Novel Toll-like receptor-4 antagonist (+)-naloxone protects mice from inflammation-induced preterm birth

Toll-like receptor 4 (TLR4) activation by bacterial infection, or by sterile inflammatory insult is a primary trigger of spontaneous preterm birth. Here we utilize mouse models to investigate the efficacy of a novel small molecule TLR4 antagonist, (+)-naloxone, the non-opioid isomer of the opioid receptor antagonist (−)-naloxone, in infection-associated preterm birth. Treatment with (+)-naloxone prevented preterm delivery and alleviated fetal demise in utero elicited by i.p. LPS administration in late gestation. A similar effect with protection from preterm birth and perinatal death, and partial correction of reduced birth weight and postnatal mortality, was conferred by (+)-naloxone administration after intrauterine administration of heat-killed E. coli. Local induction by E. coli of inflammatory cytokine genes Il1b, Il6, Tnf and Il10 in fetal membranes was suppressed by (+)-naloxone, and cytokine expression in the placenta, and uterine myometrium and decidua, was also attenuated. These data demonstrate that inhibition of TLR4 signaling with the novel TLR4 antagonist (+)-naloxone can suppress the inflammatory cascade of preterm parturition, to prevent preterm birth and perinatal death. Further studies are warranted to investigate the utility of small molecule inhibition of TLR-driven inflammation as a component of strategies for fetal protection and delaying preterm birth in the clinical setting.

Precocious inflammation leading to preterm labor can be triggered by infection or by sterile pro-inflammatory insults, originating either within gestational tissues or systemically, from tissues distal to the placenta and fetus 1,16 . Detection of infection is mediated by sensing molecules known as Toll-like receptors (TLRs) that bind conserved molecular structures carried by microorganisms, known as pathogen-associated molecular patterns (PAMPs). TLRs also sense silent inflammation resulting from non-infectious insult, through ligation of endogenous agents released by injured tissue (damage-associated molecular patterns, DAMPs). The archetypal receptor in this detection system is TLR4, which is activated by the bacterial cell wall component lipopolysaccharide (LPS). TLR4 is extensively expressed throughout the female reproductive tract including the fetal membranes 17,18 and placental trophoblasts 19,20 , as well as leukocytes, epithelial and mesenchymal cells in the cervix 20 and uterus 21 . TLR-driven pathways are involved in normal on-time parturition, since mice with null mutation in Tlr4 or Tlr2 exhibit a delay in the timing of term labour 22,23 .
Infection-induced preterm labor can be modelled in mice by administration of LPS systemically or locally into the gestational tissues [24][25][26] , or delivery of killed E. coli into the uterine cavity or fetal membranes to mimic ascending infection and chorioamnionitis 27,28 . TLR4 is a key sensor for infection-induced preterm birth, since protection is afforded in mice by spontaneous mutation of Tlr4 in C3H/HeJ mice 27,29,30 , null mutation of Tlr4 in C57Bl/6 mice 22,30 , and genetic deficiency in the MyD88 and TRIF adaptor proteins required for TLR4 signaling 28 . TLR4 inhibitors including a lipid A mimetic and anti-TLR4 neutralising antibodies have previously shown promise in attenuating preterm birth in mouse infection models 24,30 , but neither of these are suitable for human clinical application.
These observations underpin our hypothesis that TLR4 inhibition with a novel TLR4 antagonist (+)-naloxone might comprise a useful strategy for targeted suppression of inflammation underlying preterm birth. (+)-Naloxone is a newly identified TLR4 antagonist that is the non-opioid isomer of the opioid receptor antagonist (−)-naloxone 31 , a well-described nonselective antagonist of the μ -opioid receptor commonly prescribed for opioid addiction, including in pregnant women and neonates 32 . (+ )-Naloxone binds MD2 to prevent LPS engagement with TLR4 33 leading to diminished immune NFκ B activation and IL1B, IL6 and TNF synthesis [34][35][36] . In contrast to anti-TLR4 neutralizing antibodies 24 , (+ )-naloxone is a small molecule which can penetrate across the placenta 37 , with a pharmacokinetic profile affording it short systemic exposure or longer term delivery if required. (−)-Naloxone has anti-inflammatory activity and provides protection against sepsis in animal models 38,39 . (+)-Naloxone has similar properties but unlike (−)-naloxone, does not have opioid receptor-binding activity, and is a specific antagonist of TLR4 signaling 31 .
In view of the attractive pharmacological properties of (+ )-naloxone as a TLR4 antagonist and our previous observation that (+ )-naloxone can delay the timing of natural on-time parturition 22 , we sought to investigate the possible utility of (+ )-naloxone as a novel anti-inflammatory tocolytic agent. Here, we use two common mouse models of infection-induced preterm birth to show (+ )-naloxone protects against preterm delivery and alleviates in utero and postnatal loss induced by TLR-driven inflammation.
Treatment with (+ )-naloxone protected pregnancy from LPS challenge, completely abrogating PTD (0/14 mice delivered before gd 18.0) (Fig. 1A,B). The low survival rate of fetuses in utero after LPS challenge was also reversed by (+ )-naloxone, such that the mean number of viable fetuses per dam was not less than the control PBS-treated group (Fig. 1C-E). Moreover, administration of (+ )-naloxone alone in the absence of LPS did not induce PTD (0/14 mice) or affect fetal survival rates (Fig. 1A-E).
When fetuses surviving in utero at gd 18.5 were analysed for effects of treatments on fetal and placental weight, mixed model ANOVA showed that LPS reduced fetal weight (p = 0.033) ( Fig. 2A) and elevated placental weight (p = 0.019) (Fig. 2B), causing a substantial reduction in fetal:placental weight ratio (p = 0.001) (Fig. 2C), a parameter indicative of placental functional efficiency. Conversely (+ )-naloxone exerted a positive impact on fetal weight (p = 0.045), and did not affect placental weight or fetal:placental weight ratio. Post-hoc analysis to compare individual treatment groups demonstrated a 15.2% reduction in fetal:placental weight ratio after LPS treatment compared to PBS control (p = 0.036) (Fig. 2C), but (+ )-naloxone treatment was not able to correct this.
These data indicate that (+ )-naloxone can effectively inhibit TLR4 signaling to prevent PTD and rescue fetal loss elicited by LPS administration in late gestation. However, fetuses that survive LPS challenge in utero exhibited altered placental efficiency indicated by reduced fetal:placental weight ratio, regardless of (+ )-naloxone treatment.
(+)-Naloxone prevents heat-killed E. coli-induced preterm delivery. Given the ability of (+ )-naloxone to prevent LPS-induced PTD, we next investigated whether (+ )-naloxone can also prevent PTD induced by intact bacteria, which elicit PTD by ligation of multiple TLRs including TLR4 and TLR2 42,43 . In this experiment, we used a proven model of chorioamnionitis-induced PTD wherein heat-killed E. coli are surgically t-test, with mother as the subject, adjusted for litter size. a,b,c Different letters indicate differences between groups, p < 0.05. *Significant effect of LPS, independent of (+ )-naloxone, p < 0.05. # Significant effect of (+ )-naloxone, independent of LPS, p < 0.05. Number of pregnant dams per group is shown in parentheses.
Scientific RepoRts | 6:36112 | DOI: 10.1038/srep36112 instilled into the uterine cavity between gestation sacs 27,28 . The dose of heat-killed E. coli was selected to elicit ~50% PTD after titration of E. coli in pilot studies. After E. coli administration on gd 16.5, mice were monitored for time of delivery and the survival of pups. Instillation of killed E. coli into the uterus on gd 16.5 caused PTD in 70% (7/10) of treated mice (Fig. 3A), with an associated 36 h average reduction in gestation length (Fig. 3B) and 66% reduction in viable litter size at birth (Fig. 3C). The timing of birth and live birth rate were substantially rescued by (+ )-naloxone treatment, with none of 10 mice delivering before gd 18.0, a comparable gestation length, and no change in mean number of viable pups born compared to PBS control ( Fig. 3A-C). Again, (+ )-naloxone alone had no impact on timing of birth or litter size ( Fig. 3A-C).
When surviving delivered pups were analysed for effects of treatments at 12-24 h after birth, mixed model ANOVA showed E. coli administration had a strong negative impact on pup birth weight (p < 0.001). (+ )-Naloxone did not affect birth weight independently, but a significant interaction with E. coli was present (p < 0.001). Post-hoc analysis revealed that co-administration of (+ )-naloxone alleviated the reduction in birth weight caused by E. coli treatment, such that the pups born after exposure in utero to both (+ )-naloxone and E. coli were not different in weight to pups of PBS-treated control dams (Fig. 3D).
The majority of pups born small were lost in the post-natal and pre-weaning period, with only 13% of pups born alive from dams administered E. coli surviving to 3 weeks of age, compared to 90% and 88% respectively of pups from dams given PBS or (+ )-naloxone controls. Co-administration of (+ )-naloxone rescued the majority of pups with 56% surviving to 3 weeks of age (Fig. 3E), at which time their growth was comparable to those of the control groups (Fig. 3F).
This experiment demonstrates the utility of the TLR4 antagonist (+ )-naloxone in preventing prematurity and perinatal death, and alleviating fetal and postnatal growth impairment, in a physiologically relevant model of preterm birth and in utero inflammation.

(+)-Naloxone abrogates fetal and uterine cytokine expression induced by heat-killed E. coli.
E. coli-induced preterm birth is elicited when ligation of TLRs activates an inflammatory cascade, initiated when activation of NFκ B triggers expression of pro-inflammatory cytokines Il1b, Il6 and Tnf in the fetal membranes and other gestational tissues 22,28,44 . In turn, cytokines induce uterine activation genes to cause prostaglandin synthesis and signaling, cervical ripening and uterine contractions 15,45 . To examine the mechanisms underlying the protective action of (+ )-naloxone, we evaluated whether pro-inflammatory cytokines induced by intrauterine instillation of E. coli are influenced by (+ )-naloxone administration. The placenta and fetal membranes, as well as maternal decidua and myometrium were collected from the two fetal units located immediately adjacent to the injection site, 4 h after E. coli administration with or without (+ )-naloxone treatment, as before. In the fetal membranes and placenta, the sites of earliest response to TLR ligation to trigger the labor cascade 22 , E. coli induced a 9.1-fold and 2.1-fold increase in Il1a expression, a 62-fold and 6.3-fold increase in Il1b, a 33-fold and 3.3-fold increase in Il6, a 131-fold and 3.7-fold increase in Tnf, and 95-fold and 4.3-fold increase in Il10, compared to tissues from control mice given PBS (Fig. 4A-J). Administration of (+ )-naloxone substantially reduced E. coli-driven cytokine expression particularly in fetal membranes, meeting statistical significance for Il1b, Il6, Tnf and Il10, which were reduced by 78%, 84%, 66% and 73% respectively ( Fig. 4G-J).
In the maternal tissues of the uterine decidua and myometrium, which are more spatially distant from the injection site, E. coli elicited increases in some of the same pro-inflammatory cytokines, although to a lesser extent than in fetal tissues ( Fig. 5A-L). (+ )-Naloxone given with E. coli challenge reduced decidual expression of Il1a by 62% (Fig. 5G), and reduced Tnf in both the myometrium and decidua by 43% and 45% respectively ( Fig. 5D-J). Most uterine activation genes (Gja1, Oxtr, Ptger4, Ptgfr and Ptgs1; data not shown) were not elevated by E. coli treatment, likely because of the short 4 h time window between treatment and tissue analysis. However, Ptgs2 encoding prostaglandin-endoperoxide synthase 2 (cyclooxygenase-2, Cox2) was increased in response to E. coli challenge in both the myometrium (12.1-fold, Fig. 5F) and decidua (7.0-fold, Fig. 5L) compared to control. A trend towards decreased Ptgs2 after (+ )-naloxone treatment in E. coli challenged mice did not reach statistical significance in either tissue (Fig. 5F,L). (+ )-Naloxone administered without E. coli had no impact on gene expression in any of the fetal or maternal tissues (Figs 4 and 5).
These data show that the surge in pro-inflammatory cytokines released from fetal tissues contacting killed E. coli to trigger PTD, is suppressed by (+ )-naloxone. This is consistent with an action of (+ )-naloxone in preventing TLR4 ligation to trigger initiation of the inflammatory cascade that ultimately leads to preterm birth.

Discussion
Normal on-time delivery is an inflammatory process that involves release of endogenous TLR ligands from the maturing fetus to ligate TLRs in fetal membranes and placental tissue 22,23,46 . Preterm delivery results when infection and/or sterile inflammatory agents cause pathogen-derived or endogenous TLR ligands to prematurely activate the TLR-induced inflammatory cascade 8,26 . Here we demonstrate that administration of a novel small molecule TLR4 antagonist (+ )-naloxone is an effective treatment to suppress PTD and associated perinatal death induced in mice by systemic LPS or intrauterine administration of heat-killed E. coli. Gene expression analysis reveals that (+ )-naloxone treatment inhibits E. coli-induced progression of the labor cascade by inhibiting induction in fetal membranes of pro-inflammatory cytokines Il1a, Il1b, Il6 and Tnf, which are identified as responsible for the amplification of inflammation and eventual maturation of the myometrium to overcome quiescence, induce cervical ripening, activate uterine contraction and expel the fetus 41,47 . All dams that received four doses of (+ )-naloxone at 12 h intervals, effectively encompassing the window between inflammatory challenge and earliest gestational age compatible with post-delivery survival in mice, delivered on time with the expected number of pups. The majority of pups survived the perinatal period and developed normally to 3 weeks of age. This demonstrates that (+ )-naloxone protection of fetuses from TLR4-driven inflammation in late gestation is consistent with normal birth, postnatal survival and development of offspring at least to weaning age.
Exposure to TLR4-induced inflammation impacted fetal growth in late gestation, in association with evidence of reduced placental transport function, as described in previous studies 40,48 . Although no protective effect on mean fetal weight in viable fetuses was discernible at gd 18.5, fetuses at this time had been exposed to a   Il1a (A,F), Il1b (B,G), Il6 (C,H), Tnf (D,I) and Il10 (E,J) mRNAs were determined in placenta (A-E) and fetal membranes (F-J) by qPCR and normalised to Actb. Data are mean ± SEM relative gene expression of n = 12 tissues from n = 6 dams/group and were analysed by oneway ANOVA and post-hoc Sidak t-test. a,b,c Different letters indicate differences between groups, p < 0.05. pro-inflammatory environment for just 48 h, and effects on growth were less evident than in the E. coli model where birth weights were examined. Reduced birth weight was evident in pups exposed in utero to E. coli, and this was at least partly reversed by (+ )-naloxone treatment.
TLR4 is present in many compartments of the gestational tissues and reproductive tract and is well-positioned as a key sensing system and rapid first-line response to ascending or systemic bacterial infection. Fetal membrane TLR4 expression appears of paramount importance as the membranes comprise the site of earliest response to endogenous TLR ligation to trigger term labor 22 . The highest fold-change expression in inflammatory cytokines induced by E. coli was seen in this tissue as opposed to placental or uterine tissues, and additionally cytokine suppression by (+ )-naloxone was most profound in this compartment.
Mouse models of preterm birth often report heterogeneous effects amongst dams including preterm delivery of dead and live pups, and/or fetal injury or death in utero without delivery 48,49 . In the current study most dams delivered dead pups within 36 h of LPS or E. coli administration. With this outcome it is not obvious whether fetal death occurs in utero or in the perinatal phase, and thus whether death is the cause or consequence of early delivery. However, the observation of fetal death in utero in several dams, 48 h after LPS administration at gd 18.5, is consistent with the view that fetal injury and death can occur in situ after inflammatory challenge, independently of early delivery. Likewise, birth of live pups in a small proportion of preterm deliveries suggests fetal death is not essential for premature activation of the maternal parturition response. Consistent with previous reports 48 , we found that live pups born before gd 18.0 died soon after birth, presumably due to a combination of developmental immaturity and inflammatory injury inflicted in utero. These observations are consistent with interacting but distinct causal pathways underpinning the effects of inflammatory triggers on fetal and maternal tissue compartments.
Suppression of TLR4-induced pro-inflammatory cytokine expression in gestational tissues is inferred to be the key mechanism underlying the failure of bacterial agent-induced inflammation to progress to premature labor in mice given (+ )-naloxone, in turn protecting mice from subsequent late gestation or perinatal death of fetuses. Our data using both LPS and heat-killed E. coli to induce preterm delivery are consistent with previous studies employing these models 27,28 , where elevated IL1B, IL6 and TNF signaling are implicated in progressing both the uterine maturation and injury to fetal tissues that ultimately results in preterm birth and fetal death respectively.
Administration of (+ )-naloxone to mice given E. coli substantially suppressed cytokine expression, particularly in fetal membranes and variably in placenta, the tissues most proximal to E. coli deposition between fetal sacs. This demonstrates that inhibition of TLR4 signaling in these tissues can prevent the progression of the inflammatory cascade that culminates in preterm birth and fetal death. It is notable that (+ )-naloxone is efficacious despite the fact that E. coli activates inflammation by activation of TLRs in addition to TLR4, including TLR2 42,43 . Different TLRs can trigger different cellular responses in gestational tissues, for example in placental trophoblasts where TLR4 ligation induces cytokine expression whereas TLR2 predominantly leads to apoptosis 19 . This presumably affects placental function and fetal growth, thus explaining why despite robust protection from preterm birth and perinatal death, (+ )-naloxone was only partially effective in mitigating fetal growth impairment and protecting growth restricted pups from death in the postnatal phase following E. coli treatment in utero. Although (+ )-naloxone is a small molecule that is able to penetrate the placental barrier 37 , it is possible that (+)-naloxone accesses the maternal and fetal compartments to differing extents, perhaps reaching higher and more effective concentrations in maternal tissues than in fetal tissues. Further experiments are required to investigate this, as well as to evaluate how (+ )-naloxone affects cervical ripening, another important aspect of the parturition cascade.
In addition to pro-inflammatory genes, we examined effects of E. coli and (+ )-naloxone on expression of UAGs. Induction of UAGs is secondary to hormone changes and pro-inflammatory cytokine upregulation 15 so given the short time frame of the gene expression experiment, it is unsurprising that most UAGs were not elevated 4 h following E. coli administration. Expression of Ptgs2 which controls PGE2 synthesis was elevated 4 h after E. coli administration, and a trend to suppression in its expression was evident with (+ )-naloxone treatment. (+ )-Naloxone has previously been shown to reduce LPS-induced TLR4-mediated Ptgs2 and PGE2 synthesis in mouse macrophages 36 . PGE2 is a key factor in the prostaglandin signaling pathway required for cervical ripening and parturition 50 . Sustained delivery of (+ )-naloxone beyond a single dose may be required to achieve greater Ptgs2 suppression.
Previous studies support the notion of targeting TLR4 to prevent infection-associated preterm birth. Blockade of TLR4 signaling with anti-TLR4 monoclonal antibody reduces leukocyte activation and the incidence of preterm labor induced by LPS 24 . In mice given Fusobacterium nucleatum, a gram negative bacteria associated with preterm birth and premature rupture of membranes in women 51 , administration of a TLR4 antagonist lipid A mimetic CXR-526 acted to reduce fetal loss 30 . CXR-526 did not suppress bacterial colonization of the placenta, but did reduce the extent of necrosis within placental tissue 30 . The authors concluded that decreased placental necrosis might reflect TLR4 antagonist-mediated suppression of pro-inflammatory cytokine expression, as reported previously for CXR-526 in a mouse model of inflammatory bowel disease 52 , but cytokines were not directly investigated in that study.
Other studies identify downstream agents of TLR4-driven inflammation as additional effective targets for inhibiting infection-induced preterm birth 53 . Experiments in null mutant mice implicate IL1 and IL6, cytokines induced by TLR4 activation of NFκ B, in amplifying and accelerating the birth cascade 41,47 , while anti-inflammatory IL10 suppresses IL1 and IL6 production and protects against LPS-induced preterm birth 48,54 . As well as cytokines, immune regulators that attenuate TLR-mediated inflammation can impact susceptibility to LPS-induced preterm birth. For example, the cannabinoid receptor Cnr2 modulates dendritic cell production of IL10 and IL6, such that mice genetically deficient in Cnr2 are resistant to LPS-induced preterm birth 55 . Prostaglandins synthesized after Ptgs2 induction by LPS administration contribute to cervical ripening and other endpoints of parturition, and suppression of Ptgs2 with specific inhibitors can modulate LPS-induced PTD Scientific RepoRts | 6:36112 | DOI: 10.1038/srep36112 rate 56 . Pretreatment with rosiglitazone, which promotes placental synthesis of peroxisome proliferator-activated receptor-γ (PPARγ ) similarly attenuates responsiveness to LPS-induced PTD, in association with suppressed NFκ B signaling and reduced synthesis of TNF, IL1, IL6 and other pro-inflammatory chemokines 57 . Thus the current study adds to considerable evidence that TLR-driven inflammatory signaling can be attenuated to effectively reduce susceptibility to preterm birth.
TLR4 is an attractive drug target in preterm birth because of its role at the apex of the inflammatory pathway. (+ )-Naloxone and related drug compounds have potential benefits over neutralising antibodies and other TLR4 antagonists. (+ )-Naloxone potently blocks LPS-induced TLR4-mediated signaling in a variety of non-pregnancy models, suppressing NFκ B activation and inhibiting TNF and IL1B induction in immune cells 34,35 . As well as capacity to access and cross the placental barrier 37 , the established safety profile of closely related (−)-naloxone in pregnancy and infants is encouraging. In humans, (−)-naloxone is approved for use in pregnancy and no negative neonatal outcomes are reported 58,59 . Given the lack of opioid receptor activity of (+)-naloxone owing to the stereo-selectivity of opioid receptors 33 , (+ )-naloxone has distinct advantages over the currently available (−)-naloxone by virtue of its inability to alter endogenous opioid signaling, which in a clinical setting would afford freedom to employ exogenous opioids for maternal pain relief in labor.
Together, these results indicate that (+ )-naloxone is a useful experimental drug for investigating the utility of small molecule TLR inhibitors as pharmacological agents that might be administered in association with antibiotics to contain infection and suppress progression to preterm birth 53 . While clearly there are major challenges to consider in extrapolating this work to clinical application, we speculate that such TLR4 inhibitors may have particular value as prophylactic agents in threatened preterm labour where increased TLR4 expression disposes to elevated susceptibility, even in the absence of infection 60,61 , or when TLR4 SNPs associated with an increased risk of PTD and premature rupture of membranes are identified 62,63 . A likely advantage over current tocolytic agents is that TLR4 inhibitors act upstream of the pro-inflammatory cascade, overcoming the limitation of agents such as prostaglandin inhibitors which suppress uterine contractility, the final phase of labour, without impacting upstream pro-inflammatory activity 45,64 .
A key consideration is the impact on the neonate of in utero exposure to TLR4 inhibitors, and their effect on fetal capacity to withstand intrauterine inflammation, which is present to some extent in the majority of human preterm deliveries 8 . Sequalae of in utero inflammatory insult include fetal and newborn brain white matter disease, cerebral palsy, necrotizing enterocolitis, and chronic lung disease 8 , causing neurodevelopmental disability and a range of recurrent health problems in childhood 5 . Inflammation can provoke fetal brain injury even when inflammation is insufficient to activate parturition 49 , indicating the risk of sustained exposure to inflammatory mediators in utero. Clearly, clinical progression of this work requires extensive investigation of the benefits and risks of pharmacological delay of preterm birth for offspring, particularly effects on neurodevelopment, to ensure any treatment interventions adequately protect the fetus from inflammatory injury in utero.

Materials and Methods
Mice. C57Bl/6 (B6) mice were obtained from Laboratory Animal Services, University of Adelaide and maintained in the specific pathogen-free University of Adelaide Medical School Animal House with a 12:12 h lightdark cycle. Food and water were provided ad libitum. Animals were utilized in accordance with the NHMRC Australian Code of Practice for the care and use of animals for scientific purposes. All experimental protocols were approved by the University of Adelaide Animal Ethics Committee, using methods in accordance with the relevant guidelines and regulations.
One to three virgin female mice of 8-12 weeks of age were housed with a proven fertile B6 male and checked daily between 0800-1000 h for vaginal plugs, as evidence of mating. The morning of vaginal plug detection was designated gestational day (gd) 0.5. Females were removed from the male and housed individually. Mice were monitored for evidence of delivery at 12 h intervals after the gd 16.5 treatment until gd 18.5. Vaginal bleeding or the presence of intact or partial fetal tissue, viable or non-viable pups in the cage was noted. Delivery of live or dead pups before gd 18.0 was designated PTD. At gd 18.5 all mice were killed by cervical dislocation and the intact uterus was removed and analysed as described 40 . At autopsy, maternal outcomes were classified as pregnant with viable fetuses (at least one normal fetus of > 800 mg); delivered early (>1 live pup born gd 18.0-18.5), delivered preterm with live pups (>1 live pup born gd 16.5-18.0), or delivered preterm with dead pups (gd 16.5-18.0, all pups dead). The % maternal outcome in each category for each treatment group was calculated as [number]/[total pregnant mice] × 100. Fetal outcomes were classified as viable in utero (normal fetus > 800 mg), non-viable in utero (fetus hemorrhagic, anemic, malformed, or severely growth retarded < 600 mg), delivered preterm, viable (fresh implantation scar, viable pup), delivered preterm, dead (fresh implantation scar, no viable pup) or resorbed (hemorrhagic mass < 5 mm in diameter). The % fetal outcome in each category for each treatment group was calculated as [number]/[total implantation site] × 100. When viable fetuses were present, each fetus was dissected from the amniotic sac and umbilical cord, fetuses and placentae were weighed, and the fetal:placental weight ratio calculated.
Treatment and pregnancy outcomes, in utero E. coli model. E. coli (serotype o55:K59(B5):H-) from American Type Culture Collection No. 12014 was grown overnight in 1000 ml Luria broth (LB) with shaking at 37 °C. Bacteria were washed by centrifugation in sterile PBS and killed at 100 °C for 20 min, verified by lack of growth overnight on LB agar plates. Prior to heat treatment, aliquots of live bacteria were plated on LB agar plates at 1 in 10 serial dilutions to quantify colony forming units (cfu). The heat-killed E. coli suspension was stored in aliquots of 1 × 10 11 cfu/ml at − 80 °C until use.
Pregnant mice were anesthetised with avertin at 1100 h on gd 16.5 and a 1.5 cm midline incision was made on the lower abdomen. Heat-killed E. coli (5 × 10 10 cfu in 100 μ l) or PBS was injected into the midsection of the left uterine horn at a site between two adjacent fetuses, in a treatment shown to elicit at least 50% PTD, as described 28 . The abdominal incision was closed in two layers, using sutures through the peritoneal wall and the skin. Mice were administered (+ )-naloxone or vehicle as above immediately after wound closure, plus a further 3 doses at 12 h intervals as above, and allowed to recover. Females were monitored until birth for the time of parturition and the number of viable pups born, as above. Pups were weighed at 12-24 h post-delivery, at 8 days of age and at weaning.
Cytokine and uterine activation gene expression. A second cohort of pregnant females was administered E. coli or vehicle, as above. The site of injection was marked by a single interrupted suture through the uterus layer prior to injection. Mice were killed by cervical dislocation 4 h post-surgery and the intact uterus of each female was removed. The two implantation sites adjacent to the injection site were each dissected into uterine myometrium (between implantation sites), entire uterine decidua (at placental attachment site), placenta and fetal membranes, and snap-frozen in liquid N 2 , then stored at − 80 °C.
Tissues were homogenised using ceramic beads (Mo Bio) in Trizol (Ambion RNA, Carlsbad CA) and RNA was precipitated using isopropanol and ethanol, then DNAse-treated using Ambion DNA-free ™ Kit according to the manufacturer's instructions. RNA purity and concentration were determined by A 260 and A 280 (NanoDrop, Wilmington, DE) and RNA integrity was verified by denaturing agarose electrophoresis. First strand cDNA was reverse-transcribed from 2 μ g extracted RNA with Superscript III (Invitrogen Life Technologies, Carlsbad, CA) according to the manufacturer's instructions, with 200 ng random sequence oligohexamers (Geneworks, Adelaide, Australia) and 500 ng oligo dT 18 Table S1) and RNase free water. PCR reactions were 10 min at 95 °C followed by 40 cycles of 95 °C for 15 sec and 60 °C for 45 sec, using a Rotorgene 6000 (Corbett Life Sciences, Sydney, Australia). PCR product integrity was confirmed by High Resolution Melt analysis. Data were normalized to β -actin mRNA expression and expressed as Δ Δ CT using the formula mRNA level = Log 2 -(Ct Bactin − Ct target gene ).

Statistical analysis.
All statistical analysis was conducted using SPSS for Windows, version 20.0 software (SPSS Inc, Chicago, IL). Data were tested for normality using a Shapiro-Wilk test. ANOVA and post hoc T-tests were used when data were normally distributed. Kruskal-Wallis and Mann-Whitney U-test were used when data were not normally distributed. Pregnancy outcome and fetal outcome data expressed as percentage were compared by χ 2 analysis. Gestation length and viable litter size data expressed as mean ± standard error of the mean (SEM) are analyzed by two-way ANOVA and post-hoc LSD test. Fetal weight, placental weight, fetal:placental weight ratio, growth trajectory and body composition data is expressed as estimated marginal means ± SEM and analysed by Mixed Model Linear Repeated Measures ANOVA and post-hoc LSD t-test, with mother as subject. Growth curve data was also analysed by Area Under the Curve (AUC) analysis in R studio. qPCR data are mean ± SEM analyzed by one-way ANOVA and post-hoc Sidak t-test (to adjust for multiple comparisons). Differences between groups were considered significant when of p < 0.05.