postnatal stages (P0, P15, and P30) revealed that trisomic genes displayed elevated expression and that the transcription of a panel of disomic genes was altered. Furthermore, gene ontology analysis showed that the differentially expressed genes belong to the neural differentiation and migration class, including six transcription factor genes, of which two were from the Notch pathway (72, 281). In addition, the secondary effect could be perturbed by stochastic mechanisms regulating a substantial set of genes (316). Thus, even though studying the consequences of trisomy led to a clear conclusion about the expression of trisomic genes as a consequence of dosage imbalance (130, 351), its impact on the global transcriptome is still under debate. Most importantly, the origins of several phenotypes found in trisomic models result from alterations occurring at specific developmental time points.

Nevertheless, based on the variability in brain-regional expression of genes from various Ts65Dn individuals, Sultan et al. (351) proposed an interesting strategy to rank the trisomic genes as candidates for phenotypes. Obviously, measure of the impact of the trisomy in a transcriptome analysis is dependent on the purity of the tissue or cell type studied. A direct conclusion from these large studies is that a few pathways have been identified so far as altered in trisomic tissue, namely, NFAT-dependent transcription (14), the SHH signaling pathway (304), and the Notch signaling pathway (72, 281). As occurs in humans, further characterization and additional refinement will require more precise dissection of tissue or purification of cells to limit cross-contamination. Furthermore, the identification of crucial developmental time points will be needed to decipher the effects of genes contributing to the secondary dosage effects, whose altered expression affects their function in the corresponding tissue (249).

Considerable knowledge was acquired using these first trisomic mouse models, but we are still far away from a complete picture of the syndrome. Two main features of human patients, mental retardation and skull morphology, have been seen in trisomic mouse models. However, several other defects, including limb and gastrointestinal defects, hypothyroidism, increased incidence of acute megakaryocytic leukemia, and immune defects have yet to be explored. As a consequence, additional homologous regions that are not in the Mrpl39-Zfp295 genetic interval must play a role in the induction of the phenotypes associated with DS.

B. Genetic Tools for Creating Aneuploid Models

Chromosomal engineering may be defined as techniques for modifying the structure or the copy number of genomic regions or whole chromosomes. These are the methods of choice at the moment for generating new models of DS. Historically, generation of large chromosomal rearrangements was achieved by using x-irradiation or chemical mutagens known as clastogen or aneugen, which have the effect of inducing inversion, duplication, deletion, or chromosome loss or fusion. These classical approaches led to the generation and study of several valuable models of aneuploidy (73, 247, 367). Nevertheless, these random approaches still required a detailed characterization of the aneuploid condition, which can now be more efficiently specified using technologies for detecting copy number variation. The major breakthrough came in the mid 1990s with progress in manipulating large genomic regions or even whole chromosomes to generate new aneuploid models. The first new strategy took advantage of Cre/loxP-mediated chromosomal engineering and was described by the group of Bradley (for a recent review, see Ref. 36). Basically, the insertion of two loxP sites on either side of a region of interest (ROI) by gene targeting in embryonic stem (ES) cells followed by the action of CRE recombinase to select a designed chromosomal rearrangement. Thus, using two loxP sites in a cis-configuration and in the same relative orientation will generate either 1) a deletion if the CRE is active in the G1 phase or 2) a duplication if the CRE-mediated recombination event takes place in the G2 phase. Alternatively, a trans-configuration of the loxP site will undoubtedly lead to a balanced deletion and duplication of the ROI, maintaining a pseudo-disomic state of the cell. The in vitro strategy requires the restoration of a selectable marker, such as the Hprt minigene in deficient ES cells to select the new chromosomal configuration. To this end, 5′ and 3′ parts of a selectable minigene will be inserted, respectively, upstream or downstream of the loxP sites located on each border of the interval so that the minigene sequence is restored through the action of the CRE recombinase (288). With the use of this principle, a large set of deletions, duplications, and inversions have been engineered in the mouse genome (for a review, see Refs. 31, 33, 36, 215, 395). The strategy was used extensively to generate models of contiguous gene syndromes or aneuploidy, such as the Smith-Magenis (228, 378, 380, 396), Prader-Willy (363), and DiGeorge (217, 218, 318) syndromes. Following the same strategy, new models of DS have been obtained for the DS critical region between the Cbr1 and Orf9 genes (265) and for the entire region of MMU16 homologous to HSA21 (215). Furthermore, additional models corresponding to the deletion of the Cbr1-Orf9 interval or the Prmt2-Col6a1 region were derived and used to ascertain the role of genes and copy number variation on the physiology of the mouse (31, 161, 265, 267).

Complementary to the techniques described previously, X-ray Microcell-Mediated Chromosome Transfer (XMMCT) facilitates manipulating large chromosome fragments or whole chromosomes (for a recent review, see Ref. 364). This strategy originates in the 1970s to transfer into host cells single or small numbers of chromosomes by fusion with microcells (91, 243). Application of XMMCT to the creation of models of DS was achieved first in chimeric mice with transchromosomal ES cells containing different parts of human chromosome 21, ranging from ∼50 to ∼0.2 Mb (160, 189, 340). Those chimeras tend to display specific phenotypes, although with a high degree of interindividual variability (189, 340). The major breakthrough came from the group of Fisher, which succeeded in producing the first transchromosomic mouse line, called Tc1, that carries an almost complete extra HSA21 (263). Both techniques mentioned here completely changed our way of thinking about genetic dissection of DS phenotypes, allowing several groups to develop integrated genetic strategies to dissect the contribution of HSA21 or homologous regions to the DS phenotype.

C. Recent Outcomes From New Models for Down Syndrome

The first hypothesis tested in the recently created DS mouse models was that there exists a DSCR that contains all the major genes involved in generating DS phenotypes (79). Olson et al. (265) addressed the hypothesis using chromosomal engineering by generating new mice trisomic for the Cbr1-Orf9 genetic interval, named Ts1Rhr (Fig. 5; Table 2). The result of a craniofacial phenotypic analysis was quite surprising as these mice were not as affected as the Ts65Dn mice. Somehow, the DSCR was not sufficient to induce similar craniofacial defects in mice, even though distinct anomalies of rostrocaudal axis of the skull and of the mandible were noticed. Mice combining the deletion of the Cbr1-Orf9 region (Ms1Rhr) with the Ts65Dn trisomy confirmed the results obtained from the comparison between the Ts65Dn and Ts1Cje: while genes from the DSCR may contribute to the skull development, the genes involved in Ts65Dn craniofacial phenotypes must be located in the Sod1-Cbr1 genetic interval (265). Interestingly, the Tc1 mouse model does not display the same craniofacial defects as Ts65Dn mice. Thus the difference between the two models is dependent either on the mosaicism of the Tc1 model, which determines that every Tc1 mouse is unique, or on additional genes located outside of the Mrpl39-Zfp295 genetic interval that leads to the craniofacial phenotypes induced in the Ts65Dn mice (263). Similarly, impairment in learning and memory tasks involving the hippocampus found in Ts65Dn is not reproduced in the Ts1Rhr, while making the Ms1Rhr deletion in Ts65Dn mice restores normal performance in the Morris water maze (267). Thus, in this case too, the DSCR was not sufficient to induce the brain phenotype observed in Ts65Dn. Clearly, genes from the Sod1-Cbr1 genetic interval also contribute in learning and memory. On the contrary, the cerebellum and the cerebellar phenotypes observed in the three aneuploid mouse models lines are different both in the shape and the volume of the structure (7). These data support the hypothesis proposed by Olson et al. (265) suggesting that the effect of a single gene might not easily be seen but could contribute to an aneuploid phenotype in combination with other genes based on the specificity of their actions or through interaction with other genes.

A key step in responding to the challenge of making models of DS was achieved with the Tc1 mice (263), displaying a pleiotropic and variable “human”-like phenotype. A large number of phenotypes including behavior, locomotor activity, LTP, synaptic plasticity, cerebellar neuronal number, heart development, and mandible size were affected in these mice (127, 253, 263). Even though the transmission of the supernumerary chromosome is not completely achieved, leading to mosaic individuals, the Tc1 line showed a large set of deficits comparable to human patients. Other limitations of the model are the additional presence of small rearrangements and the random loss of HSA21. Nevertheless, Morice et al. (253) demonstrate that the Tc1 mice have a normal long-term reference memory and an impaired short-term working memory. Despite carrying the largest trisomy, Tc1 mice are not affected in the spatial Morris water maze as much as the Ts65Dn mice, whereas they are defective in short-term novel object recognition compared with the Ts65Dn mice. Moreover, Tc1 mice display motor coordination deficit (127). Interestingly, these detailed phenotypic analyses highlight more intricate phenotypes that validate the use of the Tc1 mouse as a model system for elucidating the complexity of DS phenotypes (291).

Further development of mouse models was recently achieved by the description of a new model encompassing the complete region of MMU16 homologous to HSA21 that extends from the mouse homolog of the Lipi gene to Zfp295 (Fig. 5, Table 2). The corresponding Ts1Yu mice were described by two major phenotypes (215). One affects the heart in 37% of the embryos, with defects related to human patients such as tetralogy of Fallot, atrial and ventricular septal defects, cleft mitral valves, severe coarctation of the aorta, and double outlet right ventricle. Interestingly, most of the cardiac phenotypes of the Ts1Yu were similar to those seen in the Ts65Dn, confirming that genes from the Mrpl39 to Zfp295 region are involved in heart abnormalities in patients. A second set of phenotypes affects the gastrointestinal tract with annular pancreas and malrotation of the intestine, which represents a category of rare phenotypes observed in trisomic human patients (256, 317, 327, 361). It is obvious that those new models represent a fantastic resource for dissecting the genetic contributions of HSA21 mouse homologs to conserved DS phenotypes.

D. Elucidating Therapeutic Pathways Using Mouse Models

The analysis of mouse models not only unravels the contribution of genetic region to the DS phenotypes, but it also highlights altered pathways that could be targeted for therapies. One of the first discoveries came from the study of the degeneration of the basal forebrain cholinergic neurons (BFCN) observed in 12-mo-old Ts65Dn mice (61, 142, 167). Mobley and co-workers (64) demonstrate that BFCN reduction is due to impaired retrograde transport of nerve growth factor (NGF) from the hippocampus to the basal forebrain starting at 6 mo of age. Impaired transport could be restored by direct NGF infusion of BFCNs (64). The loss of BFCN is common to DS, Alzheimer's disease (AD), and other neurodegenerative diseases. Similarities between both DS and AD were reinforced by the description of a duplication of App causing autosomal dominant early-onset form of the disease (308). Evidence of the role of App in BFCN degeneration were accumulated (172, 328), and a clear demonstration was obtained by combining the Ts65Dn trisomy and knock-out of the App gene (315). NGF retrograde transport was inhibited by the enlargement of early endosomes in Ts65Dn BFCNs induced by increased level of App, leading to BFCN apoptosis. Interestingly, additional pathways might also be considered as treatment with minocycline, a semisynthetic tetracycline with anti-inflammatory properties, that reduced the decrease of BFCN in the Ts65Dn mice, suggesting a more complex scheme for BFCN degeneration in Ts65Dn mice (169).

A second major step was achieved by the group of Craig Garner when they demonstrated the benefit of GABAA receptor antagonist on the cognitive performance of the Ts65Dn mice (114). Indeed, it was proposed that the learning and memory deficits and impaired LTP observed in Ts65Dn mice were due to an increase of GABAergic synaptic transmission (32, 199). Blocking the GABAergic inhibition by using GABAA receptor antagonists, such as picrotoxin or pentylenetetrazole (PTZ; Refs. 27, 114, 199), restores cognitive performance in several tests including Y maze, novel object recognition, and LTP. Nevertheless, such antagonists have severe adverse effects. They can provoke epilepsy at a higher dose. Thus caution should be taken especially in DS patients who are already suffering from epilepsy (88). Accordingly, a second study clearly showed that chronic treatment with PTZ induced adverse effect on locomotor function of Ts65Dn assessed in the rotarod test without modification of the sensory abilities, while learning and memory were improved in the Morris water maze (310).

New targets were identified via the modeling of DS such as Dyrk1A, encoding a serine-threonine kinase whose imbalance induces a variety of phenotypes (4, 8, 14, 47, 86, 219, 238, 240). Part of those changes could be reverted through RNA-mediated inhibition, showing that Dyrk1A-induced phenotypes are valuable targets for therapeutic approaches (196, 269, 326). To this end, molecules, such as Epigallocatechin-3-gallate, a major component of green tea (3), or harmine (21, 326), that could alter directly the Dyrk1A kinase activity, or others that would change the consequence of an hyperactive remodeling effect of Dyrk1A on chromatin (345) are current candidates.

Other promising approaches are emerging from the studies of mouse models. For example, the Kcnj6/Girk2, which encodes the G protein-coupled inward rectifying potassium channel subunit 2 (GIRK2), increased in gene-dosage may have a functional effect on the balance of excitatory and inhibitory neuronal transmission perturbed in DS models (32, 152, 343). Comparing consequences on learning and memory and LTP in mouse models also pinpoint candidate genes such as Gria1 that encodes the AMPA receptor subunit according to Schmitt et al. (322). Gria1 is involved in CA3-CA1 LTP and controls hippocampus-dependent spatial working memory but not the spatial reference memory (295, 321). Changes in the phosphorylation level of Gria1, observed in Ts65Dn hippocampal neurons (342), and reduced surface expression of Gria1, described in the postsynaptic membrane of the Tc1 hippocampus (253), suggest a role for the gene in the learning and memory and LTP defects found in those DS models. Undeniably, mouse model analysis will continue to play a key role in the identification of target genes for the design of new therapeutic strategies to reduce DS deficits (311).


Identifying the molecular genetic pathogenicity and etiology of DS mental retardation has proven arduous, despite some recent successes. DS, like other mental retardation disorders, is a complex polygenic disorder, with variable penetrance. The phenotypic heterogeneity among DS patients, and overlap with other mental retardation disorders, is not fully explored or integrated with the genetics work carried out so far. Moreover, gene-environment interactions and the effect of environmental factors (epigenetic modifications, effects of stress, infections, drugs, medications) on the expression of the phenotype have not been fully considered to date.

How may deregulation/dysfunction of a myriad of different molecules in DS affect cellular mechanisms in the brain? The affected mechanisms include those 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, and functional aspects of the adult brain, combining precise, localized analysis of the phenotype and knowledge of the genes showing dosage imbalance. However, there is a limited set of established cellular markers that are used in the current literature, and expression patterns of many genes remain uncharacterized. Since gene expression patterns are brain-regionally unique, distinguishing between localized changes in a phenotype and those affecting the entire brain is instrumental to understanding the particular developmental perturbations that result from aneuploidy.

On the other hand, the exciting new findings demonstrating considerable plasticity of the human genome may help to explain human traits with complex genetic patterns such as individual and pathological differences in cognitive profiles. CNVs may provide essential clues about complex disorders generated by gene-dosage alterations, but our current knowledge on CNVs is still very incomplete, in particular for CNVs of small size. A final understanding of the role of CNVs in clinical phenotypes may need to await the development of models that mimic the molecular consequences of gene-dosage alterations. Presumably, in DS, the situation is even more complex, as trisomic genes may also interact with genes located in regions showing copy number variation. Several insights into this field have been gained over the years through the study of aneuploidies at the chromosomal or gene level. Thus it is of major importance to develop a complete set of mouse models for DS. For the moment, trisomic models for the MMU17 and MMU10 regions homologous to the telomeric part of HSA21 are still missing. Having those models is essential to allow the generation of a model carrying the complete trisomy of region homologous to HSA21 and to assess more completely the genetic interactions among homologous regions. In addition, the availability in the future of a series of deletions covering almost the entire homologous region to HSA21 could be combined with the Tc1 mouse model to rescue the phenotypic defects and to analyze the genetic contribution (36, 364).

At the moment, it is still not clear whether phenotypes are induced by a series of interacting genes, each with a minimal contribution, or by major genes sensitive to dosage combining, with epistatic relation. New steps for DS research should thus include the following: 1) carrying out studies with groups of genes or genomic regions that may work together, in addition to studies with individual genes; 2) analysis of genetic association with discrete quantitative endophenotypes (154, 261) rather than broad mental retardation classifications (IQ); and 3) use of gene expression studies (in human postmortem brain or animal models), which are a direct reflection of gene-environment interactions, in a micro-scale genomic approach that analyzes genetic patterns at the cellular level.

Animal models will be useful tools in this endeavor because 1) gene expression studies in animal models can identify groups of genes that change together, and thus may work together, on a relatively homogeneous genetic background, with the signal not masked by the noise generated from the variable genetic background present in human studies; 2) specific phenotypes of the disorder can be deliberately mimicked in animal models with behavioral approaches; and 3) gene-environment interactions are minimized, well-defined, and well-controlled in animal model studies as opposed to human studies.

The elucidation of the aspects described above may contribute to developing new strategies to induce recovery of function in the DS central nervous system. Potential therapies could result in structural changes, such as axonal growth or synaptic reorganization in the brain or spinal cord, with expected functional recovery. This structural reorganization or repair will be more likely to lead to functional recovery if adequate training is provided simultaneously with new synapse formation.


X. Estivill and M. Dierssen are supported by SAF-2004-02808, SAF2007-60827, SAF2007-31093-E, Marató TV3, Fundación R. Areces, FIS (PI082038), and “Genoma España” (GEN2003-20651-C06-03). CIBER de Enfermedades Raras and CIBERESP are initiatives of the ISCIII. Y. Herault is funded by grants from the Centre National de la Recherche Scientifique, “Region Centre,” and “Conseil Général du Loiret.” M. Dierssen, Y. Herault, and X. Estivill are funded by the AnEUploidy project (LSHG-CT-2006-037627) supported by European Union Sixth Framework Program and the Fondation Jerome Lejeune.


We thank Monica Joana Do Santos and John Crabbe for their critical reading of the manuscript.

Address for reprint requests and other correspondence: M. Dierssen, Genes and Disease Program, Genomic Regulation Center-CRG, Pompeu Fabra University, Barcelona Biomedical Research Park, Dr Aiguader 88, PRBB building E, 08003 Barcelona, Catalonia, Spain (e-mail: mara.dierssen{at}