In addition, pregnancies resulting from ART show an increased incidence of anomalous placentation, including abnormal shape, abnormal umbilical cord insertion and presence of vasa previa (Gavriil et al., 1993; Al-Khaduri et al., 2007). A 3–5-fold increased risk of placenta previa has been associated with ART (Jackson et al., 2004; Romundstad et al., 2006). Studies in cattle show abnormal gene expression and a reduction in placentome number following somatic cell nuclear transfer (Arnold et al., 2006; Everts et al., 2008). The number of trophectodermal cells is reduced in IVF blastocysts; in addition genes involved in placental transport function are misregulated after IVF (Giritharan et al., 2007).

A number of important conclusions can be derived from these experiments. Although embryos undergoing only the embryo transfer procedure (FB group) adapt their development to the surrogate GA, the fetuses generated in vitro and transferred to recipient mothers (IVF group) show a more noticeable phenotype: their observed development is lagging behind their expected development (by 0.38 days); it is interesting to note that IVF conceptuses maintained the delay in development that was accrued after culture at the blastocyst stage (and showed neither a ‘catch up’ growth nor a magnified delay in development). IVF fetuses and placentas are smaller than FB fetuses even when controlling for DS (Fig. 1 and Table II). An improved media (KAA) partially compensates for the effect of culture; however, even optimized in vitro culture conditions affect fetal development (Fig. 1).

The placental/fetus ratio progressively and significantly increased as conditions deviated further from the in vivo situation. The placenta/fetus ratio is considered a marker of intrauterine stress (Barker et al., 1990). Whereas both fetal and placental weights decrease as a result of manipulation in vitro, there is a greater decrease in the size of the embryo than in placenta size (Fig. 1).

The reasons for the increase in placental/fetus ratio could be diverse and a combination of different factors may be involved. It is important to note that an increased placenta/fetus ratio has been described in human IVF pregnancies and has been associated with hypertension later in life (Barker et al., 1990; Moore et al., 1996; Koudstaal et al., 2000).

These effects cannot be attributed to differences in the number of embryos transferred, or to differences in implantation rate or litter size. In fact, a similar number of embryos were transferred in FB and IVF groups and a similar number of embryos implanted (Table I), allowing a consistent comparison between groups. However, there is a higher incidence of abortive sites in IVF mice when compared with FB animals. Interestingly there was a higher twinning rate in IVF_WM fetuses: although the increase was not significant, it is worth noting, since twinning rate is very rare in naturally mated mice. It remains to be evaluated if this effect is induced by ovulation induction or embryo culture.

Because of the experimental design, we do not know if placental size and ratio will be maintained at birth. The altered placenta/fetus ratio was not associated with an abnormal placental morphology. IVF_WM, in vivo and FB placentas displayed a similar labyrinth and spongiotrophoblast development. The total placenta and labyrinth areas were larger in the FB group compared with in vivo and IVF_WM group but the ratio labyrinth/total area was similar, suggesting normal function. It has been noted that changes in the labyrinth/total area ratio are associated with impaired placenta function (Rossant and Cross, 2001). Because histological results were similar in IVF_WM and FB groups, we did not repeat the experiments with IVF_KAA group. Placental fetal blood vessels, maternal blood spaces and trophoblast compartments were similar in FB and IVF_WM. Only Ki67 staining was increased in 12.5 day GA IVF placentas (P < 0.05). We have evidence (data not shown) that placenta morphology and proliferation is not different at GA Day 8.5 in FB and IVF_WM groups. It is possible that an increase in mitosis at GA 12.5 days represents a compensatory mechanism to rescue an originally restricted placental tissue mass. However, because Ki67 is a marker both of proliferation and endoreduplication (typical of trophoblast giant cells) our findings do not necessarily indicate an increase in proliferation and placental growth.

Placental size can be affected by gamete manipulation or embryo culture. In cattle, large offspring syndrome is associated with normal placental size (Sinclair et al., 1999) and interestingly, the gene expression profile of the endometrium of IVF and in vivo generated cows is different, suggesting that the endometrium functions as a biological sensor of in vitro manipulation technologies (Mansouri-Attia et al., 2009).

It has yet to be elucidated whether IVF alters placental function. Collier reported dysregulation of placental steroid metabolism in mouse pregnancies conceived through ART (Collier et al., 2009). Proteomic analysis reveals the presence of an abnormal protein profile after ART (Zhang et al., 2008). Our results support epidemiologic data, indicating that IVF term singletons weigh less than naturally conceived children (McDonald et al., 2009). These findings also validate blastocyst gene expression data, which show that in vitro embryos have an altered gene expression pattern with increase in apoptotic and ‘stress’ pathways (Rinaudo and Schultz, 2004; Rinaudo et al., 2006; Giritharan et al., 2007). Our laboratory has shown that IVF blastocysts contain fewer trophectodermal cells (Giritharan et al., 2007) which may lead to a smaller placenta, resulting in an even smaller fetus.

The mechanism by which IVF affects placental or fetal size is unknown. One possibility is that epigenetic differences are responsible for these findings. Indeed Rivera et al. (2008) show that the process of embryo transfer alone is sufficient to alter the methylation pattern of several imprinted genes. Perturbations in embryonic metabolism and gene expression may be additional mechanisms through which in vitro culture influences embryonic viability and subsequent development.

It remains to be determined whether these perturbations of fetal growth and development will be associated with longer-term health consequences.

These findings need further investigation since they suggest a possible suboptimal pattern of placentation in IVF that differs from that of an in vivo-conceived control group. If a similar impairment occurs in humans, it may explain the obstetric and perinatal complications associated with ART.

Several considerations need to be kept in mind when analyzing the current study. First, differences between human and mouse placentation need to be taken into account. Mouse and human placenta differ in their morphogenesis and exchange functions (Rossant and Cross, 2001; Malassine et al., 2003). On the other hand, placentation is hemochorial in both species and similarities have been identified among placental cell populations (Malassine et al., 2003; Wang and Dey, 2006).

Second, the decision to sacrifice mice at 12.5 days GA was based on developmental considerations. Between 8 and 10 days post-fertilization the placenta has completed the fusion of chorion and allantois and can be defined as mature (Rossant and Cross, 2001). From this stage onwards the placental architecture is well established and spongiotrophoblast and labyrinthine zones can be distinguished clearly. Although additional time points could have been selected to observe placenta and embryo development, analysis of placenta on Day 12.5 offers sufficient sensitivity to describe a delay in placenta maturation and morphology.

Third, we used morphologic characteristics to define DS, following the protocol of Gruneberg (1943) which allows the dating of mouse fetuses from 7 to 18 days with an error of <24 h, based on the external features of the fetus. If fetuses had developmental characteristics that were between 10 and 11 days of development they were given an intermediate score (a value of 10.5 days). Alternative methods, such as somite counting or crown rump length (CRL) measurements, have an even higher margin of error. In fact, by embryo Day 10 most of the rostral somites already begin to de-epithelialize into sclerotomal cells and become difficult to count (Kei Chan et al., 2005). CRL, on the other hand, is an unreliable estimator of embryonic age, since recoil and stretching of the embryo itself can fix it in abnormal postures. It is possible that the Gruneberg method we utilized also lacks precision. However, two independent observers, blinded on the study group, analyzed all the data and reached the same conclusions. Furthermore, the in vivo group showed a close correspondence between the observed and expected development.