Development and phenotypic variability of genetically identical half mouse embryos

The capacity of the mammalian embryo to compensate for cellular loss during embryogenesis provides developmental flexibility during gestation, and also renders the embryo a pliant subject for manipulation, including bisection for the production of monozygotic twins. However, if regulation is not always complete, phenotypic variability might be introduced by embryo manipulation. In this study, genetically identical mice developing naturally and from transferred control and half embryos have been compared. Developmental success and several measures of phenotype have been made to better understand possible long-range effects of embryo manipulation.

The effect of decreasing embryonic cell number in the mouse was first studied by Tarkowski (1959a,b) who found that half embryos surviving until the 12th day of pregnancy were at the same stage of development as control embryos and that the birth weight of live young derived from half embryos was not significantly different from controls. Lewis & Rossant (1982) have presented evidence that mouse half embryos adjust to approximately normal size by the seventh day of gestation. Rands (1986) found that half embryos produced by destruction of one blastomere at the 2-cell stage regulate their size between 7·5 and 10·5 days although half embryos may again fall behind controls by 13·5 days. Similarly, Tsunoda & McLaren (1983) reported that 18-day-old mouse fetuses derived from half embryos weigh significantly less than those from controls. All of these studies indicate some extra variability among half embryos whether it is in embryonic viability or phenotypic variation among those that survive. None of these studies, however, addresses the question of whether variation among genetically identical half embryos could persist into adulthood. An answer to this question can be obtained by comparing the variances of several parameters in genetically identical populations of animals derived from either whole embryos or half embryos. Comparisons of the overall variances, and the between and within litter components of variance for these parameters provide estimates of the uniformity of individuals in these groups and thus a measure of the variation due to incomplete regulation following embryo manipulation.

The comparison of control embryo transfers with naturally reproducing controls is of interest in determining whether handling and transfer of embryos can affect their subsequent postnatal development, a matter of interest clinically and in the design of experiments. Since, 1929, when Gesell & Thompson (1929) introduced the experimental strategy of co-twin controls, monozygotic human and cattle twins have been recognized as potentially useful in the control of experimental error [see Biggers (1986) for a review]. The genetically identical half embryos described here are the most easily produced laboratory animal models for artificial monozygotic twins.

All mice were obtained from The Jackson Laboratory, Bar Harbor, ME, and were housed under standard conditions with a 14/10 h light/dark cycle. C57BL/6J females were mated with SJL/J males to provide F₁ embryos or with vasectomized C57BL/6J males to provide embryo transfer recipients. This strain combination provided coat color markers to distinguish all mice, since the parental strains are nonagouti (a/a C/C) and white (A/A c/c), while the F₁ mice are agouti (A/a C/c).

Females between 6 and 16 weeks of age were either mated naturally or after ovulation was induced with 2 5-10i.u. each of PMSG and HCG (Sigma) 44 h apart. Some naturally mated females were left to deliver their litters. Other mated females were killed on the second day of gestation (day of the vaginal plug being taken as the first day) and 2-cell embryos were recovered in a Hepes-buffered medium (Table 1). One of the two blastomeres of some embryos was lysed by piercing it through the zona pellucida with a solid glass needle (Fig. 1). These half embryos along with control embryos were incubated in the same medium but with bicarbonate buffer at 37°C for 10 min in 5% CO₂ in air. Virtually all of the punctured blastomeres lysed and in only a few cases did both blastomeres lyse.

Half embryos with one lysed blastomere and control embryos were then separately transferred to the oviducts of females in the first day of pseudopregnancy. Between 1 and 10 embryos were transferred, usually unilaterally to the left oviduct. In control transfers, the average was 7·2 embryos per oviduct (mode 7), for half embryos 7·9 embryos per oviduct (mode 10). A few bilateral transfers were done in each group and the maximum number of embryos transferred to any recipient was 18. At the time of embryo transfer, recipient females were given an energy-rich breeding diet (Mouse Chow No. 5015, Purina Mills, Inc.) which was continued until the time of weaning at 3 weeks when weanlings were put on standard chow (Rodent Laboratory Chow No. 5001, Purina Mills, Inc.). Naturally mated females providing nontransferred, control embryos were also given the energy-rich diet from the day of the vaginal plug detection.

Thus there were three groups of genetically identical mice for comparison: (1) controls - the natural offspring of timed matings, (2) transferred controls - offspring from embryos flushed at the 2-cell stage and transferred to recipients 1 day asynchronous, and (3) transferred half embryos - offspring developing from one blastomere from the 2-cell stage transferred to recipients 1 day asynchronous.

The day of birth and number of offspring was recorded. All females from the embryo transfer groups that failed to deliver were killed and examined for implantation sites, usually within a few days of the expected delivery date. At 5 days, the offspring were weighed to the nearest 0·01 g and their tail lengths were measured to the nearest mm from base to tip. When necessary, toes were clipped to distinguish individuals and all mice were given a distinguishing earclip at the time of weaning. Near the expected times, mice were checked daily for eye opening and vaginal opening. At weekly intervals between 2 and 5 weeks, each mouse was weighed and the tail measured.

All data was stored in a Lotus 123 spreadsheet. Initially, the data were examined by methods of exploratory data analysis (Hoaglin et al. 1983; Chambers et al. 1983). In particular, notched box plots were used to compare the order statistics (median, 10, 25, 75 and 90 percentiles) of observations on the three groups of offspring (McGill et al. 1978) using the Number Cruncher Statistical System (NCSS; Jerry L. Hintze, Kaysville, Utah). The normality of the distributions of the residuals after fitting a oneway linear model was tested using the Generalised Linear Interactive Modelling System (GLIM) (Numerical Algorithms Group Ltd., Oxford, UK) to compute normal probability plots. This procedure allowed the detection of outliers. Although several outliers were detected only one animal was eliminated from the data. This animal was a male runt in the control group which could be eliminated on biological grounds. The estimation of the between litter and within litter components of variance and their variance-co-variance matrix was made with a custom written program using the method of maximum likelihood (Searle, 1971; Swallow & Monahan, 1984). The intraclass correlation coefficients were calculated from the between and within components of variance (Donner, 1986).

The fate of control, transferred control and transferred half embryos is compared in Table 2. Among control females, 60% gave birth following mating. Following transfer of control or half embryos, 59% and 45 % of recipients carried a litter to term. Among the pregnant recipients, 51 % and 41 % of the transferred control and half embryos were born. Among recipients that failed to deliver, only a single female from the half embryo group had implantation sites in the uterus. That female had five large moles at 24 days post-transfer.

Fig. 2 presents the distributions of the number of offspring born per female in the three groups. The transferred control and the transferred half embryo groups are much more variable than the controls. The low median value in the half embryo group is due to the high incidence of litter sizes of one and two.

The male/female ratio among offspring surviving to 7 days was 1·05, 0·98 and 1·00 in the controls, transferred controls and transferred half embryos, respectively. These ratios do not differ from the expected 1:1 ratio for any group.

The duration of pregnancy for transferred embryos was timed from the day of the plug of the recipient female since timing of implantation is determined by the mother rather than the embryo in asynchronous transfers (Noyes et al. 1963). The data are summarized in Fig. 3 for the three groups. The duration of pregnancy was more variable and significantly longer in the transfer groups than the control group, but the transfer groups were not significantly different from each other. The mean duration of gestation in control females was 18·7 days, while for transferred control and transferred half embryos it was 19·2 and 19·8 days, respectively. There was a significant negative correlation (-0·68, P< 0·001) between the duration of pregnancy and the number of half embryos born, whereas no significant correlations were seen in the other groups.

Fig. 4 summarizes the pregnancy rates following natural mating or embryo transfer, embryonic loss during gestation and neonatal loss in the different groups. Overall embryonic survival to term was 34·4% for transferred controls and 18·2% for transferred half embryos.

Among 12 control litters only 4 offspring died perinatally, all from a single litter of 7. Among transferred control and half embryos bom, 86 % and 80 %, respectively, survived beyond 7 days. The 15 deaths among transferred control animals were from 7 Utters, 4 of which consisted of 1-4 offspring that all died. The 21 deaths among half embryos were from 16 litters, 12 of which consisted of only 1-2 offspring that died.

The distributions of the ages at which the eyes opened in the three groups of animals is shown in Fig. 5. There is no difference between the three groups, the overall mean age of eye opening being 13·9 days. The distributions of the ages at which the vagina opened in the female offspring in the three groups is also shown in Fig. 5. There appear to be no significant differences between the three groups with respect to this parameter, the mean age of vaginal opening being 28·6 days.

The mean body weights of the male and female offspring of the three groups of mothers for the first five weeks of postnatal life are shown in Fig. 6. Within each sex these appear to be very similar although, as we shall see, there are significant differences between body weights. A sexual difference can be detected by the third week, the weight of the males being greater.

The overall variances of the body weights of the male and female offspring in the three groups are shown in Table 3. The results show clearly that there are no significant differences in the variation of the three groups during the first week after birth. However, from the second to the fifth weeks, the variances of the offspring derived from transferred control and half embryos are significantly more variable than those from the controls. In contrast, there are only a few significant differences in the variability of the transferred controls and the half embryos and the magnitude of these differences is relatively minor. These differences are confined to the female groups, the offspring from the transferred control group being more variable.

The large changes in the distributions of the body weight of the offspring that occur between day 5 and week 2 after birth are displayed for each sex in Fig. 7 as notched box plots. The body weights from the transferred control and transferred half embryos in both sexes are not significantly different but they are larger than the median body weights of the controls. Also in both sexes the interquartile range (25-75 percentile) of the body weights of the transferred control and transferred half embryo offspring at 2 weeks is considerably greater than the controls. This result demonstrates that the increased variability is not due to a few extreme individuals.

In contrast to the body weights, there is no difference between the tail lengths of the two sexes within each group. The mean tail lengths of the combined male and female offspring of the three groups of mothers for the first five weeks of postnatal life are shown in Fig. 8. The overall variances of the tail lengths of the male and female offspring in the three groups are shown in Table 4. As with the data on body weight, the results show clearly that there are no significant differences in the variation of tail length in the three groups during the first week after birth. However, from the second to the fifth weeks, the variances of the offspring derived from transferred control and half embryos are significantly more variable than those from controls. In contrast, there are only relatively minor significant differences in the variability of the transferred controls and half embryos. These differences are confined to the female groups, the offspring from the transferred control group being more variable.

The large changes in the distributions of the tail lengths of the offspring that occur between day 5 and week 2 after birth are displayed for each sex in Fig. 9 as notched box plots. In both sexes, the interquartile range (25-75 percentile) of the tail lengths of the transferred control and half embryos at 2 weeks is considerably greater than the controls. This result demonstrates that the increased variability is not due to a few extreme individuals.

Since the data have been obtained on newborn from several mothers, the overall variances of the body weights and tail lengths shown in Tables 3, 4 can be partitioned by a components of variance analysis into the variation between litters and the variation within litters. This analysis has been done assuming the oneway random effects linear model (Searle, 1971), given by:

where y_ij is the observation on the jth embryo in the ith litter, μ is the overall mean, l_i is the deviation of the ith litter, e_U is the deviation of the ijth embryo (sometimes referred to as a residual), r is the number of litters, n_i is the number of offspring in the ith litter.

It is further assumed that the l_i, are distributed with mean zero and variance σ_i² and the e_ij are distributed with mean zero and variance σ₁². The σ₁² and σ_e²are called the between and within litter components of variance respectively. Estimates of σ₁² and σ_e² are of prime interest in the study of natural variation (Fisher, 1925; Fieller & Smith, 1951). However, a derived statistic, the intraclass correlation coefficient (Donner, 1986), given by:

is of significance in the planning of some types of experiment. For example, incomplete block designs using identical twins are of no advantage unless the intraclass correlation coefficient is above a critical value (Biggers, 1986).

The estimates of the within and between litter components of variance of the body weights of the male offspring in the three groups are shown in Table 5. The increases in overall variation observed between the control and transferred groups for both parameters can be due to increases in the within and between litter components of variance separately or in combination. More importantly the ratios of the within to the between litter components of variance (σ_e²/σ₁²) for both parameters (Equation I) tend to be the same in all three groups. Consequently the intraclass correlation coefficients for both the body weights and tail lengths that are observed in the transferred control and transferred half embryo groups are as high as, and sometimes greater than, those observed in the control groups even though the components of variance are greater (Table 6).

Similar results were obtained in components of variance analyses for the body weight data of the females and the tail length data of both sexes. The intraclass correlation coefficients of all data are shown in Table 6.

Published accounts of half embryo survival vary greatly (Papaioannou & Ebert, 1986) and it has been eported as high as 65 % for bisected 2-cell embryos near term (Tsunoda & McLaren, 1983). In the present study, we achieved 18 % survival of half embryos to term overall or 41 % among pregnant recipients compared with 34 % overall and 51 % among pregnant recipients for transferred control embryos. The decreased success of half embryos can be accounted for by a combination of lower pregnancy rate, higher losses during gestation and higher neonatal loss. In general, a lower pregnancy rate could be the result of failure of embryos to be fertilized, of abnormal development or of hormonal failure of the mother. Our study indicated that as high as 40 % of natural matings fail to result in delivery of offspring at term and a similar figure was found following control embryo transfer to females mated with vasectomized males. Transfer of half embryos, however, resulted in an even lower pregnancy rate, with 55 % failure. Our results indicate that this increased loss of half embryo litters is embryo-related, reflecting either a diminution of the ability of half embryos to initiate implantation or a lowered viability such that embryos implant but whole litters die before parturition. The absence of moles, implantation sites or scars in the uteri of females that did not deliver offspring argues either for failure to implant or death very soon thereafter (Edwards & Fowler, 1959).

Among females that delivered at least one offspring at term, embryo loss during gestation was slightly greater among half embryos than transferred controls. We could not distinguish between pre- and postimplantation death since the uteri of these recipients were not examined postpartum. Together with the pregnancy failure, the lower viability of half embryos results in a 16% greater loss at term than transferred controls, in agreement with the decreased viability of half embryos compared with controls reported by others (Tsunoda & McLaren, 1983).

Higher neonatal mortality was also found in mice that developed from transferred control or transferred half embryos compared with naturally developing controls. It is known that small litter size increases the duration of pregnancy and that this in turn is associated with increased neonatal loss (McLaren, 1970). In our study, embryo transfer, regardless of the embryos transferred, resulted in small litters and the duration of pregnancy was increased in inverse proportion to the litter size at least among the transferred half embryos. Thus, the increased neonatal loss can be explained as an indirect effect and not as a reflection of decreased neonatal survival potential of mice developing from half embryos. While surviving half embryos appear to regulate their cell number and size shortly after implantation (Lewis & Rossant, 1982; Rands, 1986), the considerably lowered viability may indicate that embryos that do not successfully regulate are eliminated during the early stages of gestation (Papaioannou, 1989).

In this study, we have chosen to produce half embryos by the destruction of one blastomere rather than by embryo bisection to produce twin pairs. Blastomere destruction is easy to accomplish and results in a high rate of developmental success, thus allowing us to compare large numbers of animals for accurate estimates of variation. It does not, however, allow us to address the issue of the similarity in developmental potential of two halves of the same embryo. Using a small number of twin pairs produced by bisection of 8-cell embryos, Gartner & Baunack (1981) and Baunack et al. (1986) reported results in agreement with ours with respect to the similarity of mean and overall variation in body weights between half embryos and controls. In addition, they found that bisected pairs were more similar to one another than were non-bisected but genetically identical control pairs.

They interpret this result as indicating some environmental influences on the early embryo (prior to bisection) that create postnatal variation. Thus bisected pairs that share this early effect are more similar to each other than are randomly selected pairs. However, since only 5 % of bisected pairs survived in their study, an alternative explanation is that dissimilar pairs were developmentally disadvantaged and only bisected pairs in which each pair had equal potential were capable of survival to term.

The estimates of variation in body weight and tail length in the different groups at different times after birth shown in Tables 3-5 arise from several causes. The increase in variation of both parameters that occurs over time is typical of growth data and is due to scaling. In comparing the changes in variation over time, however, it is important to note that the estimates arise from repeated measurements on the same individuals and that the estimates are serially correlated. More important are the independent comparisons between the treatment groups at each time. The differences are best interpreted in terms of the forces that influence the development and growth of an individual.

Kempthome & Osborne (1961) discussed theoretically the forces that influence the variability of monozygotic human twins. They classified these forces into non-competitive forces which favor uniformity of siblings, and competitive forces which promote diversity. Similar principles can be used where litter size is greater than two. A list of the non-competitive and competitive forces adapted to the present experiment follows using the same notation as Kempthorne & Osborne (1961).

(G) The individuals in a litter have the same genetic structure with a high degree of probability,

(M) The individuals have passed through development in the same uterus,

(P) The individuals have shared the same gross prenatal environment; they have not only shared the same uterus but they have done so at the same time, so that parity, maternal age and maternal health were constant,

(N) The individuals have shared the same neonatal environment until the time of weaning,

(I) The individuals after weaning have shared a constant animal house environment.

(Z) The individuals may receive unequal cellular contributions due to differences in the number and nature of the blastomeres,

(V) The individuals may experience intrauterine competition for space and nutriments,

(S) The individuals may experience competition after birth, before weaning, such as competition for milk.

Each of these factors are associated with a variance (σ₁²). At any one time, the overall variation is a linear combination of these separate variances. However, only the competitive forces vary between the treatment groups in the experiment described in this paper.

The observed variability in a given parameter will depend on its susceptibility to competitive forces. In this respect, the four parameters we have studied vary widely. The time of eye opening is probably the least sensitive, being a highly inherited character. The variation in the body weight of mice for up to seven days after birth is largely influenced by prenatal factors (El Oksh et al. 1967), while by the age of two weeks the variation is largely determined by postnatal factors particularly milk supply (Cox et al 1959; Young et al. 1965; El Oksh et al. 1967; Nelson et al. 1976). At about two weeks after birth, when the eyes open and the offspring begin to eat solid food, other influences on the variation in body weight become operative. Vaginal opening is partially dependent on body weight and the number of males in the litter (Drickamer, 1976). The growth of the tail in mice, a major heat loss organ, is sensitive to body temperature, being long in animals reared in hot temperatures and short in those reared at low temperatures (Sumner, 1909, 1913, 1915; Barnett, 1956). The variation in length is influenced by both endothermic factors, such as an individual’s metabolic rate, and ecto thermic factors, such as local temperature in the nest (Barnett, 1956). The latter are likely to dominate in the 10 days after birth since up to this time mice are unable to regulate their body temperatures (Sumner, 1913; Barnett, 1956).

The overall variances of the body weight and tail length of both sexes are not significantly different during the first week after birth. However, the median body weights of males and females in the control group are lighter than those in the transferred control and transferred half embryo groups (Fig. 7A,B), which themselves are not significantly different. In contrast, the median tail lengths of the offspring of both sexes do not differ significantly between the three groups (Fig. 9A,B). The lower weights of the controls can be explained by the fact that the litter size of this group is larger than in the other two groups. Although this suggèsts that there was some competition for space in the uterus (competitive factor V), it was insufficient to increase significantly the variance. The fact that no significant differences have been detected between transferred half embryos and transferred controls suggests that the loss of a blastomere at the two-cell stage is fully compensated for in those individuals that survive to birth.

From the second to the fifth week inclusive, the variances of the body weights of the two transferred groups are significantly greater than those of the control group. In contrast the variances of body weights of the transferred groups are not significantly different. The lower body weights of the control offspring persist until the fifth week after birth, but by this time the males in all groups are significantly heavier than the corresponding females. Since the increase in variation in the transferred groups only occurs after the first week of age, it is presumably caused by postnatal environmental influences (Factor S). These could arise as the result of the smaller litter sizes than in the controls and be partially due to scaling because the offspring from the transferred groups are larger.

The difference in body weight between the groups is not reflected by a significant difference in the time and variation of vaginal opening, which occurs about the 28th day after birth. We may not have detected a relationship between the variables, since body weight was not necessarily measured on the day of vaginal opening.

From the second to the fifth week, inclusive, the variances of the tail lengths of the two transferred groups are significantly greater than those of the control group. In contrast to the body weight, the increased variance in the transferred groups cannot be attributed to scaling since the tail length of the transferred groups is no longer than the control group. Since the increase in variance in both transferred groups occurs only after the first week, it is caused by postnatal environmental factors to which the offspring are particularly sensitive. This factor could be the ambient temperature in the microenvironment of the nest which may be more variable in small litters.

Genetically identical litter mates, even if there are only two, can be usefully exploited to reduce experimental error in the design of comparative experiments by using incomplete block designs. The larger overall variances of body weights and tail lengths observed between offspring of transferred control and half embryos once they are two weeks old suggests that animals resulting from embryo transfer will have limited use in comparative experiments. However, the usefulness of litter mates depends on the value of the intraclass correlation coefficient (ρ). Provided ρ>(l-E), where E is the efficiency of the experimental design, the litter mates will be useful (Biggers, 1986). Equation I shows that p is inversely related to the ratio of the within and between litter components of variance and not on their absolute values. The estimates of the intraclass correlation coefficients in the control and transferred groups at all ages presented in Table 6 are relatively invariant compared with the large differences in the components of variance from which they were derived. Thus, litter mates produced by embryo transfer, even after surgical manipulation, may be sufficiently uniform that they can be utilized in the design of comparative experiments to reduce experimental error.

The work described in this paper was supported by a grant from the National Institutes of Health (RR02510). We wish to thank Dr Joel Lawitts and Carol Kountz for assistance in the production of this manuscript.