The undifferentiated ES cell inoculates were allowed to develop for 30 days, equivalent to the time span of entire intrauterine, and 10 postnatal days of mouse development. This extended period encompasses the critical interval (P0–P6), in which the biggest difference in neuronal precursor cell proliferation was observed between mouse models of DS, and their euploid littermate controls [11, 15, 16]. The use of syngeneic inbred mouse strains also minimizes the individual variation, which is further reduced by the paired analysis of tumours which grow in the same recipient animal. Using this system, we have observed a powerful inhibition of neuroectodermal fate in transchromosomic +HSA21 containing pluripotent mouse ES cells, compared to their parental control (Fig 1A,B.). This conclusion is not without caveats: the inhibition of neural fate could be specific to the cell line used, or it could be related to the sheer presence of a supernumerary chromosome, not specific to HSA21. Though it is impossible to rule out these caveats at this stage, we believe on balance of probabilities, that this system shows a measurable phenotype with potential to map it to a segment of HSA21. The reduced Tubb3 levels (Fig 1C) clearly indicate a reduction in mature neuron numbers in trisomic teratomas, confirmed by neuron specific staining (Fig 2). The reduction in GFAP levels could be partly the result of reduced numbers of mature astrocytes, as well as reduced numbers of neuronal precursors, corresponding to GFAP+ cells found in the subgranular zone and hilus . These GFAP+ precursors divide and give rise to immature neurons (DCX+PSA-NCAM+) . Radial glial cells, which are the main source of precursors for neurogenesis in the dentate gyrus, also express GFAP . These very cellular layers are the site of the biggest deficit in mitotic activity at postnatal day 6 in mouse models of DS .
Spontaneous deletions of the supernumerary HSA21 were observed in two thirds of transchromosomic tumours (Fig 3). As there were sufficient numbers of cells of many different tissue types in all tumours, and multiple HSA21 PCR reactions (including RT-PCR of ubiquitously expressed HSA21 genes) were negative in the deleted regions, deletions must have occurred before any differentiation of ES cells in the tumours took place, probably during the early cell divisions at the very start of tumour growth. Alternatively, a subset of cells with a major deletion might have been present within the inoculum of 47-1 cells; a less likely explanation, as 3 independent cultures gave rise to a similar proportion of tumours with, and without deletions. We cannot exclude the possibility that the inocula in all 3 experiments contained a mixture of two clones, the 47-1, and a segmental trisomy clone. However, the deleted segments also show subtle variations in the pattern of markers present, making this possibility less likely. Whether the deletions are caused by an instable DNA sequence element, and whether or not they provide a selective advantage to the proliferation of cells, remains to be investigated. Regardless of the cause of deletions within the tumours, they had no effect on the statistical significance of any differences shown in Fig. 1. Though the data are not conclusive on this point, it can be hypothesized that the segmental trisomy of two regions common to all tumours (Fig. 4), comprising less than a third of the HSA21 gene complement, could be sufficient to cause a powerful Inhibition of Neuroectodermal DIfferentiation (INDI) in pluripotent ES cells in vivo.
The telomeric INDI region partially overlaps with trisomic regions in mouse segmental trisomy models for DS, such as Ts65Dn and Ts1Cje (Fig. 4), which both have a reduced number of granular layer neurons in the cerebellum and hippocampus, compared to euploid littermates [11, 16, 27]. The overlap between the telomeric INDI segment and the 3 mouse models shown in Fig. 4. is restricted to 23 genes (based on comparisons, see refs [1, 28]. The centromeric INDI region is less likely to play a causative role in INDI, as the degree of cerebellar neuron number reduction is similar between the Tc1 mouse model  which has this region in trisomy, and the Ts65Dn and Ts1Cje models that do not . The exception are only 3 genes from the centromeric INDI region (BTG3, YG81 and PRSS7) which are deleted in the third chromosome of the Tc1 model , and therefore remain unchecked. The telomeric INDI segment of 23 genes overlapping with trisomic mouse models is fully contained within the Ts1Rhr, the 33 gene-trisomy mouse model, which does not show the reduction in cerebellar volume, or neuron density . This mouse model also shows no hippocampal volume change, and no electrophysiological and behavioural defects associated with hippocampal functions, therefore suggesting that trisomy of this segment is not sufficient to cause the brain pathology in the mouse . However, this 33 gene segment was found necessary for most of the cerebellar and hippocampal pathology of Ts65Dn model, as when its trisomy is reversed to disomy, the pathologies disappeared . This segment also probably contains highly and bi-directionally dose sensitive genes regulating brain development, as a monosomy of this segment (mouse model Ms1RhR) produces striking changes in cerebellar and hippocampal volume and neuron density . The data in our system could differ from those in the Ts1RhR model due to the human, rather than mouse, origin of the third chromosome. Interestingly, very recent data in human DS show that duplication of a 4.3 Mb segment, completely contained within the 33 gene equivalent segment in the mouse, is sufficient to cause (in three members of the same family) a range of DS phenotypes, including brachycephaly, intellectual disability, mental retardation, speech learning impairment, and a typical facial gestalt of DS . Our data suggest that further studies of individual gene dosage effects within the INDI region could reveal major candidate contributors to DS-related hypo-cellularity of the CNS, and suppression of neuroblastomas.