Fetal and Maternal Transforming Growth Factor-β1 May Combine to Maintain Pregnancy in Mice1

Abstract One of the mysteries of pregnancy is why a mother does not reject her fetuses. Cytokine-modulation of maternal-fetal interactions is likely to be important. However, mice deficient in transforming growth factor-β1 (TGFβ1) and other cytokines are able to breed, bringing this hypothesis into question. The phenotype of TGFβ1 null-mutant mice varies with genetic background. We report here that, in outbred mice, the loss of TGFβ1- deficient embryos is influenced by the parity of their mother. This is consistent with the loss of mutants being due to immune rejection. An inbred line of TGFβ1+/− mice that supported TGFβ1-deficient fetuses had high levels of TGFβ1 in their plasma. Analysis of the amniotic fluids in this line indicated that biologically relevant levels of maternal TGFβ1 were present in the TGFβ1−/− fetuses. These data are consistent with maternal and fetal TGFβ1 interacting to maintain pregnancy, within immune-competent mothers.


INTRODUCTION
We are immunologically distinct from our mothers. Consequently, successful pregnancies involve mechanisms that prevent the mother's immune system from rejecting her concepti. The placenta has a primary role here as it separates the maternal and fetal blood supplies, thus limiting the exposure of the fetus to the mother. The placenta, however, is not a perfect barrier. Fetal cells enter the maternal blood supply [1] and maternal T cells and antibodies can be detected in viable fetuses [2]. Other mechanisms must therefore be acting in concert with the placenta to suppress fetal rejection.
Immune rejection in the adult is regulated by complex interactions involving multiple cytokines. These interactions are also likely to be important in pregnancy, with successful pregnancy being dependent on an appropriate balance of cytokines at the maternal-fetal interface. T-helper (Th) 2/3-type cytokines, such as the transforming growth factor-betas (TGF␤s), interleukin-10 (IL-10), and colonystimulating factor-1 (CSF-1), appear to promote pregnancy whereas proinflammatory Th1 cytokines, such as tumor necrosis factor-␣ and IL-2, are detrimental [3][4][5].
The TGF␤s are a small family of proteins with three mammalian members, all of which are potent regulators of the immune system. In most circumstances, the TGF␤s suppress the activities of immune cells [6]. The three TGF␤ isoforms are present in the uterus [7,8] and concepti [9], but have distinct spatial and temporal patterns. TGF␤2 is produced by the placenta [7,10] and has been implicated in creating immune tolerance of the placenta [5]. The distribution of TGF␤2 within the fetus is, however, comparatively limited [11], making it unlikely that it is primarily responsible for suppressing immune cells that pass through the placenta. TGF␤1 is a better candidate for this role, as it is ubiquitously present in fetal tissues [12,13]. TGF␤1 could also be important for maintenance of the placenta, as decidual cells produce TGF␤1, as well as TGF␤2 [7,8].
The phenotype of TGF␤1 Ϫ/Ϫ mice is complex and varies with genetic background [14,15], but to date, there have been no reports of maternal rejection of TGF␤1-deficient concepti (fetus plus extraembryonic membranes). However, all of the TGF␤1-deficient colonies studied have inbred background that limits the immunological distinction between the mother and her fetuses. We have therefore undertaken a retrospective examination of the breeding records of an inbred and an outbred colony of TGF␤1 ϩ/Ϫ mice, seeking evidence of maternal rejection of fetuses.
As a result of the limited exchange of material between the mother and her conceptus, immune-related conditions such as Rh hemolytic disease are rare during a first pregnancy [16]. This provides a means to assess the physiological role of cytokines in preventing immune-mediated abortion. Simply, if TGF␤1 prevents immune rejection of fetuses in vivo, then the loss of TGF␤1-deficient fetuses should increase with the parity (number of pregnancies) of their mother. We report here that the loss of TGF␤1 Ϫ/Ϫ and TGF␤1 ϩ/Ϫ fetuses in outbred mice follows this pattern.

Animals
All experiments were approved by The University of Otago's Animal Ethics Committee. The majority of the mice were originally bred for experiments that are unrelated to the present study [17,18].
Swiss Webster mice were obtained from the University of Otago's colony.
The inbred TGF␤1 ϩ/Ϫ colony was a derivative of Prof. T. Doetschman's [14] and was established with mice purchased from The Jackson Laboratory (Bar Harbor, ME). These mice had a mixed 129/Sv ϫ C57BL/6J background. The colony was maintained by mating TGF␤1 ϩ/Ϫ studs and dams. The genotypes of the mice were determined by polymerase chain reaction, as previously described [17].
The nude (Whn Ϫ/Ϫ ) TGF␤1 Ϫ/Ϫ colony was maintained by breeding Whn ϩ/Ϫ , TGF␤1 ϩ/Ϫ dams with Whn Ϫ/Ϫ , TGF␤1 ϩ/Ϫ studs and are referred to as outbred/nu. The mothers therefore produced T cells and TGF␤1. The founder nude TGF␤1 ϩ/Ϫ mice were generated by breeding TGF␤1 ϩ/Ϫ dams with nude (Whn Ϫ/Ϫ ) studs. The nude mice had an outbred Swiss Webster background (Crl:CD-1nuBR) and were from a local colony, ex Charles River Laboratories (Wilmington, MA). Nude dams were periodically introduced to the colony to maintain its outbred character.  (4), and 20 (5)(6)(7)(8). The values for all the experienced mothers (parity Ͼ2) were statistically different from the monoparous (parity ϭ 1) value (Student t: #P Ͻ 0.05; *P Ͻ 0.01). The data were also analyzed using binomial logistic regression. For this analysis, the standard error of the means were adjusted for clustering of dams, as each dam has litters in multiple parity classes. The reduction in the proportion of mutant pups in parity 2 was not significant (P ϭ 0.105). For higher parity, the reduction was highly significant (*P ϭ 0.000). There was no correlation between maternal age and number of mutant pups. The presented data are from the outbred/nu colony.
Timed pregnancies were generated by mating TGF␤1 ϩ/Ϫ males and females at 1700 h. The females were examined the following morning at 0700 h for the presence of copulatory plugs. The males were removed to prevent copulation during the daytime. Noon on the day of detection of a plug was defined as E0.5.
An outbred/wt colony was established by crossing inbred TGF␤1 ϩ/Ϫ studs with Swiss Webster dams. The male TGF␤1 ϩ/Ϫ pups resulting from these crosses were then crossed with Swiss Webster dams. The F2 TGF␤1 ϩ/Ϫ dams and studs were mated to analyze the survival of TGF␤1 ϩ/Ϫ fetuses in outbred Swiss Webster mice, lacking the nude mutation.

Collection of Amniotic Fluid
Pregnant dams were anesthetized with pentobarbitone and their abdomens opened. The amniotic fluid of each conceptus was collected by inserting a syringe with a 16-gauge needle through the chorion and amnion. The amniotic fluid was centrifuged at 12 000 ϫ g for 5 min at 4ЊC to remove cells and immediately snap frozen in liquid nitrogen. The samples were stored at Ϫ80ЊC. The tail of each fetus was collected and used for genotyping [17].

Collection of Platelet-Deficient Plasma
Platelet-deficient plasma was prepared using a standard technique. Mice were anesthetized with diethyl ether and 400 l of their blood removed by cardiac puncture using a syringe and 16-gauge needle that had been washed with 2% EDTA. The blood was added to a tube containing 40 l of 2% EDTA at 4ЊC and then spun at 12 000 ϫ g for 15 min at 4ЊC.

ELISA
The concentrations of TGF␤1 in amniotic fluid and plasma were measured using the Promega Emax Immunoassay system (Promega, Madison, WI) according to the manufacturer's instructions. Each sample was acidified with HCl to a pH of 1.6 for 15 min, neutralized with NaOH, and diluted with the manufacturer's sample buffer to ensure that all measurements were made in the middle of the range of the ELISA (31-1000 pg/ ml). The intraassay variation was 7.7%. The ELISA was specific to TGF␤1, with a cross-reactivity of less than 5% with TGF␤2 and TGF␤3 at 10 ng/ml (Promega).

Loss of Mutant TGF␤1 Concepti Is Related to Parity
Retrospective analysis of the breeding records of two colonies of TGF␤1 ϩ/Ϫ mice was undertaken to assess whether the loss of TGF␤1-deficient fetuses was related to parity. The frequency of TGF␤1 Ϫ/Ϫ mice surviving to birth in the outbred/nu colony was strongly linked to parity. The loss of TGF␤1 Ϫ/Ϫ concepti increased substantially after the first pregnancy, with no mutants surviving in mothers carrying their third or subsequent pregnancy (Fig. 1). The survival of TGF␤1 Ϫ/Ϫ concepti was not related to the age of the mother, indicating that the association with parity was not an indirect consequence of mothers being young during their first pregnancy.
We then examined whether experienced outbred/nu mothers rejected a proportion of their wild-type and heterozygous concepti, along with the mutants. The number of wild-type pups per litter averaged 3.3 and was not correlated with parity (not illustrated). The expected ratio of heterozygous to wild-type concepti is two, which was observed in the litter from monoparous mothers. The ratio with experienced mothers (two or more pregnancies) was 1.5, which is significantly less than expected (P Ͻ 0.01, chi-squared). The extent of this loss was maximal with the second litter and relatively constant thereafter (Fig. 2). This suggests that experienced outbred/nu mothers reject approximately one in four of their TGF␤1 ϩ/Ϫ concepti. TGF␤1 Ϫ/Ϫ concepti were also lost in the inbred colony (Table 1). However, in marked contrast with the outbred/ nu colony, the extent of this loss was unrelated to parity, with mothers able to carry a proportion of their TGF␤1 Ϫ/Ϫ embryos to term, even after 11 pregnancies ( Table 1). The   ratio of heterozygous to wild-type pups was also normal in the inbred colony for all parities (Table 1). Inbred TGF␤1 ϩ/Ϫ were then bred to Swiss Webster mice to produce an outbred TGF␤1 ϩ/Ϫ colony that is wild type for the nude mutation (outbred/wt, colony). The number of mutant pups in this colony was also strongly linked to parity (Fig. 3), indicating that the Swiss Webster background is sufficient for the loss of TGF␤1-deficient concepti to occur.

Maternal TGF␤1 Is Present in Concepti
The amount of maternal TGF␤1 in concepti was assessed by comparing the levels of TGF␤1 protein in the amniotic fluids of TGF␤1 ϩ/ϩ , TGF␤1 ϩ/Ϫ , and TGF␤1 Ϫ/Ϫ fetuses within a common TGF␤1 ϩ/Ϫ mother (see Table 2). The concentration of TGF␤1 in the amniotic fluids of the mutant fetuses was approximately half that of their wildtype littermates (Fig. 4B), indicating that significant maternal transfer of TGF␤1 protein had occurred. Heterozygotic concepti had TGF␤1 levels that were intermediate between their mutant and wild-type littermates (Fig. 4B). A gene dose relationship was also observed for the concentration of TGF␤1 in maternal plasma (Fig. 4A).

Plasma TGF␤1 Is Constant During Pregnancy
The concentration of TGF␤1 in the plasma of humans rises during pregnancy [19]. In contrast with this, the level of TGF␤1 in the TGF␤1 ϩ/Ϫ dams used in the present study was not different from that of their nonpregnant counterparts (Fig. 4A). The possibility that maternal TGF␤1 levels fluctuate during pregnancy was then further examined using Swiss Webster mice. No significant variation in plasma TGF␤1 was observed at any stage of pregnancy (Fig. 5).

Parity-Related Loss of TGF␤1-Deficient Concepti
The frequency of TGF␤1 Ϫ/Ϫ and TGF␤1 ϩ/Ϫ mice surviving to birth was strongly linked to parity in the outbred/nu colony. This phenomenon was also observed in the outbred/ wt colony, indicating that the presence of the nude (Whn Ϫ/Ϫ ) mutation is not essential for parity-related loss of concepti.
The loss of TGF␤1-deficient concepti in the outbred col- onies was not associated with maternal age, indicating that the association with parity is not an indirect consequence of more experienced mothers being older. Parity-related reproductive failure is rare and is usually indicative of the involvement of the maternal immune system. With the present study, this is particularly likely, as the fetuses being lost are deficient in a potent suppressor of immune responses.
The loss of TGF␤1 Ϫ/Ϫ concepti in the inbred line was unrelated to parity. Several factors may combine here to create this phenomenon. First, the capacity of the inbred mothers to mount an immune attack against their concepti may be very limited. In inbred lines, the mother and her fetuses are genetically similar, which may attenuate (but not abolish) the ability of the mother to recognize her fetuses as foreign. Additionally, the inbred mice analyzed in this study had C57BL/6 ancestry, which can lead to diminished immune responses in some circumstances [20,21] (see also www.informatics.jax.org/external/festing/search form.cgi).
Second, the inbred mice had higher plasma levels of TGF␤1 than outbred mice (cf., Figs. 4A and 5), which may make the survival of their offspring less dependent on fetal sources of TGF␤1 (see below). Last, successful pregnancy is likely to depend on multiple genes, which may be differentially expressed in the inbred and outbred mice. This could lead to different strains of mice having different dependencies on TGF␤1 (see below).

Loss of TGF␤1-Deficient Concepti in Monoparous Mother
Some TGF␤1 Ϫ/Ϫ concepti die early in development due to inadequate development of the yolk sac [14,22]. The variability of this early death is due to allelic variation in a region of chromosome 5: TGF␤1 Ϫ/Ϫ concepti survive when the chromosome is from NIH/Ola or 129 mice but not when it is from C57BL/6 mice [15,23]. The inbred colony analyzed in this study had a mixed C57BL/6 ϫ 129 background. The loss of TGF␤1 Ϫ/Ϫ concepti that occurred in the inbred colony is thus probably due to the selective elimination of the mice with the unfavorable C57BL/6 allele. The loss of TGF␤1 Ϫ/Ϫ concepti in the monoparous outbred mothers is suggestive that the unfavorable C57BL/ 6 allele is common in outbred colonies. However, we do not discount the possibility that immune rejection may also contribute to the loss of TGF␤1 Ϫ/Ϫ concepti in monoparous outbred litters.

Dual Sources of TGF␤1
The TGF␤1 in amniotic fluid was observed to be of dual maternal and fetal origins. Amniotic fluid is a mixture of fetal urine and lung secretions [24]. The presence of maternal TGF␤1 within the amniotic fluid is thus indicative that maternal TGF␤1 passes through the fetus. Consistent with this, TGF␤1 protein is bound to the connective tissues of TGF␤1 null mutant concepti [18,25] and intact iodinated TGF␤1 is recoverable from fetuses after injection into the mother [25]. Thus, all fetal tissues appear to be bathed in TGF␤1 from both the mother and from sources within the fetus (see [26] for the locations of fetal TGF␤1 mRNA).
In the inbred line, the amniotic fluid from TGF␤1 mutant fetuses had half the level of TGF␤1 protein of their wildtype equivalents (see Fig. 4B and Table 2). This implies that the amniotic fluid from the wild-type concepti contained very similar amounts of maternal and fetal TGF␤1. This method may, however, underestimate the maternal contribution, as the dams used in this study were TGF␤1 ϩ/Ϫ , which have half the wild-type levels of plasma TGF␤1 (Fig.  4A). Thus, in wild-type pregnancies, the amount of maternal TGF␤1 in a conceptus may be up to twice that of fetal TGF␤1. However, we do not favor too exact a ratio being placed on the relative levels of maternal and fetal TGF␤1, as the level of plasma TGF␤1 varied with genetic background (cf. Figs. 4 and 5). As a conceptus is genetically different from its mother, it is likely that the ratio of maternal to fetal TGF␤1 varies in a natural population. The importance of our observations is that they show that a conceptus has biologically significant amounts of TGF␤1 from both its mother and itself.

Are Maternal and Fetal TGF␤1 Redundant?
The existence of two sources of TGF␤1 raises the issue of whether both are essential for the maintenance of pregnancy. When considering this question, it is important to bear in mind the following. TGF␤-dependent phenomena exhibit different dose sensitivities. Mice that are heterozygous null for one of the TGF␤s exhibit specific defects, even though they contain 50% of the normal levels of the TGF␤ [27,28]. Other defects are only observed when both alleles are mutated [14,29] (Table 1). Here, 50% of the normal levels of TGF␤ are sufficient for full function. Further defects emerge when deficiencies in TGF␤ signaling occur in concert with either deficiencies in other signaling pathways [30,31] or allelic variations in genes with no obvious relationship to signaling pathways [32]. Consequently, we need to consider the possibilities that 1) maternal and fetal TGF␤1 may jointly regulate some aspects of pregnancy/development while independently controlling other aspects and 2) the importance of TGF␤1 in controlling immune rejection may be modulated by multiple factors, including the genetic backgrounds of both the mother and her concepti, the immunological similarity between the mother and her concepti, and the immune history of the mother.
With inbred mice, the mothers have high levels of plasma TGF␤1 and no parity-related loss of TGF␤1-deficient concepti occurs (this study). This indicates that fetal TGF␤1 is not essential to prevent parity-related loss of concepti in this colony. This mirrors the situation with inbred neonates where maternal transfer of TGF␤1 via milk is suf-ficient to suppress autoimmune attack in TGF␤1 Ϫ/Ϫ pups [14,33]. Both of these phenomena are, however, in marked contrast with the TGF␤1-dependent development of the yolk sac, which appears to be entirely under the control of fetal TGF␤1 [32,34]. This is not surprising, as the defect in the yolk sac is lethal before extensive development of the placenta has occurred. The ability of maternal TGF␤1 to reach the concepti would thus be limited.
The situation with the inbred mice also contrasts with that of the multiparous outbred mothers where fetal TGF␤1 production is an important determinant of survival: wildtype concepti survive; TGF␤1 ϩ/Ϫ concepti have reduced survival; TGF␤1 Ϫ/Ϫ concepti do not survive, even though their amniotic fluids contain approximately 50% of TGF␤1 of the wild-type littermates. This suggests that the TGF␤1dependent suppression of parity-related loss of concepti only occurs if the level of TGF␤1 exceeds a threshold, as in the deficits observed in TGF␤1 ϩ/Ϫ adults [27].
In summary, the data reported here show that concepti are exposed to significant levels of both fetal and maternal TGF␤1. The levels of fetal TGF␤1 appear to be a determinant of whether a conceptus is rejected or not, although the evidence clearly indicates that maternal TGF␤1 and other factors are also important. In at least some circumstances, these factors may be sufficient to fully compensate for a low level of fetal TGF␤1.