Down syndrome (DS) is the most common genetic reason for learning disability and congenital malformations in the human population, occurring with an incidence of around 1/600 newborns. It is now almost half a century since the genetic background was identified, i.e. that most people with DS have an extra small chromosome, some of which are mosaics with only a proportion of trisomy 21 (T21) cells, while a minority have the relevant part of chromosome 21 translocated to another chromosome, leading to this type running in families with a substantially increased risk for parental carriers [1].

During the interim five decades much research has been devoted to trace the origin of the common type of DS characterised by an extra free chromosome 21 [2–4]. There is unequivocal evidence from large-scale family linkage studies comparing DNA markers between parents and affected children that the extra chromosome 21 most often originates from the mother, and there may be aberrant maternal recombination along the length of the long arm of chromosome 21. In addition, it is well known that the recurrence is increased for younger mothers and the incidence increases dramatically with maternal age (Figure 1). The underlying biological reasons for this spectrum of clinical and experimental observation are still unknown, constituting outstanding biological enigmas. Most recently it has been suggested that the situation is highly complex, dependent on multifactorial traits [2–4].

Nevertheless it has become generally accepted that the main problem concerns abnormal chromosome behavior of the two chromosomes 21 during the first meiotic oocyte division, so-called primary non-disjunction (Figure 2). We here present data to indicate that the common type of DS in children may in fact result from their mothers themselves being mosaics, carrying a proportion of T21 oocytes in their ovaries (Figure 3). The implication is that the decisive problem is instead the segregation of the three pre-existing chromosomes 21, i.e. obligatory secondary non-disjunction during the first meiotic oocyte division (Figure 4), rather than primary non-disjunction of two chromosomes 21 that has become the generally accepted dogma.

All procedures were performed with informed consent and ethical approval from the local ethical committees. Foetal ovarian cells were obtained from eight foetuses at gestational age 14–22 weeks, following termination of pregnancy for social reasons with all the foetuses having a normal phenotypic appearance. Ovaries were removed within a few hours post-mortem and placed in L-15 (Leibovitz) medium (Life Technologies) containing 0.3% bovine serum albumin (Sigma). Pieces of ovaries were frozen at -80°C; thawed cells were prepared for immunoflourescence and fluorescence in situ hybridisation (FISH).

Only a small proportion of each ovary from the initial collection of foetuses [1–4] was used for this study, with the majority divided amongst other experiments, the results of which are described previously [5–7]. In this project we used an antibody against the meiosis-specific protein STAG3 for an initial round of immuno-fluorescence to discriminate between germ cells entering first meiosis prophase in relation to pre-meiotic germ cells and ovarian stroma cells [8]. Parts of the tissue samples from the later collection of foetuses [5–8] were used to prepare direct imprints from the cut surface of the foetal ovary and the remaining material processed by microspreading [9, 10].

Microscopy slides for FISH analysis were fixed in methanol: acetic acid then washed in 2X standard saline citrate (SSC) and treated with pepsin (0.1 mg/ml) in 0.01 M HCl for 8 min at 37°C. After additional washing in phosphate-buffered saline (PBS), paraformaldehyde (1%) fixation and dehydration through series of alcohol the slides were left to air-dry at room temperature. Hybridization was performed according to the manufacturers' instructions with two DNA probes positioned near the end of the long arm of chromosome 21 (Cat No: 32-190002, Abbot Molecular Inc, USA and Cytocell, Cat No. LPT21QG/R, Cytocell Technologies Ltd. UK). The DNA probes were mixed and added to the slides followed by denaturation, hybridization and post-hybridization washing. After dehydration slides were mounted in glycerol containing 2.3% DABCO (1, 4-diazabicyclo-(2, 2, 2) octane) as antifade and DAPI (4, 6,-diamino-2-phenyl-indole) 0.5 mg/ml for nuclear counterstaining.

Fluorescent signals were analyzed using a Zeiss Axioskop 2 microscope equipped with a cooled CCD camera (CoolSnap; Photometrics Ltd, USA) controlled by a Power Macintosh computer. Grey scale images were captured, pseudocolored and merged using the SmartCapture 2 software (Digital Scientific Ltd, UK).

The same FISH procedure was applied to preparations of blood lymphocytes obtained from subjects, serving as a control population with respect to the efficacy of the dual FISH probe analysis for appropriate chromosome counts [10]. This included three children with T21 and three children ascertained because of learning disability without any signs of DS, shown to have normal karyotypes by metaphase analysis.

In these T21 figures we have only included those showing three clear double signals in comparison to the normal two. This is illustrated in Figure 3. The left image shows the position of the two fluorescence probes near the end of the long arm of chromosome 21 (Figure 3a). An example of a normal cell nucleus containing two double chromosome 21-specific fluorescence signals (one red and the other yellow-greenish) is shown in Figure 3b. Cell nuclei, each containing three double signals, thus interpreted to have three copies of chromosome 21 and labelled as representing T21 cell nuclei, are illustrated in Figure 3c, d.

Pre-staining with an antibody against the meiosis-specific protein STAG3 performed in cases 1–4 allowed differentiation between germ cells entering meiosis in relation to pre-meiotic germ cells and ovarian stroma cells. The proportion of T21 cells identified by the STAG3 antibody/nuclear size as meiotic germ cells was on average 0.41% with a range of 0.31–0.49%. There was no statistically significant difference between samples in this respect (SD 0.17; P > 0.05).

Using the same criteria as for FISH analysis of foetal ovaries we did not find any indication of a similar type of mosaicism as regards in vitro cultured blood lymphocytes from three children with T21 and three children with normal karyotypes by traditional metaphase analysis. Thus, in a total of 12,320 cells analyzed we did not detect any cell nucleus showing extra dual chromosome 21 signals.

In this study we have, for the first time, documented that T21 ovarian mosaicism is common in normal human female foetuses. T21 mosaicism has also previously been documented in ovaries from adult human females, who have had one or more children with T21 DS (Table 1). Thus, in a total population of 15 human females, where ovarian cells have been investigated in this respect, T21 mosaicism of varying degrees has been found in all. On the basis of these observations, together with the expected implications as regards the maternal age effect, the higher recurrence of DS in younger women, the aberrant maternal meiotic recombination patterns and the low incidence of DS of paternal origin, we suggest that gonadal mosaicism is a prevalent unifying reason for T21 conceptions in the human population.

In this paper we challenge the current dogma that disomic maternal and paternal gametes are most often caused by failure of the two homologs 21 to separate at the first meiotic division, so called primary non-disjunction. Instead we propose that obligatory secondary non-disjunction of three homologs 21, which is the expected outcome of gonadal T21 mosaicism, may constitute a common reason. One relatively straight forward way to obtain further information on this situation would be to count copy number of chromosome 21 at the meiotic metaphase I stage to find out what proportion contain two respectively three copies. To our knowledge no T21 spermatocytes have so far been identified at the metaphase 1 stage in testicular biopsy samples from any men other than those diagnosed as having T21 DS, but this nevertheless requires further study. In particular, there are as far as we are aware no such data available on human oocytes, where collection of a large enough sample of ovulating oocytes presents one of the outstanding hurdles. Further studies in this regard are underway. Obviously more work is also required to find out to what extent gonadal mosaicism does exist for other chromosomes than 21, including those that are of special relevance for common genetic disease in the human population. On another note, we are also wondering if the intriguing species difference as regards constitutional aneuploidy, where it would appear that our own species is much more affected than other mammals, could be caused by more stringent control of embryonic cells divisions and therefore a lower incidence of gonadal aneuploidy mosaicism. Similar studies as we have here reported on chromosome copy number in human foetal ovaries should be straight forward and allow this outstanding question to be readily answered.