The human blastocyst: cell number, death and allocation during late preimplantation development in vitro

Despite the diversity of eutherian mammals, their early development is very similar (Wimsatt, 1975). Following fertilization, the zygote, enclosed in an acellular envelope, the zona pellucida, undergoes a number of cleavage divisions, compacts to form a morula and, at the blastocyst stage, accumulates fluid in a central blasto-coel cavity. The blastocyst expands, hatches from the zona and implants. However, there are considerable differences between species in the timing of these events and their relationship to the timing and extent of implantation. For example, the elephant shrew forms a blastocyst at the 4-cell stage, approximately 2 days after fertilization, whereas in the mouse, blastulation coincides with the completion of the fifth cleavage division at the 32-cell stage, 4 days after fertilization (Smith and McLaren, 1977; Handyside and Hunter, 1986). Also, implantation occurs within 36–48 h of blastulation in the mouse whereas in many domestic species (e.g. horse, pig, sheep and cow) implantation is delayed until a much later stage at which primitive streak formation has been initiated.

In the mouse blastocyst, there are two distinct cell types: an outer epithelial layer of trophectoderm (TE) responsible for blastocoel fluid accumulation and specialized for implantation, and an inner cell mass (ICM). Following implantation, the TE only gives rise to components of the placenta and extraembryonic membranes whereas the ICM forms all three germ layers of the fetus as well as complementary contributions to the extraembryonic membranes (Gardner and Papaioannou, 1975). Allocation of cells to these two primary lineages occurs at the preceding morula stage mainly during the fourth cleavage division (Handyside, 1981); although an additional contribution from the outer cells to the ICM has been demonstrated during the fifth cleavage division (Pedersen, 1986; Fleming, 1987).

The numbers of TE and ICM cells in mouse blastocysts developing in vivo have been studied by serial sectioning (Copp, 1978) and more recently by differential labelling of the nuclei with polynucleotide-specific fluorochromes (Handyside and Hunter, 1984, 1986). Over a period of 36 h following blastocyst formation, the number of TE cells increases exponentially before reaching a plateau. This coincides with the initiation of endoreduplication and giant cell formation in the mural TE surrounding the blastocoel cavity during implantation (Barlow et al. 1972). At the same time, cell number increase in the ICM slows down and there is a wave of embryonic cell death, or apoptosis (El-Shershaby and Hinchliffe, 1974; Copp, 1978; Handyside and Hunter, 1986). Hence, the proportion of ICM cells steadily declines as the blastocyst expands, hatches and begins to implant.

With the human embryo developing in vitro, compaction occurs late on day 3 following in vitro fertilization (IVF), between the 8- and 16-cell stages, blastulation on day 5 or 6 (Hardy et al. 1989), and implantation in vivo between days 7 and 10. The time taken to complete the first three cleavage cycles is similar to the mouse, on average about 60 h (Edwards et al. 1981; Sundstrom et al. 1981; Trounson et al. 1982). At later stages, there is very little information about cell numbers or the allocation of cells to the TE and ICM at the blastocyst stage. By differential labelling of the nuclei of the TE and the ICMs of human blastocysts developing in vitro, we report here cell numbers in newly expanded blastocysts on day 5, the increase in cell numbers on days 6 and 7, and the allocation of cells to the two lineages. In addition, we have examined polyspermic and partheno-genetic blastocysts and the nuclei of a series of embryos arrested at cleavage and morula stages. The results suggest that the numbers of cells allocated to the TE and ICM of the human blastocyst are similar to the mouse. However, despite evidence of increasing cell death between days 5 and 7 in both the TE and ICM, the proportion of ICM cells remains higher.

The work has been approved by both the local ethical committee of the Royal Postgraduate Medical School, Hammersmith Hospital and the Voluntary Licensing Authority for Human in Vitro Fertilization and Embryology of the Medical Research Council and the Royal College of Obstetricians and Gynaecologists.

Surplus human embryos were obtained with permission from patients undergoing in vitro fertilization (IVF), using a method of superovulation described by Rutherford et al. (1988). After pituitary-gonadal suppression with a luteinising hormone releasing hormone (LHRH) agonist (Buserelin, Hoechst), patients were superovulated with human menopausal gonadotrophin (hMG; Perganol, Serono). 10OOOi.u. human chorionic gonadotrophin (hCG; Pregnyl, Organon) were given 34 h before egg collection.

Oocytes were collected, preincubated, inseminated (day 0) and checked for pronuclei the following day, as described previously (Hillier et al. 1984). Embryos were then cultured in 1ml of medium containing 10% heat-inactivated maternal serum under a gas phase of 5 % CO₂, 5 % O₂ and 90 % N₂. The medium was either T6 (Quinn et al. 1982) or Earle’s balanced salt solution (Gibco) supplemented with 25HIM-sodium bicarbonate (BDH, Analar) and 0.47 mM-pyruvic acid (Sigma). On day 2, each fertilized embryo was examined and up to 3 were selected on the basis of their morphology for embryo transfer, on days 2 or 3. After confirming the patients’ consent, the surplus embryos were allowed to develop in vitro either in the original 1 ml of medium or in 20 /il drops of the same medium under paraffin oil (BDH) to facilitate scoring.

Oocytes and embryos were initially classified according to the number of pronuclei visible on day 1 (16 h after insemination) as follows: (i) fertilized embryos (with two pronuclei); (ii) polyspermic embryos (with three or more pronuclei); and (iii) unfertilized oocytes (with one or no pronucleus). Later on day 2, embryos were graded according to eveness of blastomeres, fragmentation and presence of cellular debris from perfectly symmetrical embryos with no fragmentation (grade 1) to embryos having one intact blastomere with gross fragmentation (grade IV) or being totally degenerate (grade V) (Dawson et al. 1987). In addition, unfertilized oocytes that cleaved were reclassified as parthenogenetic embryos. Blastocysts were selected for labelling on days 5 to 7.

The TE and ICM nuclei of blastocysts were differentially labelled with polynucleotide-specific fluorochromes using a modification of the method described for mouse embryos (Handyside and Hunter, 1984). Briefly, the nuclei of the outer TE cells are first specifically labelled with the fluorochrome propidium iodide. This fluorochrome is excluded from viable ICM cells but labels TE cells undergoing antibody-mediated complement lysis during immunosurgery. The whole embryo is then rapidly fixed and both the TE and ICM nuclei labelled with a second fluorochrome, bisbenzimide. Since the emission spectra of the two fluorochromes are different, the labelled nuclei can be distinguished by the colour of the fluorescence using appropriate filter combinations. The modification involved treating human blastocysts with trinitrobenzenesul-phonic acid (TNBS) to label cell surface proteins with covalently bound trinitophenol (TNP) groups. This allowed substitution of a rabbit antiserum against dinitrophenol (DNP) covalently linked to bovine serum albumin (BSA) (which cross-reacts with TNP-labelled proteins), for the antispecies antiserum used with mouse blastocysts for immunosurgery (Surani and Handyside, 1983). Preliminary experiments with fluorescein-conjugated goat anti-rabbit antiserum demonstrated uniform labelling of the outer cells of human embryos at cleavage and blastocyst stages. The protocol was as follows: the zonae were removed in acid Tyrodes solution (Nicolson et al. 1975) and zona-free blastocysts washed in Medium M2 containing 4 mg ml⁻¹ BSA (Quinn et al. 1982). They were then incubated in 10mM-TNBS acid (Sigma) in ungassed Medium 16 (Whittingham, 1971) containing 4mgml⁻¹ polyvinylpyrrolidone on ice for 10min, washed three times in M2, and incubated in O.lmgml⁻¹ anti-DNP-BSA (ICN ImmunoBiologicals) in M2 at 37 °C for 10 min. After further washing in M2, blastocysts were finally incubated in a 1:10 dilution of guinea pig complement serum (Sigma) in M2 containing 0.01 mg ml⁻¹ propidium iodide (Sigma) at 37°C for 10–15 min. The degree and eveness of lysis of the outer TE cells were assessed and the blastocysts were briefly washed in phsophate-buffered saline without Ca²⁺ and Mg²⁺ before fixing and labelling in 0.05mm-bisbenzimide (Hoechst 33258) in absolute alcohol overnight at 4 °C. Labelled blastocysts were washed for between 24 and 48 h in absolute alcohol before examination. Some zona-free blastocysts and arrested embryos were not differentially labelled but simply fixed and labelled in bisbenzimide in absolute alcohol.

Embryos with labelled nuclei were individually mounted on microscope slides in glycerol underneath a coverslip and initial examination was carried out in whole mount. Using a combination of two filter sets, i.e. UV excitation and band pass filters (A2; Leitz) with the emission filter from the FITC filter set (12; Leitz), ICM nuclei labelled with bisbenzimide appear green and TE nuclei labelled with a combination of bisbenzimide and propidium iodide appear orange. After photography and description of the staining pattern, the embryo was carefully squashed and disaggregated using a pencil eraser and examined again for counting.

Estimates of cell numbers in blastocysts are based on the assumption that they are equivalent to the number of nuclei counted. Multinucleate cells are commonly observed in normally fertilized early cleavage stage embryos. However, they are considered to result from defective migration of chromosomes at mitotic anaphase (Tesarik et al. 1987) and are therefore unlikely to make a major contribution to more advanced embryos. Normal nuclei had a distinct nuclear outline, brightly staining nucleoli and even shape. Cells at the metaphase stage of mitosis were counted as single cells. Estimates of the number of dead cells were based on the presence of degenerating nuclei characterized by discrete clusters of labelled nuclear fragments. These dead cells were not included in the overall total of cell numbers. Mitotic and dead cell indices were calculated as follows:

Examination of the variation in blastocyst cell number in blastocysts from the same patient and between different patients was carried out by Analysis of Variance. Differences in total cell number and the numbers of TE and ICM cells in blastocysts between days 5 and 7 were analyzed for significance using the Student’s t test.

The development of a total of 181 surplus embryos of various classifications from 42 patients (1–17 surplus embryos per patient), was observed from day 2 to day 5, 6 or 7 in vitro (Table 1). 63 out of 149 (42%) normally fertilized embryos developed to the blastocyst stage by day 5 or 6, and 6 of these hatched (Fig. 1A). Some of these blastocysts (15/63), however, were morphologically abnormal and were characterized by one or more of the following features: unincorporated blastomeres, multiple cavities (possibly intracellular vacuoles), no discrete ICM visible or low numbers of mural TE cells. The remainder of the normally fertilized embryos arrested at cleavage and morula stages. Of 21 polyspermic embryos, 6 reached the blastocyst stage by day 6. Three of these were morphologically abnormal. Two out of a total of 11 parthenogenetic embryos developed to the blastocyst stage. Both were morphologically abnormal.

Differential labelling of 50 out of 71 blastocysts was attempted (Fig. IB). 4 were lost during processing. In 17 cases, careful examination in whole mount suggested that some of the TE nuclei had not been differentially labelled. The majority of these (10 out of 17) were either morphologically abnormal blastocysts, polyspermic or parthenogenetic blastocysts. Among the remaining 7 normally fertilized blastocysts, remnants of the zona pellucida after acid Tyrodes treatment appeared to have prevented the immunosurgical lysis of part of the TE. This phenomenon has been observed in the pig (Papaioannou and Ebert, 1988). These blastocysts were, however, adequately labelled after fixation for determination of total cell numbers (see below). The numbers of TE and ICM cells in the remaining 30 normally fertilized blastocysts on days 5–7 are presented in Table 2. Newly expanded blastocysts on day 5, with a total of 58.3±8.1 cells, had 37.9±6.0 and 20.4±4.0 TE and ICM cells, respectively. Amongst expanded blastocysts on day 6 (some of which had reached this stage on day 5), the number of TE cells was virtually unchanged, 40.3±5.0; however, the number of ICM cells, 41.9±5.0, was double. By day 7, TE cell number was double (80.6±15.2), but the number of ICM cells was almost the same (45.6±10.2) as on day 6. All of these normally fertilized blastocysts had ICMs although the range, from 5 to 113 cells, was considerable. The proportion of ICM cells varied from about a third on day 5 and 7 to a half on day 6. Cell death, characterised by fragmenting nuclei (Fig. 1C), was widespread and increased between days 5 and 7; 28 out of the 30 blastocysts examined between days 5 and 7 had evidence of cell death in both the ICM and the TE. The mitotic index of both the TE and the ICM was relatively high on day 5 but was reduced significantly by day 7. Differential labelling of two morphologically abnormal blastocysts on day 6 revealed 15 and 9, and 51 and 35 cells in the TE and ICM, respectively.

The total cell numbers, mitotic and dead cell indices for 45 normally fertilized blastocysts, 16 with abnormal morphology and 6 abnormally fertilized blastocysts on day 5, 6 and 7 are summarized in Table 3. This series of embryos included all 50 blastocysts which were differentially labelled (see above) together with 17 others that were simply fixed and labelled with a single fluorochrome. Total cell numbers in this larger series of normally fertilized blastocysts on days 5 to 7 were almost identical to those ascertained after differential labelling (Table 2). The range of total cell numbers in normally fertilized blastocysts increased progressively from day 5 to day 7 (Fig. 2A-C). Newly expanded blastocysts on day 5 ranged from 24 to 90 cells, on day 6 from 27 to 136 cells and on day 7 from 60 to one hatched blastocyst with 283 cells.

Normally fertilized blastocysts on day 6, which were morphologically abnormal, had only half the cells of their morphologically normal counterparts (Fig. 2D, Table 3). One morphologically abnormal blastocyst on day 5 had only 18 cells. Polyspermic blastocysts on day 6 had a similar number of cells to morphologically abnormal blastocysts and large numbers of dead cells. Two parthenogenetic blastocysts had a total of only 6 and 13 cells.

46 arrested embryos were labelled to examine nuclear morphology on days 6 and 7. These included eleven 2- to 7-cell-stage embryos, ten 8-cells, four uncompact morulae with greater than 8 cells, seven morulae, eight cavitating or vacuolated late morulae and six totally fragmented embryos. Ten of the arrested embryos had more nuclei than would be expected on the basis of their morphology. Four arrested 4-cell embryos had between 7 and 35 nuclei including one with 23 normal nuclei and 12 dying nuclei in one cell and none in the other three cells. One 8-cell embryo had 16 normal nuclei with a few additional fragments of nuclear material. Two morulae contained large numbers of nuclei; one with 90 including 35 degenerate nuclei, the other 30 including 7 degenerate nuclei. Interestingly, three embryos that were scored as being totally fragmented had 20, 24 and 45 normal nuclei. Also, seven arrested embryos had no nuclei apart from a few bright specks of fluorochrome-labelled material. One arrested embryo had a highly polyploid metaphase (Fig. ID). Finally, six other arrested embryos had ‘flocculent’ nuclei, with labelled chromatin clumped on the nuclear membrane (Fig. IE).

The use of in vitro fertilization (IVF) for the treatment of infertility presents a unique opportunity to study the early development of the human embryo in culture. The early cleavage divisions, before transfer on day 2, have been extensively described (Sundstrom et al. 1981; Edwards et al. 1981; Trounson et al. 1982; Mohr and Trounson, 1984). However, apart from a series of ultrastructural studies of human blastocysts (Lopata et al. 1982; Mohr and Trounson, 1982), there is relatively little information about cleavage rates in late preimplantation development, mainly due to the paucity of surplus embryos after transfer or cryopreservation. The recent advent of treatment with agonists of luteinising hormone releasing hormone prior to superovulation, has increased the number of oocytes collected and fertilized and resulted in a higher pregnancy rate (Rutherford et al. 1988). The concomitant increase in the number of surplus embryos after transfer has thus facilitated analysis of later development in vitro. In this study, we examined a total of 181 surplus embryos, including both normally and abnormally fertilized embryos donated by 42 patients with, on average, 4.9±0.5 surplus embryos per patient.

The timing of development in vitro from day 2 onwards appears to be comparable to the few reliably timed observations in vivo. For example, Buster et al. (1985) recovered 19 embryos by uterine lavage 5 days after artificial insemination, 5 of which (26%) had developed to the blastocyst stage. This is similar to the proportion reported by us in another recent study, in which 10/43 (23 %) reached the blastocyst stage by day 5 in vitro (Hardy et al. 1989). Notably, amongst the embryos examined here, those reaching the blastocyst stage by day 5 were already more advanced on days 3 and 4 than those embryos not reaching the blastocyst stage until day 6. The total cell numbers in these day 5 blastocysts from individual patients were remarkably similar and not significantly different, whereas there was significant between patient variation (P<0.001). This suggests that among embryos from individual patients, the timing of fertilization and cleavage may be similar. However, on days 6 and 7, there was significant variation both within and between groups of embryos from different patients. This increase in variability may be caused by the increased cell death at these stages (Tables 2 and 3).

The minimum number of cells we observed in normally fertilized, expanded blastocysts on days 5 and 6 was 24 and 27 cells, respectively (Fig. 2). It is likely, therefore, that blastulation is initiated between the 4th and 5th cleavage divisions. This is earlier than in the mouse, where blastulation coincides with completion of the fifth cleavage division (Smith and McLaren, 1977; Handyside and Hunter, 1986); but similar to the pig, where it occurs mainly after the 4th cleavage division (Papaioannou and Ebert, 1988).

Total cell numbers in normally fertilized blastocysts increased from day 5 onwards and significantly between days 6 and 7 (P<0.05) (Table 3). Thus cell division continues in vitro at least up to day 7, although the mitotic index steadily declined to a low level. The hatched blastocyst with 283 cells on day 7 is particularly encouraging for these in vitro culture conditions. On average, the first three cleavage divisions take about 60 h (Edwards et al. 1981; Sundstrom et al. 1981; Trounson et al. 1982). Between days 3 and 7, the increase in total cell numbers indicates that, on average, a further four cleavage divisions have been completed in the same number of days. Thus, the overall cleavage rate of human embryos in vitro at morula and blastocyst stages appears to be no faster than the previous cleavage divisions, with one division every 24 h. However, this must be an underestimate because of the extensive cell death occurring at these stages. In the mouse, cleavage rates over this period are initially higher, with about one division every 12h at the morula stage. At the blastocyst stage, the rate decreases because of a combination of giant cell formation in the mural TE and slower division and cell death in the ICM (Handyside and Hunter, 1986). Cleavage rates in sheep embryos, however, are equally slow (Handyside et al. 1987).

Previous data on the numbers of cells in human blastocysts are sparse; however, our data are comparable. Hertig et al. (1954) recovered two blastocysts, estimated to be 5 days postfertilization, from uteri after hysterectomy. Cell counts were made by counting nuclei in serial sections and were 58 and 107 cells of which 5 and 8, respectively, were considered ICM cells on the basis of their position. Croxatto et al. (1972) obtained one blastocyst from uterine flushing which upon serial sectioning was found to have 186 cells but was thought to be more than 5 days old. Finally, Steptoe et al. (1971) counted the nuclei of two blastocysts grown in vitro for 7 days after IVF. One had 112 nuclei, 16 of which were in mitosis, the other 110 nuclei.

Not surprisingly, morphologically abnormal blastocysts and those derived from abnormally fertilized embryos had lower total cell numbers (Table 3). However, one blastocyst derived from a polyspermic embryo had 103 cells on day 6 which is well within the range for normally fertilized blastocysts (Fig. 2) suggesting that it had undergone normal cleavage. Van Blerkom et al. (1987) reported that some zygotes, which at the light microscope level appeared to have multiple pronuclei, have upon closer examination at the electron microscope level the normal two pronuclei closely associated with a membrane-bound cytoplasmic vacuole. These pseudo-multipronuclear embryos developed normally and have given rise to pregnancies. Therefore, it is possible that this blastocyst was in fact derived from a pseudo-multipronucleate embryo.

The allocation of cells to the TE and ICM of the blastocyst is of fundamental importance for later development. In the mouse, allocation to the two lineages occurs during the fourth (Handyside, 1981) and fifth cleavage divisions (Pedersen, 1986; Fleming, 1987) at the morula stage. Following implantation, the TE only gives rise to the placenta and extraembryonic membranes whereas the ICM forms all three germ layers of the fetus as well as complementary contributions to the extraembryonic membranes (Gardner and Papaioannou, 1975). With human blastocysts, apart from sporadic estimates of the number of these cells in sectioned embryos (see above) and descriptions of the ultrastructural features of the two cell types, cell allocation has not previously been examined in detail.

The timing and numbers of cells initially allocated to the inner and outer positions at the morula stage may be difficult to ascertain in the human embryo. Focal intercellular contacts and junctions in 8- to 16-cell human morulae have been observed (Lopata et al. 1983; Tesarik, 1989; Hardy, unpublished observations). However, compaction, in terms of the close apposition of cells and loss of spherical shape, is less complete than in the mouse embryo. Specific antibody labelling of the outer morula cells may, therefore, not be a reliable approach (Handyside, 1981). Instead, we have attempted to estimate cell allocation at the 16- and 32- cell stages by extrapolating from TE and ICM numbers in day 5–7 blastocysts (Fig. 3). On this basis, and ignoring cell death, at the 16-cell stage there would be approximately 6 cells allocated to the ICM, and 10 to the TE and, at the 32-cell stage, there would be approximately 12 and 20 cells allocated to the ICM and TE, respectively. These numbers are similar to the numbers found in the mouse (Handyside, 1981; Handyside and Hunter, 1986).

Between days 5 and 7, the numbers of TE and ICM cells in normally fertilized human blastocysts both doubled. This contrasts with mouse blastocysts in which, over a similar period, the numbers of TE cells quadruple, before reaching a plateau coinciding with the initiation of giant cell formation, whilst the number of ICM cells only doubles (Handyside and Hunter, 1986). Also, the maximum number of TE and ICM cells in mouse blastocysts was 147 and 49, respectively. With human blastocysts, the maximum number of TE cells was similar, 170, whereas the maximum number of ICM was higher, 113. The overall proportion of ICM is high: 34 % on day 5, 51 % on day 6 and 37 % on day 7. This contrasts with both the pig, where the proportion of ICM declines from ≈25 % at the expanded blastocyst stage to ≈17% at the hatched blastocyst stage (Papaioannou and Ebert, 1988), and the mouse, where the proportion declines from ≈40 % in nascent blastocysts to ≈17% before implantation (Handyside and Hunter, 1986). However, the high proportion of ICM in the human is similar to that in the sheep, where the ICM constitutes ≈44% of the expanded blastocyst (Handyside et al. 1987).

The widespread cell death, which we have observed as fragmented labelled nuclei in human blastocysts at all stages (Fig. 1C), was equally prominent in both the TE and ICM and increased markedly with culture of blastocysts to day 7 (Table 2). Cell death is a widespread feature of embryogenesis in both vertebrates and invertebrates (Glucksmann, 1951; Saunders, 1966) and generally occurs by apoptosis, characterized by cellular shrinkage with condensation of nuclear chromatin and cleavage of chromatin into oligonucleosome chains (Kerr et al. 1972; Wyllie, 1980). It has been observed in a wide variety of mammalian embryos at the blastocyst stage. Ultrastructural studies show cell death in the ICM of the cow and the human (Mohr and Trounson, 1982), the mouse (El-Shershaby and Hinch-liffe, 1974) and in both the TE and the ICM of the Rhesus monkey (Enders and Schlafke, 1981). Differential labelling confirmed that cell death occurs in the ICM of the mouse (Handyside and Hunter, 1986) but, in contrast, using a similar technique, none was observed in the pig blastocyst (Papaioannou and Ebert, 1988). In the cow, Rhesus monkey and mouse, the blastocysts had been flushed directly from the uterus, suggesting that cell death is a normal phenomenon occurring in vivo. The significance of these dead cells is not fully understood but it has been suggested that, in the mouse, it may involve the elimination of ICM cells retaining the potential to form TE (Handyside and Hunter, 1986). With human embryos, some of the degenerating nuclei may be derived from the multinucleate blastomeres observed at cleavage stages (Lopata et al. 1983; Tesarik et al. 1987).

Embryos arrested for several days at cleavage or morula stages had extensive cell death but also other nuclear and chromosomal abnormalities (Fig. 1C-E). The ‘flocculent’ nuclei observed in six arrested embryos on day 6 had the appearance of DNA clumped in vesicles on the nuclear membrane. It is of interest that in ultrastructural studies of cell lines undergoing induced apoptosis, margination of nuclear heterochromatin in dense clusters on the nuclear membrane has been seen (Bardon et al. 1987). Thus it is possible that these ‘flocculent’ nuclei are a feature of an intermediate stage of apoptosis, before fragmentation of the nuclei, or simply the early signs of degeneration after embryonic arrest.

Over a third, 63/149 (42%), of normally fertilized embryos examined developed to the blastocyst stage on day 5 or 6 in vitro (Table 1). Assuming that a similar proportion of embryos would have developed to the blastocyst stage in vivo and that all of these blastocysts had a high probability of implanting and establishing a pregnancy, the expected pregnancy rate, including multiple pregnancies, after transfer of three embryos would be about 70% (i.e. if the probability of an embryo arresting is 2/3, the probability of all three embryos arresting is (2/3)³=0.296, assuming that they are independent of each other. Thus 29.6% of the women would not become pregnant). In fact, of the 41 patients for whom information is available, only 17 (41 %) became pregnant, as determined by raised hCG levels. However, 15 out of these 63 blastocysts were morphologically abnormal and had significantly fewer cells and high levels of cell death (Table 3). In addition, amongst the other blastocysts low numbers of ICM cells on day 6 and 7 were associated with a high proportion of dead cells (Fig. 4A). This raises the possibility that blastocysts with smaller numbers of ICM cells may be less viable, and responsible for biochemical pregnancies where there is an initial increase in the levels of hCG but no fetal development. Whether these blastocysts with low cell numbers result from suboptimal conditions in vitro or inherent defects will require further investigation.

The authors are grateful to Karin Dawson, IVF laboratory director, and members of her team, for their care and expertise in the collection and establishment of the surplus embryos used in this study, and to Jaroslav Stark for assistance with the statistical analysis.