Dosage-sensitive genes.

The simplest model for gene action in DS is of a single dosage-sensitive gene that acts by itself to produce a phenotype, independent of effects by other genes or the environment to either buffer or exacerbate its dosage effect. In this sense, the gene is Mendelian in its function. A number of transgenic mice have been engineered to express elevated levels of Hsa21 genes or their mouse orthologs (see [13]). For the most part, these models have not been used to compare quantitatively the phenotypes in mice with segmental trisomy for the same (plus other) genes. A number of early transgenic studies used constitutive promoters to obtain high levels of expression without regard to normal spatial and temporal patterns for that gene. These types of studies may provide some insights into possible roles for a gene, but they are at best several steps removed from the conditions that produce specific phenotypes in DS.

The single dosage-sensitive gene model underlies hypotheses of “critical regions” on Hsa21, chromosome segments believed to include a dosage-sensitive gene or genes that are responsible for a given aspect of the DS phenotype. Shortly after the discovery of trisomy 21 as the cause of DS in 1959 [14], rare individuals with partial trisomy 21 were identified who had two complete copies of Hsa21 and a third copy of a subset of genes from this chromosome due to cytogenetic rearrangements [15,16]. Comparison of the triplicated regions from individuals who shared a given phenotype of DS could sometimes identify a common region of overlap believed to contain the “critical” gene(s) in a Down syndrome critical (or chromosomal) region (DSCR). The best-described DSCR extended about 5.4 Mb on Hsa21 [17,18]. This region was associated with several of the major DS phenotypes, including protruding tongue and flat facies (largely a function of hypoplastic mandible and craniofacial skeleton, respectively), short stature, mental retardation, joint hyperlaxity, muscle hypotonia, and a variety of dermatoglyphic abnormalities. The DSCR hypothesis predicted that a gene or genes in this region were sufficient to produce these DS features when present in three copies.

Several features attributed to this DSCR have direct parallels that can be measured precisely in Ts65Dn mice. With these phenotypic “readouts” for the predicted functions of a critical region gene or genes, Olson et al. [9] made a critical region model by re-engineering Mmu16 such that the region corresponding to this DSCR was duplicated or deleted. Mice carrying the duplication, which had segmental trisomy involving only the critical region genes, did not display the effects on stature nor the midface hypoplasia, small mandible, or dysmorphology of the skull predicted by the DSCR hypothesis. Thus, no gene(s) from this region was sufficient to produce these phenotypes. Next, mice deleted for the critical region segment were crossed to Ts65Dn mice (which display all of these DS characteristics), thus returning critical region gene dosage to normal in an animal that carried the majority of Ts65Dn segment genes in three copies. These mice had a somewhat attenuated presentation of phenotypes seen in Ts65Dn, indicating that while critical region genes made some contribution when present in three copies, they were largely not necessary for these effects. This result suggests that for those specific phenotypes, the DSCR hypothesis of single gene effects is not correct. Rather, multiple genes are required to produce these complex alterations to structures that are the products of intricate developmental processes.

Some aspects of DS may in fact be due primarily to the effects of a single dosage-sensitive gene on Hsa21. For example, elevated expression of endostatin, a protein that inhibits angiogenesis required for tumor growth, may explain at least part of the cancer resistance seen in DS [19]. However, it seems to us unlikely that many aspects of the DS phenotype that show highly variable presentation and derive from changes in structures that are the product of a long span of development are likely to reflect the effects of a single dosage-sensitive gene. Indeed, the classical understanding of Mendelian “single-gene” mutations as independently acting elements has been qualified with the greater appreciation for the roles of modifier genes on the phenotype.

Interacting genes of major effect.

A simple extension of the single dosage-sensitive gene model is to imagine additive effects of multiple dosage-sensitive genes interacting in a specific cell type during development. This could occur due to co-expression of two or more genes of major effect in the same cell at the same time or at different stages in the developmental history of that cell population. The effects of multiple dosage-sensitive genes might be amplified (or attenuated) when they occur within the same biochemical pathway. Possible trisomy 21 effects on a number of pathways have been posited [20], prioritizing them as targets for molecular analysis. However, the functions and interactions of most Hsa21 (and other) genes are not catalogued to this level. The combinatorial possibilities for testing groups of genes present an obvious challenge to direct interrogation by undirected screens. A further complication is that even in the mouse, few phenotypes are defined with sufficient precision to consistently detect small changes if one or two genes make an incremental contribution to the trisomic phenotype.

Ultimately, it may be less important to tease out “sub-phenotypic” consequences of individual genes than to identify the pathways and processes that are perturbed by trisomy. Correcting unbalanced pathways, regardless of the precise genetic cause, is a logical approach to attenuation of the phenotypic consequences [21].

Allelic variation on Hsa21.

Dosage sensitivity may be manifested in another fashion. Allelic variants of Hsa21 genes are present in different ratios in an individual with trisomy than in the diploid state. In the case where a mutant allele results in lower levels of gene product, this mutation will display recessive inheritance when the presence of one wild-type allele is sufficient to carry on normal function. A trisomic condition resulting in two copies of the loss-of-function mutation plus one wild-type copy would probably not alter the phenotypic outcome in this case. However, a recessively inherited phenotype can also occur when a mutant allele produces a gain or change of function, one copy of which does not produce a detrimental effect in the presence of a single wild-type allele, but two copies of which may be sufficient to “overcome” the buffering of a normal allele in a trisomic individual.

Another possible manifestation of trisomy at the molecular level is heterotrisomy, in which alleles from three grandparents are present in every cell [22]. This will occur when trisomy results from an error in meiosis I, the most frequent origin of the extra chromosome in DS [3]. For multimeric proteins assembled from multiple peptides, such as the collagens, the combinatorial possibilities become large. (COL6A1, COL6A2, and COL18A1 are all encoded on distal Hsa21.) Individuals with trisomy will produce combinations of multimers that cannot occur in euploid individuals. Baptista et al. described a region of Hsa21 between D21S167 and HMG14 that was frequently heterotrisomic in individuals with DS and CHD [22].

Both “recessive dosage” and heterotrisomy should be amenable to genetic analysis. However, standard statistical methods do not account for the possibility of three alleles in one individual. Sherman, Feingold, and colleagues have established statistical methodologies for genetic association studies to identify genes that affect the DS phenotype when triplicated [23].

Coincident small effects.

The preceding examples describe situations in which phenotype is altered due to increased expression of one or a few trisomic genes of major impact. However, small coincident or additive effects of the many genes over-expressed in every trisomic cell may also contribute to trisomic phenotypes. Recent observations confirm that transcript levels are elevated about 1.5-fold for the majority of trisomic genes in a few tissues from humans with trisomy 21 [24,25] and across a broad range of tissues that can be measured in trisomic mouse models [11,12].

In this model, an individual triplicated gene might have no demonstrable impact on phenotype by itself, whereas the collective effect of dozens of genes affecting multiple cellular processes is sufficient to result in a significant impact on phenotype (see [26,27]). Proving this model presents significant experimental challenges, but might be approached by considering quantifiable phenotypes in animal models. Several phenotypes have been measured precisely in trisomic mouse models, allowing comparison between the Ts65Dn mouse, with about 130 Hsa21 orthologs in three copies, and Ts1Cje mice (91 triplicated genes). Behavioral, structural, and functional brain phenotypes, dysmorphology of the skull, and gene expression in the cerebellum all show patterns in Ts65Dn that are similar but attenuated when fewer genes are trisomic in Ts1Cje [6,28–31]. Attenuation of the phenotype when fewer genes are triplicated is consistent with (but not proof of) additive small effects by neighboring trisomic genes.

Note that not only trisomic genes show altered expression in tissues from individuals with trisomy. In some but not all studies, the perturbation of gene expression levels has been demonstrated to extend beyond trisomic genes to those that are disomic, affecting expression levels of a substantial proportion of transcripts in trisomic tissues in mice [32,33]. In one study of trisomic mouse cerebellum, up to one-third of disomic gene transcript levels were subtly altered [32]. Very few of these genes showed a statistically significant difference with euploid when considered individually, but collectively, disomic gene expression robustly distinguished the trisomic and euploid cerebellar transcriptomes. Conflicting results regarding the question of perturbation of disomic gene expression have been reported in human studies [24,25,34]. The controversy is perpetuated by the use of different analytical approaches for array analysis in different studies.

Modifier genes.

Most of the features that occur frequently in DS are variable in severity (expressivity) and, except for a few characteristic phenotypes, in occurrence (penetrance). None of the commonly described DS phenotypes are unique to DS or other chromosomal abnormalities but also may occur in euploid individuals [35]. This wide degree of variation suggests that a particular phenotype in a given individual is affected by genetic and environmental variation, and it is reasonable to assume that genetic background (the specific allele set inherited by an individual) affects the severity of outcome.

Preliminary data support the supposition that genetic modifiers contribute to CHD, for which trisomy 21 is the largest risk factor. About half of all individuals with DS have some form of CHD, and most of these involve septal defects. Complete atrioventricular canal occurs in one of five individuals with trisomy 21, compared to 1/10,000 in the euploid population [4]. However, since 80% of those with DS do not have complete atrioventricular canal and 50% have no clinical presentation of heart defects, trisomy 21 is not sufficient to cause CHD by itself. Evidence from patient studies now links the occurrence of CHD in individuals with DS to mutations in disomic genes known to affect septation in mouse models and in non-syndromic atrioventricular canal (C. Maslen, S. Sherman, G. Capone and R. Reeves, unpublished data).

The increased frequency of several important medical conditions in DS, including CHD, childhood onset leukemia, and Hirschsprung disease, suggests that additional genetic factors may be involved. That is, the occurrence of these diseases is greatly increased—but not caused—by trisomy 21. Predisposing modifier genes may combine with the effects of trisomy 21 to reach a threshold effect, resulting in an observable phenotype (see Box 1). Genetic studies in “sensitized” DS populations can be especially effective for identifying genetic variation that contributes to these conditions, regardless of ploidy.

Not all predisposing conditions caused by trisomy have a negative impact. Individuals with DS have reduced frequencies of solid tumors [36,37] and may have a lower incidence of atherosclerosis as well [38,39]. Characterization of these effects could indicate approaches to reducing cancer incidence or cardiovascular disease in all people.