Superpregnancy in the Mouse: 1. Implantation and Foetal Mortality After Induced Superovulation in Females of Various Ages

In 1927 Engle used pituitary implants to induce ovulation and oestrus in adult female mice mated to fertile males. At post mortem examination of their uteri 9−10 days later he found large numbers of embryos, ranging from 19 to 29. He did not follow the fate of such embryos into later pregnancy. But his observation raises a question which remains unanswered today : ‘Could the reproductive output of mice and other mammals be artificially increased by gonadotrophin treatment?’

Not only adults but also immature female mice respond to pituitary gonadotrophins by the ovulation of large numbers of eggs (superovulation) (Runner, 1950 ; Runner & Palm, 1953). If treated immature females are paired with fertile males copulation occurs, followed by fertilization and cleavage of the eggs (Runner & Gates, 1954). These eggs can be shown by transfer to uterine foster-mothers to possess normal viability and developmental capacity (McLaren & Michie, 1956; Gates, 1956).

The failure of implantation to occur in immature females following induced ovulation and mating was investigated by Smithberg & Runner (1956). The corpora lutea of the immature animals began to regress 4 or 5 days post coitum instead of developing into corpora lutea of pseudo-pregnancy (for terminology see Snell, 1941). Implantation could be induced, and pregnancy maintained to term, by supplying the missing progesterone artificially. The rare animals in which implantation occurred spontaneously were invariably found to have functional corpora lutea.

When pituitary gonadotrophins are administered to adult females without regard to their oestrous phase, a majority respond by superovulation, mating and implantation (Fowler & Edwards, 1957). The proportion of mated females in which implantation occurs is lower than in natural mating. Following implantation, embryonic mortality during pregnancy and parturition is abnormally high, so that the number of young born alive is no greater than normal.

On receiving privately from Drs Fowler and Edwards an account of their work, we made a preliminary study‡ of the ages at which superpregnancy could be induced. In the strains of mice used, superovulation and mating at 30 days of age was regularly followed by pregnancy, although spontaneous sexual activity, and hence the possibility of natural pregnancy, does not appear until some 5−10 days later. We shall refer to this stage as adolescence. It is preceded by the juvenile stage, studied by Runner & Gates (1954) and by Smithberg & Runner (1956), in which induced ovulation is followed by mating and conception but not by implantation, and is succeeded by the adult stage in which oestrus and ovulation occur spontaneously.

It is clear from the foregoing facts that before we can hope to increase the reproductive yield, or decrease the reproductive age, by artificial means, many questions still require answers. The following are among the questions to which the present work is addressed :

1. When does adolescence begin?

2. How do adolescent females compare with adults in their capacity to initiate and maintain superpregnancy?

3. What is the pattern of embryonic mortality in adolescent and in adult superpregnant females?

4. Are the large prenatal losses in superpregnancy due to the excessive number of embryos carried or to some other attribute of the experimental treatment?

The experiment to be described also provided information on embryonic growth. The presentation of this material will be deferred to a later paper.

The mice belonged to a random-bred colony of Theiler’s Original strain (TO strain), maintained at the National Institute for Medical Research, Mill Hill, London. All females were virgin. They had been reared on the M.R.C. diet 41, but were fed throughout the experiment on diet 86 of the Rowett Research Institute, supplemented with mixed grain. Ovulation was induced by the intraperitoneal injection of 0·2 i.u. of pregnant mare’s serum (‘Gestyl’, manufactured by Messrs Organon Ltd.) per g. body weight at about 5 p.m. 2 days before pairing with males, followed by the intraperitoneal injection of 0·2 i.u. of human chorionic gonadotrophin (‘Pregnyl’, manufactured by Messrs Organon Ltd.) per g. body weight at 12 noon on the day of pairing. Females were weighed individually immediately before the first injection, and hormone doses were calculated from these weights.

Treated females were paired singly with males of proved fertility. The occurrence of mating was detected by the presence of a vaginal plug the following morning. The females were thereafter stored in groups of five in larger cages. Untreated control females were also kept in groups of five, with a fertile male in continuous residence. They were examined daily for vaginal plugs.

Seven days after the finding of a plug a laparotomy was performed. The uterine horns were exposed through a ventral incision and the number of swellings in each horn was counted. Care was taken to handle the uterus, and especially the swellings, as little as possible. We have found that this operation neither interferes with the maintenance of pregnancy nor increases embryonic mortality. In normal pregnancies the number of swellings counted at $712$ days post coitum agrees well with the number of implants found in the same animal at $1812$ days. When the number of swellings is increased by superovulation, the $712$-day count becomes more difficult. Occasionally two embryos which have implanted very close together are contained in a single swelling. In cases of extreme crowding the whole uterine horn becomes thickened and the individual swellings are poorly defined. Where there is early embryonic mortality the $1812$-day count is also less reliable. None the less, our $712$-day and $1812$-day counts tallied in the great majority of cases (see Table 1).

Most of the pregnant females were killed at $1812$ days post coitum, i.e. shortly before term. Their uterine contents were then classified into live embryos and dead embryos, and the latter were subdivided into early, middle, and late deaths. A few females were found at $1812$-day autopsy to show the white blotches on the liver characteristic of Tyzzer’s disease. These animals have been excluded from the data.

The classification of embryonic death into early, middle and late was made as follows. Implants which were represented at $1812$ days by nothing more than small brown wart-like objects were described as early deaths. These wart-like objects, which we shall term ‘moles’, are quite distinctive and are familiar to all who have dissected pregnant mice. They have been variously referred to as ‘moles’ or ‘solid moles’ (Snell & Picken, 1935; cited by Grüneberg, 1952, p. 144), as ‘dead implantation sites’ (Grüneberg, 1952, p. 500) and as ‘deciduo-mata’ (Bateman, 1958), a term which calls attention to the overwhelming predominance of maternal tissue in these structures. Like Bateman, we have found no evidence of any tendency for the deciduoma to shrink and disappear by resorption after the death of the embryo. In the present work, and also in unpublished material, we have found that if an embryo lives long enough to cause a localized swelling at $712$ days it is always represented by a recognizable relic at $1812$ days. We are here referring to normal pregnancies (‘spontaneous ovulation’ in Table 1). In superpregnancies, other factors, as explained earlier, affect the accuracy of the $712$-day count.

The stage at which early death, as defined above, occurs does not seem to be very accurately or certainly known. Bateman (1958) cites Russell, Russell & Kimball (1954) as believing that it occurs before $1012$ days. He himself considers that such embryos have been dead ‘long before $1012$ days and probably since implantation began’ (i.e. since 5 days). Both Russell et al. and Bateman were considering early deaths induced by X-irradiation of the father’s testes, but they presumably intend their statements to hold for the early deaths occurring in their control material. In the case of early deaths among the offspring of males carrying X-ray-induced translocations Grüneberg(1952, p. 500) says: ‘The bulk of the embryonic mortality… happens after implantation, presumably during the formation of the germ layers’ (i.e. about 7 days). We shall give the period of early death the relatively non-committal designation of < 10 days.

Any implant which was represented at $1812$ days by a well-formed (but typically undersized) placenta without a recognizably formed embryo was classified as a middle death. We assume that these placental relics correspond to deaths occurring after day 9 (otherwise there would not be a well-defined placenta) and before day 13 (otherwise there would be, as we have inferred from our experience with other material, an embryo of grossly recognizable bodily form). However, we emphasize that these assumptions are based on informed guess-work only, and may possibly be in error by as much as one day in either direction.

The third category of embryonic death which we distinguished was based upon the recovery at autopsy of an embryo with recognizable bodily parts. We take such embryos to have died at some time after 12 days.

The vital statistics of the experiment are set out in Table 2, which shows the numbers and fate of the females allotted to the various groups. It presents a number of striking features, which are depicted in Fig. 1.

In the four groups of immature females, the proportion which mated following treatment rose from 11 /18 and 10/18 in the two youngest groups to 15/16 and 16/17 in the two oldest. This trend is significant at the 1 % probability level as judged by a χ² test for regression (Fisher, quoted by Holt, 1948). Evidently in the youngest groups there is a proportion of mice in a pre-juvenile stage in which oestrus, or at least mating behaviour, cannot be evoked by the gonadotrophin treatment.

The mating rate, having attained a peak at 30 days, falls to a lower level among the treated adults. The observed proportion, 30/53, differs from the 30-day value, 16/17, at the 1% level of significance as assessed by Fisher’s exact method. The possibility suggests itself that in some of the adults a clash may occur between the treatment and the phase of the spontaneous oestrous cycle in which the female happens to be at the time of the treatment. But in a more recent experiment, in which the oestrous phase at PMS injection was determined in 60 treated adults by the vaginal smear method, the mating rate was found to be independent of oestrous phase.

We have defined adolescence as the stage in which induced ovulation and mating are followed by implantation, but spontaneous sexual activity does not yet occur. The onset of adolescence can be accurately charted from the data of Table 2. In the 24-day-old group, implantation occurred in only 2 out of 11 mated females, as judged by the presence of $712$-day uterine swellings. The proportion rose rapidly, via 7/10 in the 26-day-old group and 14/15 in the 28-day-old group, to 16/16 in the 30-day-old group. The average age at which adolescence begins in this particular strain can be estimated from these data as about 25 days. It should, however, be emphasized that females of the TO strain mature unusually early. It is clear, for example, from Smithberg & Runner’s (1956) work that there are strains in which adolescence does not begin until after 35 days of age.

The mean numbers of implants in the various groups, as judged by $712$-day uterine swellings, are given in Table 3. It is clear that there is no difference between adolescent and adult treated females. Both show slightly more than twice the mean number of implants found in untreated control females. The consistent slight excess of the right horn over the left in mean number is in our experience a usual feature of normal mouse pregnancies also.

A very striking difference, however, is seen between adolescent and adult treated females in the amount of scatter about the mean. This is shown graphically in Fig. 2. A possible explanation may again be sought in the existence of a spontaneous oestrous cycle in the adult, but not in the adolescent, females. If the treatment happens to be in phase with the spontaneous cycle of the female, the effects of the treatment may be accentuated ; but if it clashes with the spontaneous cycle, its effects may be reduced. In a random assortment of females some will be at a favourable, and others at an unfavourable, phase of oestrus when the treatment is applied, while others will occupy a neutral position. The net result will be an increased variability in the responses to the treatment, as compared with adolescent animals, but no shift in the mean.

If this explanation is correct we expect that the whole of the observed difference in variability will be due to greater variation from female to female, and not at all to any difference in the amount of side-to-side variation. We can put the matter to test by performing an analysis of variance, taking individual uterine horns as the units of analysis. The results, shown in Table 4, accord fully with this expectation. The mean square estimating the variance within mice (i.e. between the two uterine horns of the individual mice) is approximately the same in the two groups, the adolescents actually having a slightly higher value. But the ‘between mice’ mean square for the adults is about 4 times the corresponding value for the adolescents (P < 0·001). To bring out the full force of the comparison, the ‘between mice’ item is also represented in Table 4 by the appropriate component of variance from which effects of side-to-side variation are entirely eliminated. The ‘between females’ component of variance for the adults is 12 ·8 times that for the adolescents.

A more familiar, but less informative, way of expressing these results is to say that there is a greater positive correlation between sides in adult than in adolescent implant numbers. The ratio of the ‘between mice’ to the ‘within mice’ mean square is in fact equal to (1 + r)/( 1 − r), where r is the intraclass correlation coefficient. The values of r derived in this way are given in the table.

The spontaneously ovulating controls had a similar ‘within females’ mean square (8·3333) to that of the treated groups, but a very small ‘between females’ item (2·1664). The corresponding correlation coefficient is −0·59. This presumably reflects the negative correlation between the numbers of eggs shed from the two ovaries in spontaneous ovulations in the mouse (Bowman & Roberts, 1958).

Turning back to Table 2 we see that a proportion of pregnancies failed, either by the disappearance (presumably by abortion) of the embryos between day $712$ and day $1812$ or by the presence at $1812$ days of dead embryos only. In Table 5 we have tabulated the frequency of failure in relation to the female’s age and to the number of $712$-day uterine swellings. There is a suggestion that the younger females are least able to maintain pregnancy to term. But the chief factor responsible for the loss of the experimental pregnancies seems to be the number of implants. The table shows that in the treated groups the mean number of $712$-day uterine swellings is consistently higher in treated females which subsequently lost their pregnancies than in their more successful contemporaries. The phenomenon is especially marked in the three youngest groups.

The average numbers of live and dead embryos in those females which carried at least one live embryo until autopsy at $1812$ days are set out in Table 6 and summarized in Fig. 3. The adolescents are seen to be at least as capable as the treated adults of maintaining large numbers of embryos alive. Indeed, the treated adults inflicted a mortality rate upon their embryos which is about $112$ times that found in the adolescents, and succeeded in bringing, on average, 2·7 fewer live embryos to $1812$ days. The additional mortality is entirely accounted for by the large numbers of early deaths (< 10 days), which amount to no less than six per pregnancy.

The gross mortality for the various groups is also entered in the table. This is the percentage of all implants which failed, for whatever reason, to be represented by living embryos at $1812$ days. Thus not only the partial loss but also the total loss (i.e. all embryos gone or dead at $1812$ days) of pregnancy contributes to the estimate. It has approximately the same value for the adult as for the adolescent treated females, since the heavier early mortality in the adults is counterbalanced by the higher proportion of total failures of pregnancy among the adolescents.

The occurrence of very large numbers of implants is, as we have said, a major factor in the loss of entire pregnancies. Is it also largely responsible for the partial losses sustained through the death of individual embryos ? In examining this question we can distinguish between embryonic deaths occurring during the early, middle and late periods of post-implantation existence. This distinction turns out to be of great importance. To anticipate our conclusions, the incidence of early death bears no relation to the total number of implants either per female or per uterine horn, whereas death at later stages (after 9 days) is found to be strikingly dependent on intra-uterine crowding.

In Table 7 the percentage incidence of early death (< 10 days) is shown after dividing the females into those with small and those with large numbers of implants. It is clear that there is no relation between the early death rate and the number of implants.

The cause of the heavy early embryonic mortality in the treated females, which, surprisingly, affects the adults even more than the adolescents, is obscure. Bateman (1958) says of early embryonic death in the mouse that it ‘seems to have few possible causes other than true dominant lethals’. But lethal genes cannot explain the differences shown in Table 7, since the three groups of females and their mates were closely similar in genetic constitution. Bateman also mentions, as a possible rare cause of early death, maternal inadequacy in the development of the deciduomatal reaction at implantation. One might reasonably suppose such inadequacy to occur more frequently after induced ovulation than in normal pregnancy, since disturbances in the luteinization of the ovarian follicles might follow their forced maturation and rupture. In that event, the immediate cause of early deaths would be progesterone insufficiency around the time of implantation. However, in a supplementary experiment where 10 out of 20 superpregnant adults received a subcutaneous implant of 100 mg. of progesterone (‘Lutocyclin’ implantation tablet, CIBA Ltd.) on day 3 post coitum, the incidence of early death showed no significant reduction in the females receiving progesterone.

There is a third possible cause of early embryonic death, not mentioned by Bateman but suggested by the work of Runner & Palm (1953). This is lack of synchronization between the shedding of the eggs and the arrival of the sperm, resulting in the fertilization of over-ripe eggs which develop for a time and even implant before dying.

A very different picture emerges when we consider embryonic death during the middle period, namely days 10−12, which is marked by the transference of the embryo’s nutritional dependence from the yolk-sac to the true chorioallantoic placenta. It is the persistence of this structure unaccompanied by a recognizably formed embryo which we have taken as our criterion for assigning embryonic deaths to this stage. In Table 8 we have arranged the relevant data according to the number of implants in each uterine horn which survived the first 9 days of life— i.e. the total number of implants less the number of ‘moles’. In this way we assess the possible effects of living competitors in the same uterine horn upon an embryo’s chances of survival during the middle period.

These effects are seen to be very great. Where the number of implants in a uterine horn which survive to the middle period is 8 or less, the death rate during this period is moderate and does not significantly exceed the control level. But as soon as the number rises above 8, the death rate in both the adolescent and adult treated groups jumps to 20−30% of all embryos surviving to the beginning of the middle period. The abruptness of the transition is brought out strikingly in Fig. 4. This entire holocaust occurs within a short span of time, probably three days, and must be regarded as a major catastrophe of placentation. It is the more remarkable in that its full force appears to be visited even upon those females which transgress the limit of eight by only one extra live implant, and shows no subsidiary gradation within the ‘8 or less’ and ‘9 or more’ categories. The daily injection of 3·6 mg. of progesterone from day 9 failed to suppress the phenomenon in a further batch of 14 superpregnant females. It is worth noting that, despite an aberrant control female which had ten implants in one horn and one in the other, eight is usually the maximum number of implants which a single horn is called on to bear in normal pregnancies.

We may inquire whether the whole of the effect is dependent upon the degree of crowding of the individual uterine horn, or whether it is to some extent also affected by the load carried by the opposite horn of the same female. An appropriate arrangement of the data, presented in Table 9, reveals a striking difference between the adolescent and adult treated females in this respect. The latter show no effect of crowding in the opposite horn. But in the adolescent group, doubly crowded females (i.e. with more than eight implants in each horn) have more than three times the middle death rate found in crowded horns when the other horn is not crowded. The difference is significant at the 10⁻⁶ probability level.

In considering the effect of uterine crowding upon the middle death rate we have ignored possible effects of early deaths. A superficial inspection of our complete data suggested that early deaths might have some protective effect upon the incidence of later death, in that there was a negative correlation between the number of ‘moles ‘in the horn and the proportion of the remaining implants which are found as placental relics at autopsy. Fuller analysis, however, showed this to be no more than an indirect reflexion of the phenomenon which we have just described: horns which had suffered a heavy early death rate tended to have the number of implants surviving beyond 9 days reduced to eight or less, while horns with few or no ‘moles’ tended to carry larger numbers of implants alive into the middle period and hence to qualify for a heavy middle death rate. After allowance for this, early death was found to have no effect: the occurrence of ‘moles’ in a uterine horn constitutes in itself neither a good nor a bad prognosis for the survival of the remaining implants.

Compared with early and middle death, the incidence of late death is low, amounting even in the superpregnancies to no more than 9 % of all deaths and 3 % of all implants. None the less, the numbers are sufficient to give a suggestion, as shown in Table 10, that here again the number of embryos carried is a causative factor.

Once again we need to know whether it is the total reproductive load or the local crowding of the individual uterine horn which is primarily responsible. We can answer the question by classifying each of the thirty-three late deaths which occurred in the experimental pregnancies according to whether it occurred in the more crowded or the less crowded of the two uterine horns. For this purpose we have taken as our measure of crowding the number of implants which survived beyond 9 days (i.e. middle deaths + late deaths + live embryos). At the same time we have calculated for each pregnancy the number of late deaths expected to occur in the two horns on the assumption that crowding does not play a causative role. For example, if two late deaths occur in a pregnancy with fifteen implants entering the late period alive, six in the right horn and nine in the left, the expected number of late deaths in the right horn is 2 × 6/(6 + 9) = 0·8. The expectations are summed and compared with the corresponding numbers of late deaths which are observed in the more crowded and the less crowded horns respectively. The result is striking and clear-cut, as shown in Table 11. The adolescent and adult females were in close agreement and have been combined. 82 % of the late deaths occurred in the more crowded horn, as compared to a chance expectation of 54%. The difference is statistically highly significant. Hence the effect of crowding is to a large extent local, that is, confined to the individual uterine horn.

When middle deaths are not included for the purpose of deciding which of two horns is the more crowded, the effect is still present but is reduced below the level of significance. It is possible that embryos which die during the middle period may effectively compete with their surviving siblings by contributing to intra-horn crowding. This would contrast with embryos dying in the early period, which, as we have seen, do not in any way affect the chances of those which outlive them. Doomed embryos which survive into the period of placentation might perhaps limit the size of the placentae formed by their surviving neighbours. This is in harmony with the parallel finding, reported by Healy, McLaren & Michie (1959), that embryonic growth is retarded by the presence of other embryos in the same horn if, and only if, these survive beyond the early period. We may suppose that limitation of placental size would restrict the rate of uptake of nourishment, and hence diminish embryonic growth, and, in the more extreme cases, viability.

A further group of adult (6 weeks) TO strain females was allowed to go to term following induced superovulation and mating. $712$-day laparotomy was performed and the number of uterine swellings was recorded. In Fig. 5 the number of liveborn young is plotted against the number of $712$-day swellings. A strong inverse relation is seen, showing that birth represents a second major stage of ‘densitydependent’ loss of young in superpregnancy, the first being the middle period of prenatal life described earlier. When the number of implants is large, the number of young born alive is actually less than in normal pregnancy, a result which calls to mind that of Parkes (1943) who found that the litters produced by adult rabbits following superovulation and mating were abnormally small. It should, however, be pointed out that we used mice which had only just crossed the threshold of sexual maturity, so that their youth may have contributed to the observed effect.

The relation of embryonic survival to the site of implantation in the uterine horn, independent of intrauterine crowding, has been described and discussed elsewhere (McLaren & Michie, 1959a).

At the outset of this paper the possibility was mentioned of artificially increasing the reproductive output of mice and other mammals. The question is not entirely academic. The suggestion has been made (Adams, 1954) that the induction of pregnancy in sexually immature animals might be used in livestock farming to diminish the interval between generations. Not only an increase in productivity but also an acceleration of progress in selective breeding could in principle be thereby achieved. Studies have already been made (e.g. Robinson, 1951) of induced superovulation in sheep and other farm animals as a possible means of increasing productivity; these studies have not been extended, so far as we know, to sexually immature females.

In mice we at first thought that the adolescent phase of sexual immaturity, in which induced superovulation and mating is followed by implantation and pregnancy, would prove too early a stage to obtain good yields of live young. It seemed likely that embryonic mortality would prove heavier than in superpregnant adults, and so offset the time gained in using younger females. Our results have shown that this is not the case. Partial loss of pregnancies through the death of individual embryos is in fact somewhat heavier in adult than in adolescent superpregnancies, owing to the greater (unexplained) incidence of early embryonic death (see Fig. 3). The loss of entire pregnancies is commoner in the younger adolescent groups (24 and 26 days old at pairing) than in the adults, but not in the older adolescents (28 and 30 days) (see Table 5). Hence 28−30 days would be the age of choice if we were set the task of extracting the greatest number of live full-term embryos in the shortest possible time out of a given number of TO strain female mice supplied to us, say, at weaning. This choice is re-inforced by considering the proportion of treated females which mate and the proportion of mated females which become pregnant. The mating rate is near to 100% at 28−30 days and falls to below 60% in both the younger and older groups. The pregnancy rate among mated females falls steeply as we reduce the age of pairing below 28 days, but is high in the 28-day, 30-day and adult groups.

A further disadvantage of adult females relative to adolescents is the enormous variation which they show in respect of the number of implants (Table 4). This may reflect heterogeneity either in the number of eggs ovulated, or in the proportion of embryos which implant, or in both. In either case the greater variation in number of implants is undesirable from the practical standpoint.

Even in adolescent females, the death rate at all embryonic stages is very much higher following superovulation than in normal pregnancies. Deficiency of progesterone has been shown experimentally (Smithberg & Runner, 1956; Hall, 1957) to be capable of causing embryonic death at some stages of pregnancy. But in our material progesterone therapy was not effective either against early death or against the catastrophe of placentation which occurs during the middle period when the number of implants in a single uterine horn rises above 8 (Table 8). There is no reason, of course, why progesterone should not alleviate the ill-effects of intra-terine crowding when these are due to sheer compression of the embryo, since the hormone’s effect in relaxing the wall of the gravid uterus is well known. For this reason a progesterone supplement might avert late deaths in superpregnancy. These resemble middle deaths in being mainly dependent upon the degree of crowding of the single uterine horn. But the existence of a sharply demarcated threshold at eight implants per horn, above which the middle death rate increases abruptly, suggests that the limiting factor may be the amount of uterine surface available for the placentation of a linear series of implants, rather than the overall distension of the uterine horn. In the extreme case the interference between neighbouring sites of placentation is such that placental fusion occurs. The occurrence of this phenomenon in artificially crowded uteri has been made the subject of a separate paper (McLaren & Michie, 1959c).

An additional cause of embryonic death during the middle period which remains to be discussed is the double crowding effect displayed by superpregnant adolescents but not by superpregnant adults. Uterine horns in which more than eight implants enter the middle period alive suffer, as we have seen, an increased incidence of deaths in the middle period. But in the treated adolescents this incidence was further greatly increased when the opposite horn also contained more than eight implants surviving into the middle period.

A systemic factor must be sought with effects upon young mice only. An obvious candidate is the nutritional stress involved in supporting the growth of large numbers of embryos while the mother is herself still in the growing period. In support of this idea is the fact that the pregnant adolescent females were invariably found at $1812$-day autopsy to have exhausted their reserves of intra-abdominal fat. This was never observed in adult females. On the other hand it is difficult to see why nutritional stress should make itself felt so early in pregnancy. It is here, perhaps, that we may ascribe a role to progesterone insufficiency, arising from a partial failure of adolescent females to form luteal tissue. Such failure is total in pre-adolescent females after induced ovulation and mating (Smithberg & Runner, 1956).

Returning to our imagined task of obtaining the greatest number of live $1812$-day embryos from a given number of weanling female mice in the shortest time, we can say that the induced superovulation and mating of 30-day-old adolescents will give yields of about thirteen live embryos per treated female within 3 weeks of weaning the females, as compared with about nine in normal pregnancies within 5−6 weeks. It is worth noting that, despite the effects which we have described of intra-horn crowding upon embryonic mortality in the middle and late periods of pregnancy, we have found no signs of a ceiling to the number of embryos which a female mouse can maintain alive to term. This is brought out in Fig. 6, where we have plotted from our data the average number of live embryos against the average number of implants. The slight tendency shown by the adolescents for the curve to flatten out at the upper end of the range is not statistically significant. In an attempt to find a ceiling using another method, namely the transfer of fertilized eggs, we obtained a similar result (McLaren & Michie, 1959b). Robinson (1951), in a study of superovulation in sheep, found that, with increasing numbers of implants, the number of surviving embryos at autopsy reached a ceiling at a level somewhat under four per pregnancy. This is nearly four times the normal litter size of sheep, and so does not necessarily conflict with our findings in mice.

We have as yet said nothing of the greatest hazard of all, namely birth. Our evidence suggests that the greater the number of implants the fewer the number of young born alive. The cause of the heavy parturitional losses is unknown.

But even were these losses overcome the survival chances of new-born mice in the large experimental litters might be poor, since such mice are extremely small, see Healy et al. (1959). In the pig, increased litter sizes have in the past not been wholly desirable, since the small size and heavy mortality of sucklings in the largest litters more than cancels out the initial gain in numbers. But methods of bottle-feeding piglings from birth have recently been developed, so that it might one day become possible to bring a litter of, say, 20 or more live-born piglings from an induced superpregnancy to marketable age without undue losses.

We wish to express our thanks to the Agricultural Research Council for financial support, and to Lord Rothschild, F.R.S., for his personal interest in the work.

We are also indebted to CIBA Laboratories Ltd. for supplies of progesterone and for technical advice.