Main

Understanding the mechanisms coordinating the rate of fetal maturation with the length of gestation and the initiation of labor are of fundamental biologic interest and substantial medical importance. Preterm labor continues to be a frequent complication of pregnancy, with sequelae for premature infants not yet prepared for viability ex utero including hyaline membrane disease, intraventricular hemorrhage, necrotizing enterocolitis, growth failure, and sepsis (1, 2). The support of these premature newborns exacts significant emotional and financial costs, not only in terms of the acute complications and intensive care management they require, but also because of the development of chronic sequelae such as bronchopulmonary dysplasia and cerebral palsy. For these reasons, intervention intended to prevent or significantly delay premature labor would be of great utility.

Characterization of the initiators of human labor has proven difficult. In part, this difficulty stems from the limited ability to extrapolate mechanisms from nonprimate systems to humans. For example, in the sheep, activation of the fetal hypothalamic-pituitary-adrenal axis, and the accompanying GC surge, clearly precipitates labor (3). In this species, rising serum cortisol concentration induces 17-hydroxylase/17, 20 lyase activity, which reduces progesterone production, resulting in increased PG synthesis and labor (4). In women, this mechanism is not maintained (5). GC administration does not precipitate labor, although recent studies suggest that placental CRH, which could augment GC production, may contribute to the timing of parturition (6–8). Administration of exogenous GC to fetal sheep also results in acceleration of lung maturation, suggesting that the endogenous GC rise serves to promote both fetal maturation and the onset of labor (9, 10).

The application of molecular genetic techniques that allow mutational analysis in vivo shows promise for elucidating relevant molecular pathways that control parturition. To date, transgenic and KO mice are the most readily accessible mammalian systems for such studies (11–13). The maintenance of pregnancy in the mouse differs from that in humans in that progesterone and estradiol are synthesized in the corpus luteum of the ovary throughout gestation, whereas luteal function in the human is necessary to maintain pregnancy for only the first trimester (14, 15). Additionally, the termination of pregnancy in the mouse is precipitated by a decline in serum progesterone, whereas the termination of pregnancy in humans is not. Instead, local modulation of progesterone synthesis or action within the uterus and fetal membranes has been suggested as a mechanism by which diminished progesterone action could still result in labor in humans (16–18). Nevertheless, conserved regulation and critical functions for several of the same molecular pathways in mouse and human labor have emerged, as would be expected for a process as fundamental as parturition. Targeted inactivation studies using homologous recombination in ES cells (13) to generate mice completely deficient in specific neuropeptides and interacting molecules have recently implicated the importance of these systems in the maintenance of gestation and the control of parturition in humans. The results of three such systems will be reviewed.

CORTICOTROPIN-RELEASING HORMONE

Our initial investigations of neuropeptides in the modulation of the timing of parturition and regulation of fetal maturation centered on defining the contribution of CRH to these processes. CRH is a 41-amino acid peptide initially isolated by Vale et al. (19, 20) on the basis of its ability to augment ACTH release from the pituitary and modulate the function of the hypothalamic-pituitary-adrenal axis. In addition to CRH localized to paraventricular nucleus of the hypothalamus and destined for release into the hypophyseal portal blood supply, this peptide has been found in other brain regions, as well as nonneural sites (21–23). Importantly, CRH in human serum has been found to increase dramatically during pregnancy because of high-level expression by the placenta (6, 8, 24, 25). Moreover, CRH-binding protein, which limits the bioactivity of circulating CRH, decreases in abundance late in gestation (26, 27). The timing of the placental increase in CRH synthesis, together with increasing CRH bioactivity, has been implicated as an important contributor to the timing of labor in humans, although the function of CRH during parturition remains uncertain (6, 8).

To directly examine the role of CRH during murine pregnancy, and GC whose synthesis would be altered as a consequence of CRH deficiency, we generated mice homozygous for CRH deficiency by homologous recombination in ES cells (23). Mice homozygous for CRH deficiency derived from heterozygote matings were present in the expected Mendelian ratios and appeared healthy (28). Despite this normal appearance, CRH KO mice exhibited adrenal atrophy and deficient GC production (Fig. 1). Female CRH KO mice generated serum GC (corticosterone in mice) concentrations approximately 30% of WT concentrations after stress, and the male CRH KO mice exhibited even more profound adrenal insufficiency, displaying no significant increase in their serum corticosterone above basal concentrations after a variety of stresses including restraint, fasting, and ether (28).

Figure 1
figure 1

Adrenal hypoplasia in CRH KO mice. Hematoxylin-eosin stained sections of adrenals from 8-wk-old CRH KO (a) or WT (b) mice are shown. The normal strandlike appearance of the adrenal cortex, predominantly the GC-producing zona fasciculata (ZF), is seen in the WT adrenal, whereas the CRH KO has very little ZF between the adrenal medulla (M) and zona glomerulosa (ZG), located beneath the adrenal capsule. Reproduced with permission from Muglia et al. (28).

Full size image

CRH KO male and female mice proved fertile. Despite the evidence of placental CRH serving an important function in human pregnancy (6) and fetal hypothalamic CRH serving an important function in ovine pregnancy (29), the completely CRH-deficient gestations resulted in a normal number of pups delivered at the expected time. The normal timing for labor in the CRH KO pregnancies does not exclude a role for CRH in human pregnancy, because mouse placenta does not synthesize appreciable amounts of CRH (23). In striking contrast to CRH KO progeny of heterozygote matings, however, all pups arising from KO × KO matings died on the first day of life (28). Inasmuch as CRH KO progeny of heterozygote matings are viable, the heterozygous mother must provide a factor that crosses the placenta and rescues the homozygous KO fetus. Maternofetal transfer across the placenta of CRH, a peptide hormone, in amounts sufficient to rescue the KO progeny of heterozygote matings, would be unlikely. Therefore, we tested the hypothesis that GC, present in normal concentrations in the heterozygous mother and deficient in the CRH KO dams, crossed the placenta and restored viability to pups of KO × KO matings. Indeed, administration of corticosterone in the drinking water to gravid CRH KO females resulted in birth of viable progeny (28).

To determine the basis of CRH and resultant glucocorticoid deficiency leading to fetal demise, we performed histologic analysis on the progeny of timed KO × KO and WT matings. These studies revealed a failure of lung morphologic maturation after day 16.5 of mouse gestation, corresponding to the progression from the early to the late canalicular phase of pulmonary development (Fig. 2) (28, 30). The CRH KO lung was hypercellular, as reflected by increased wet weight, dry weight, and DNA content in comparison to WT lung at 18.5 d gestation. This hypercellularity in the CRH KO lung did not result from impaired apoptosis, but rather increased cell proliferation as determined by sustained proliferating cell nuclear antigen immunoreactivity late in gestation (30).

Figure 2
figure 2

Lung morphogenesis in CRH KO and WT mice. Hematoxylin-eosin stained sections of lung from fetuses of 14.5 through 18.5 d gestation photographed at ×200 magnification are shown. (a) WT 14.5 d; (b) CRH KO 14.5 d; (c) WT 16.5 d; (d ) CRH KO 16.5 d; (e) WT 17.5 d; (f) CRH KO 17.5 d; (g) WT 18.5 d; (h) CRH KO 18.5 d. A divergence in architecture is apparent at day 17.5, with the CRH KO maintaining a dense, pseudoglandular appearance. Reproduced with permission from Muglia et al. (30).

Full size image

Despite earlier studies on pulmonary development that implicated endogenous GC as a modulator of surfactant apoprotein gene expression, our findings in the CRH KO mice demonstrated no substantial differences in either surfactant apoprotein or lipid biosynthesis (30). These findings, which are consistent with the normal levels of the surfactant apoprotein mRNAs in the GC receptor KO mice at birth (31), argue against an absolute requirement of GC for surfactant apoprotein gene expression and suggest that the essential effects of GC on fetal lung maturation are on cell proliferation and architectural maturation. This concept is further supported by the analysis of markers of epithelial differentiation in the CRH KO mice, which again reveal that the fatality observed in CRH-deficient mice results directly from an overall delay in the timing of pulmonary maturation secondary to GC insufficiency (30). Taken together, the striking discordance between the rate of fetal maturation and the onset of labor in the CRH-deficient mice underscores the essential role of GC in the timing of fetal development such that viability ex utero is achieved.

OXYTOCIN

OT, a cyclic nonapeptide produced primarily in the hypothalamic paraventricular and supraoptic nuclei, augments uterine contractions and endometrial and amnion PG production when given to humans in either pharmacologic or physiologic doses (32). Additionally, the abundance of uterine myometrial OT receptors increases more than 10-fold as pregnancy nears term (33). Thus, OT has been considered an important facilitator of labor at term. Other studies, however, have shown no increase in plasma OT concentrations before the onset of labor (34, 35). Moreover, OT antagonists fail to alter the onset of labor (36, 37), and humans with damage to the posterior pituitary, the site of release of OT synthesized in the hypothalamus, have been observed to progress through labor normally. Confusing the determination for or against a necessary role for OT in parturition is its paracrine production and action within the uterus and decidua, a setting in which blockade by an antagonist could be incomplete (33).

Using a strategy similar to that in creation of CRH KO mice, we (38) and other groups (39, 40) generated mice with OT deficiency by homologous recombination in ES cells. Most surprisingly, gravid OT KO females, including those that had been mated with OT KO males, initiated parturition at the expected time of 19.5 d of gestation and did not demonstrate prolongation of active labor. Therefore, neither maternal nor fetal OT is required for normal labor to occur in the mouse. Progeny arising from OT KO females die in the first days of life because the OT KO females do not lactate normally (39, 40). As expected from the myoepithelial localization of OT receptors, this lactation defect results from defective milk letdown rather than production.

CYCLOOXYGENASE-1

One pathway that, like OT, can pharmacologically accelerate labor is that of PG synthesis. PG induce luteolysis, augment uterine contractions, and promote the controlled inflammatory response that results in dilatation and thinning of the cervix and disruption of the connective tissue of the decidua (41). Indeed, the administration of nonsteroidal anti-inflammatory drugs, which block PG production, have proven efficacious in attenuating the progression of term and preterm labor in animal model systems (42, 43), as well as in human studies (44, 45). Not only do amniotic fluid concentrations of PG and other cytokines rise as labor proceeds (46, 47), but it has recently been demonstrated that the increase in amniotic fluid PG occurs before the onset of labor in humans (48).

All PG synthesis starts with the conversion of arachidonic acid to PGH2. This production is catalyzed by COX (PGH synthase), an enzyme with two isoforms, designated COX-1 and COX-2. These COX isoforms share similar structures and are the primary sites of action of the nonsteroidal anti-inflammatory drugs (49, 50), but differ significantly in their pattern of regulation. COX-1 is constitutively expressed in most tissue types. Although stimulation of expression can occur in certain cell types after growth factor exposure, little regulation has in general been demonstrated, including in extrafetal membranes during parturition (51–53). COX-2, in contrast, is in general undetectable in most tissues but can be expressed at high concentrations in macrophages and other inflammatory cell types after cytokine exposure (51, 54). COX-2 is expressed in the amnion, and rises in cultured amnion cells in response to IL-1β administration (55).

Molecular genetic studies on PG synthetic enzymes and PG receptors have confirmed the central role of PG in murine parturition. PGF2α receptor- (56) and cytoplasmic phospholipase A2- (57) deficient mice fail to initiate labor because of impaired luteolysis and persistent progesterone production. Because of this failure of luteolysis, myometrial OT receptors are not induced (56). In addition to inhibiting induction of OT receptor expression in the uterus, the sustained production of progesterone by the corpus luteum would be anticipated to directly impair OT receptor activity (58). If OT receptors are induced in these mice by ovariectomy, however, labor is initiated and progresses normally (56). Thus, PGF2α appears to be required for luteolysis in mice, but is not essential for uterine contractions if a fall in progesterone occurs by other means. Although PGE2 action has been implicated in closure of the ductus arteriosus in mice deficient in one of the PGE2 receptors, EP4 (59), and fertilization of the ovum in EP2 receptor-deficient mice (60, 61), parturition defects have not been described in any mouse mutants deficient in the four known PGE2 receptors (62).

To determine which of the COX isoforms generates the PGF2α necessary for the onset and progression of labor at term in mice, we evaluated the parturition phenotype in COX-1 KO mice. The initial characterization of COX-1 KO mice indicated that both males and females were healthy and resisted arachidonic acid-induced inflammation (63). Pregnancies arising from COX-1 KO male × COX-1 KO female matings, however, culminated in delivery of nonviable fetuses (63). To determine whether fetal demise at term in the COX-1 KO mice was caused by an abnormality in the onset or progression of labor, we established timed pregnancies of COX-1 KO females mated with COX-1 KO or COX-1-intact males (38). In marked contrast to the onset of labor at day 19.5 of gestation in WT mice, COX-1 KO females delivered their litters at day 21.5–22.0. Administration of PGF2α to the gravid COX-1 KO females at day 18.5–19.5 resulted in delivery of viable pups, demonstrating that fetal demise was caused by postdates gestation and not by a developmental abnormality arising from COX-1 deficiency. Blastocyst transfer of COX-1 KO or WT embryos to pseudopregnant WT females further demonstrated that fetal COX-1 was not required for the normal onset of labor (38). Thus, these studies revealed that maternal COX-1 is necessary and sufficient for normal murine labor.

Biochemical analyses of the mechanism of action of COX-1 showed that COX-1 was induced in the uterine decidua at high concentrations in late mouse gestation and accounted for the vast majority of PGF2α production (38) (Fig. 3). The induction of COX-1 mRNA was surprising given the usual housekeeping activity this isoform is thought to mediate. The insufficient PGF2α production in the COX-1 KO mice inefficiently initiated luteolysis and induction of OT receptors, consistent with studies in the PGF2α receptor and cytoplasmic phospholipase A2 KO mice (56, 57). The poor ovulatory frequency and impaired blastocyst implantation found in COX-2 KO mice (64) has thus far precluded genetic determination of the consequences of complete COX-2 deficiency for the onset and progression of term labor in mice.

Figure 3
figure 3

COX-1 mRNA induction and uterine PGF2α during pregnancy in mice. (a) Northern blot analysis of total uterine RNA. Samples from nongravid (NG) or pregnant females of the gestational age in days shown above the corresponding lanes were hybridized to COX-1 or cyclophilin (CYC) probes. (b) Impaired uterine PGF2α production in COX-1 KO and COX-1 KO/OT KO mice. PG were extracted from uteri of the indicated genotypes, and PGF2α determined by ELISA. Assay specificity was confirmed by administration of indomethacin (Indo) to a subset of WT mice at day 15.5 of gestation to suppress PG production. Data are presented as mean, with day 19 samples ± SEM. *p < 0.05 vs WT at day 19. Reproduced with permission from Gross et al. (38).

Full size image

Genetic analyses in mice facilitate the ability to dissect redundant in vivo pathways by generation of animals with combined deficiencies. To determine whether redundant OT and PG action on the myometrium allowed both the normal time of onset and normal duration of labor in OT KO mice, and the normal duration of labor despite the delayed onset of labor in COX-1 KO mice, we generated mice deficient in both COX-1 and OT (COX-1 KO/OT KO) (38). We hypothesized that with deficiency in both PG and OT activity, not only would there be a delay in the initiation of labor, but also a prolongation in the duration of labor. In part, this expectation was fulfilled with the COX-1 KO/OT KO females demonstrating a markedly prolonged duration of labor with delivery of pups occurring during a 2-d period. Surprisingly, OT deficiency restored the onset of labor to normal (day 19.8). COX-1 KO/OT KO mice underwent luteolysis in a manner similar to WT mice, although they did not require the usual increase in PG found in WT mice before the onset of parturition (38) (Fig. 4). These findings suggest that OT not only facilitates uterine contractions, but also serves to maintain corpus luteum function in late gestation. Whether OT action occurs directly on the corpus luteum, or results from modulation of other peptides, such as prolactin, known to affect luteal function, is currently under investigation.

Figure 4
figure 4

Corpus luteum histology at day 19 of gestation. Corpora lutea of WT and COX-1 KO/OT KO mice demonstrate an amorphous cellular appearance and disruption of luteal architecture consistent with luteolysis. In contrast, the COX-1 KO corpus luteum, which has not undergone luteolysis, maintains prominent vascular spaces, a uniform cellular appearance, and an organized luteal architecture (×200 magnification). Reproduced with permission from Gross et al. (38).

Full size image

SUMMARY AND FUTURE DIRECTIONS

The molecular genetic analysis of the control of parturition in mice with targeted gene mutations has allowed dissection of the mechanism of labor initiation in this species (Fig. 5). Arachidonic acid release from cellular lipid stores by the action of cytoplasmic phospholipase A2 provides substrate to COX-1 for generation of PGF2α. Induction of COX-1 in the uterus during gestation then results in a dramatic increase in PGF2α late in gestation. The increased PGF2α, via action on PGF2α receptors on the corpus luteum, causes luteolysis and a decrease in serum progesterone. The fall in progesterone induces a state of increased uterine contractility, by diminished direct antagonism of OT receptor signaling, and induction of OT receptors, contractile PG receptors, gap junctions, and other myometrial proteins that increase sensitivity to uterotonic agents. OT and PG actions at the level of the myometrium appear to be at least in part redundant inasmuch as only in mice with combined OT and COX-1 deficiency is the duration of labor prolonged (38). The induction of COX-1 at least 3 d before the onset of labor (38) indicates that the process of parturition is set in motion well before the appearance of the known indicators of luteolysis and labor. Further analysis of COX-1 regulation should prove valuable in determining which maternal or fetal signals are involved in the activation of labor at even earlier times during gestation and thus provide new insight into the clock mechanism of parturition.

Figure 5
figure 5

Scheme for parturition in mice. Arachidonic acid released from membrane phospholipids by the action of cytoplasmic phospholipase A2 (cPLA2) serves as substrate for COX-1 to generate PGH2. PGH2 is converted by the action of PGF synthase to PGF2α. Increased uterine (and perhaps ovarian) generation of PGF2α late in gestation causes involution of the ovarian corpus luteum by action on the PGF2α receptor. The resulting fall in progesterone increases uterine contractility and promotes labor by decreased direct antagonism of the OT receptor, and induction of OT receptors, contractile PG receptors, and gap junctions, among other myometrial proteins.

Full size image

What do these experiments teach us about the control of human parturition? Extrapolation of the mechanism by which PG promote parturition in mice to humans is not straightforward (14, 15), although PG are potent modulators of labor in both species. The corpus luteum in humans does not actively contribute to progesterone production in late gestation, and progesterone does not fall before the onset of labor. Is there a functional equivalent in the human to the rodent corpus luteum, e.g. the placenta or fetal adrenal? Our studies suggest that the analysis of the consequences of PGF2α action on these structures will be useful in this regard. Several studies have investigated which of the COX isoforms is essential for the rise in PG in late human gestation, with the only clear increase occurring in COX-2 from extraembryonic tissues (53, 65–67). More important, determination of the isoform critical for imparting preterm labor in humans remains controversial (67, 68). Thus far, no naturally occurring or targeted mutations imparting preterm labor in the mouse have been described. As novel candidate genes implicated in the pathogenesis of human preterm labor are identified, in vivo genetic analysis in the mouse will greatly accelerate determining their mechanism of action. Inflammatory models of preterm labor in mice, such as lipopolysaccharide administration (43, 69), have been used and may provide a reasonable analog of some forms of preterm labor in humans. Determination of whether COX-1 or COX-2 plays the essential role in inflammation-mediated preterm labor in mice, and ultimately in humans, may allow efficacious therapy by selectively modulating the activity of only one COX isoform. By using appropriate isoform-specific inhibition, preterm labor could be stopped without imparting the many sequelae of nonselective COX blockade if activity of the targeted isoform was not also essential for other aspects of fetal and maternal organ function.