|Home | About | Journals | Submit | Contact Us | Français|
Preterm birth is a leading cause of neonatal morbidity and mortality. Although microbial invasion of the amniotic cavity (MIAC) is associated with the majority of early preterm births, the temporal events that occur during MIAC and preterm labor are not known. Group B Streptococci (GBS) are β-hemolytic, gram-positive bacteria, which commonly colonize the vagina but have been recovered from the amniotic fluid in preterm birth cases. To understand temporal events that occur during MIAC, we utilized a unique chronically catheterized nonhuman primate model that closely emulates human pregnancy. This model allows monitoring of uterine contractions, timing of MIAC and immune responses during pregnancy-associated infections. Here, we show that adverse outcomes such as preterm labor, MIAC, and fetal sepsis were observed more frequently during infection with hemolytic GBS when compared to nonhemolytic GBS. Although MIAC was associated with systematic progression in chorioamnionitis beginning with chorionic vasculitis and progressing to neutrophilic infiltration, the ability of the GBS hemolytic pigment toxin to induce neutrophil cell death and subvert killing by neutrophil extracellular traps (NETs) in placental membranes in vivo facilitated MIAC and fetal injury. Furthermore, compared to maternal neutrophils, fetal neutrophils exhibit decreased neutrophil elastase activity and impaired phagocytic functions to GBS. Collectively, our studies demonstrate how a unique bacterial hemolytic lipid toxin enables GBS to circumvent neutrophils and NETs in placental membranes to induce fetal injury and preterm labor.
Preterm birth is a leading cause of neonatal morbidity and a direct cause of one-third of neonatal deaths (1, 2). Intraamniotic infection and inflammation are major risk factors for fetal injury, early preterm births, stillbirths, and early onset fulminant neonatal infections (3, 4). The infected amniotic fluid often contains organisms typically colonizing the lower genital tract including Group B Streptococcus (GBS; Streptococcus agalactiae) (5, 6).
GBS are β-hemolytic, gram-positive bacteria that typically exist as recto-vaginal colonizers in healthy adult women. However, during pregnancy ascending GBS infection can lead to fetal injury, stillbirth or preterm birth. Despite the success of intrapartum antibiotic prophylaxis to prevent maternal to infant transmission during labor and delivery, GBS remains a leading cause of neonatal morbidity and mortality (7, 8). Effective therapies to prevent GBS fetal sepsis, preterm birth or stillbirth are lacking. A recent report indicated that maternal colonization of GBS can be associated with increased rates of infants being transferred to the neonatal intensive care unit (9). Furthermore, maternal sepsis due to GBS can predispose infants to adverse outcomes that include preterm birth or stillbirth (10). A better understanding of host immune responses in the placenta that normally protect the fetus from ascending infection of lower genital tract organisms like GBS, is pivotal to development of preventive therapies.
Although a comprehensive understanding of GBS virulence factors that enable the pathogen to breach placental membranes and induce preterm birth or stillbirth are lacking, we recently showed that increased expression of the hemolytic pigment enables GBS to penetrate human placental membranes ex vivo (11). We have also shown that hyperhemolytic/hyperpigmented GBS strains, some with mutations in the transcriptional repressor of the hemolytic pigment known as CovR/CovS (or CsrR/CsrS), can be isolated from the amniotic fluid and placental (chorioamniotic) membranes of women in preterm labor (11). Further, we and others have demonstrated that expression of the hemolytic pigment induces fetal death in pregnant mouse models of GBS infection (12, 13).
Despite these advances, there are limitations to the above model systems. For example, the use of human placental membranes ex vivo does not permit investigation of host immune cells, which may be recruited to prevent microbial invasion of the amniotic fluid and fetus during pregnancy. Also, as antibiotics are routinely administered during Cesarean sections (14–16), the transfer of these antibiotics to the human placenta can impose limitations on bacterial studies performed with placental membranes ex vivo. Although animal models of pregnancy address the role of host immune defenses during an active infection, lower mammalian models differ significantly from human pregnancy in key respects including dissimilarities in reproductive anatomy, placentation, mechanism of labor onset and sensitivity to pathogens. In contrast, the pregnant nonhuman primate (NHP) emulates human pregnancy and is considered the closest animal model for studies related to human pregnancy (17–20). Similarities of NHP to humans include reproductive anatomy, number of fetuses (singleton), long gestational period (160–170 days), type and structure of placenta (hemomonochorial), initiation of labor (hormonal control of parturition), sensitivity to pathogens and timeline of fetal lung and brain development (19, 20). In our chronically catheterized pregnant NHP model (21) we are able to inoculate bacteria at the choriodecidual space, which lies between the uterine muscle and the placental membranes, where bacteria first encounter the maternal-fetal interface during ascending infection from the lower genital tract (4, 21).
To elucidate temporal events that occur during MIAC and preterm labor, we used the chronically catheterized pregnant NHP model (21). Previous studies using this model revealed that choriodecidual inoculation of a wild type GBS strain (serotype III, strain COH1) induced cytokine production that was associated with fetal lung injury without MIAC or overt chorioamnionitis or preterm labor (21). Interestingly, this human isolate of GBS is mildly hemolytic/pigmented in contrast to certain other GBS strains (22). As GBS strains with increased expression of the hemolytic pigment were recovered from the amniotic fluid (AF) of women in preterm labor (11), we utilized the chronically catheterized pregnant NHP model to understand how hyperpigmented GBS (lacking the gene covR, GBSΔcovR) evade host immune responses in vivo during MIAC. Although CovR/S is a transcriptional repressor of the cyl genes important for hemolytic pigment expression, this two-component system also controls the expression of more than 100 genes in GBS (23–25). Therefore, to evaluate the role of the hemolytic pigment on MIAC and preterm labor in the NHP model, we also included an isogenic, nonpigmented GBS covR mutant that lacked the gene cylE important for hemolytic pigment expression (11, 22), as a control (GBSΔcovRΔcylE). Here, we show that hyperpigmented GBS rapidly invaded the AF and induced preterm labor in pregnant NHP due, in part, to the ability of the hemolytic pigment to induce neutrophil cell death and evade killing by neutrophil extracellular traps (NETs).
To understand how the hemolytic pigment may promote GBS invasion of the amniotic fluid and fetus, we utilized our unique chronically catheterized NHP model. Ten animals received choriodecidual inoculations of 1–3 × 108 colony forming units (CFU) of either hyperpigmented GBSΔcovR (n=5) or control nonpigmented GBSΔcovRΔcylE (n=5); these results were compared with saline controls (n=5) that were previously described (21).
The primary and secondary study outcomes from this study are shown in Table 1. Our primary outcome was a composite of preterm labor and/or MIAC, because either event results in a poor pregnancy outcome. We observed that inoculation of the hyperpigmented ΔcovR was associated with an adverse pregnancy outcome in 5/5 (100%) animals when compared to 2/5 (40%) animals inoculated with the nonhemolytic GBSΔcovRΔcylE or 0/5 (0%) saline controls (Table 1). Preterm labor occurred in 4/5 animals inoculated with GBSΔcovR (excluding GBSΔcovR 3) compared to 1/5 with GBSΔcovRΔcylE and 0/5 saline controls (Tables 1 and S1). In the other animal inoculated with hyperpigmented GBSΔcovR (GBSΔcovR 3), the amniotic fluid (AF) became dark and cloudy due to MIAC at 12 hours after inoculation, the decision was made to proceed with Cesarean section earlier than the defined study endpoint (preterm labor) to avoid stillbirth due to fetal sepsis. At the time of Cesarean section at 48 hours post inoculation, this animal (GBSΔcovR 3), had a sustained pattern of increased uterine contractions, but had not yet made cervical change to meet criteria for development of preterm labor. Overall, MIAC and fetal sepsis were observed in 3/5 animals inoculated with GBSΔcovR versus 1/5 animals inoculated with GBSΔcovRΔcylE and 0/5 saline controls (Tables 1 & S1, Figs. 1, S1 and S2). In all cases of MIAC, the fetus became septic and GBS could be recovered from multiple organs. Overall, the bacterial burden in GBS infected fetal organs ranged from 102–106 CFU/g tissue with consistently more bacteria recovered from the fetal lung when compared to the other fetal organs (Fig. S3).
An increase in uterine activity was often seen within a few hours following inoculation of GBSΔcovR with MIAC and preterm labor occurred rapidly in most of these cases. Three animals from the GBSΔcovR group (GBSΔcovR 1, 2, 3; see Tables 1 & S1; Fig. 1B, S1A & S1B) developed sustained uterine contractions within hours after inoculation; bacteria were recovered as early as within 15 minutes (0.25 hours) in one animal, 45 minutes (0.75 hours) in another and within 12 hours in the third case. Due to the increase in uterine contractions and cervical dilation (GBSΔcovR 1, 2) or dark and cloudy AF (due to bacteria) that imposed concerns for stillbirth (GBSΔcovR 3), a Cesarean section was performed within 6, 24 and 48 hours after inoculation, respectively (Table S1). In all these cases, GBS was recovered from fetal organs (Table S1, & Fig. S3). In the remaining two animals infected with GBSΔcovR, rapid uterine contractions and cervical dilation indicative of preterm labor were seen without MIAC resulting in Cesarean section at 24 and 72 hours post inoculation, respectively (GBSΔcovR 4 and 5; Table S1, Fig. S1C & D).
In the 5 animals infected with the nonpigmented isogenic control GBSΔcovRΔcylE (see Tables 1 & S1; Fig. 1C & S2), adverse outcomes were detected in two animals (GBSΔcovRΔcylE 1, 3). Preterm labor developed in one animal without MIAC (GBSΔcovRΔcylE 3; see Tables 1 & S1; Fig. S2D). In a second animal (GBSΔcovRΔcylE 1), MIAC without preterm labor was detected at the time of Cesarean section on day 3 (Fig. S2C). Unfortunately, AF could not be recovered from the amniotic catheter until the experimental end point on day 3 and thus we could not determine the time course of MIAC; GBS were recovered from the fetal organs of this animal (Fig. S3). The other animals did not exhibit signs of preterm labor or MIAC. Taken together, our results indicate that choriodecidual inoculation of hyperpigmented GBSΔcovR induced adverse outcomes such as preterm labor, MIAC with concerns for stillbirth and fetal sepsis more frequently than in animals inoculated with nonpigmented GBSΔcovRΔcylE or in saline controls (P =0.03, GBSΔcovR vs. GBSΔcovRΔcylE; P =0.0009, GBSΔcovR vs. saline).
We then examined cytokine responses in the AF and fetal tissues. Levels of AF interleukin-1 beta (IL-1β), tumor necrosis factor alpha (TNF-α), interleukin-6 (IL-6) and interleukin 8 (IL-8) were all significantly higher in animals inoculated with hyperpigmented GBSΔcovR versus saline controls (Table 1, all P <0.05). In animals inoculated with nonpigmented GBSΔcovRΔcylE, AF IL-1β and IL-8 were significantly higher compared to saline controls (Table 1, P <0.05), but levels of TNF-α and IL-6 were not significantly different. The effect of hemolytic pigment expression was assessed by comparing mean peak levels of AF cytokines between animals infected with GBSΔcovR compared to GBSΔcovRΔcylE. Notably, the AF mean peak levels of IL- 6 and IL-8 were significantly higher in hyperpigmented GBSΔcovR inoculated animals versus those inoculated with nonpigmented GBSΔcovRΔcylE (Table 1, P <0.05). At delivery, fetal plasma IL-1β, TNF-α, IL-6 and IL-8 levels were all significantly higher in the GBSΔcovR compared to saline controls (Table 1; all P <0.05); fetal IL-8 showed a trend towards higher levels in animals inoculated with GBSΔcovR compared to nonpigmented GBSΔcovRΔcylE groups (P =0.06). IL-1β and TNF-α were also higher in animals infected with GBSΔcovRΔcylE compared to saline controls. Regardless of MIAC, levels of TNF-α were significantly increased in fetal organs of animals infected with GBSΔcovR compared to GBSΔcovRΔcylE (Fig. S3), suggesting the onset of severe fetal systemic inflammation (26).
We next compared histological sections of the chorioamniotic membranes at the inoculation site from the hyperpigmented GBSΔcovR, nonpigmented GBSΔcovRΔcylE and saline controls. Histological lesions associated with placental infection of GBSΔcovR appeared to rapidly progress and were dominated by widespread accumulation of neutrophils and necrosis as the time interval from inoculation to delivery increased. Within 6 hours of GBSΔcovR inoculation neutrophils aggregated within blood vessels in the chorion, marginated to the endothelium and occasionally were seen surrounding small vessels within the chorionic trophoblast layer. At this early timepoint, neutrophils also lined the interface between chorion and trophoblast layers, but did not extend into the chorion or amnion (Fig. 2A and D, corresponds to animal shown in Fig. S1B). By 24 hours after inoculation, neutrophil accumulation was more pronounced with wide, dense layers of neutrophils in the chorion without invasion of the amnion (Fig. 2B and E, corresponds to animal shown in Fig. S1A). Small vessels in the chorion and trophoblast layers appeared thrombosed with focal areas of necrosis. At 48 hours post inoculation, there was widespread severe inflammation and necrosis, which was predominantly neutrophilic and involved all layers of the chorioamnion including the amniotic epithelium. (Fig. 2C and F, corresponds to animal shown in Fig. 1B).
With the exception of a single animal where MIAC was observed, there was minimal to no inflammation within the chorioamniotic membranes from animals inoculated with nonpigmented GBSΔcovRΔcylE; neutrophils were occasionally seen within deeper regions of the placenta (Fig. 2H and J; corresponds to animal shown in Fig. 1C). In the case of the animal with microbial invasion of GBSΔcovRΔcylE, neutrophil infiltration of the membranes was severe, but the extent of necrosis and neutrophil density was less than in the most affected hypervirulent GBSΔcovR infected animal (Fig. 2C and F).
Taken together, these observations indicate that neutrophil recruitment into the chorioamniotic membranes is the primary placental innate immune response to bacterial infection. However, frequent MIAC as observed with GBSΔcovR suggests that despite neutrophil recruitment, the placental innate immune response may be inadequate or rendered ineffective to prevent bacterial trafficking into the amniotic cavity.
To further investigate the innate immune response following infection by hyperpigmented (ΔcovR) and nonpigmented GBS (ΔcovRΔcylE), we performed microarray analysis on RNA isolated from the inoculation site of the chorioamniotic membranes in GBS infected animals versus saline controls (n=5/group). Out of a total of 31,740 probesets for host gene expression, 797 were differentially regulated in chorioamniotic membranes between animals inoculated with GBSΔcovR compared to saline controls and 530 were differentially regulated between animals inoculated with GBSΔcovR compared to those inoculated with GBSΔcovRΔcylE (≥1.5 fold change; P <0.005; heatmap shown in Fig. S4). A subset of differentially expressed probe sets is shown in Table S2 and the entire list is available through GEO link GSE80248 ( http://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?token=cpifasaozdulpkl&acc=GSE80248.) Approximately half of the differentially regulated probe sets for host gene expression (263) between GBSΔcovR and GBSΔcovRΔcylE infected animals were also significant in the animals infected with GBSΔcovR versus saline suggesting that the pigment has an outsized effect on gene expression compared to other GBS factors (Venn diagram shown in Fig. S5A). However, many genes that showed increased expression in the GBSΔcovR infected animals show similar trends in the one GBSΔcovRΔcylE infected animal that exhibited MIAC (Fig. S4). These data suggest that some of the pigment-mediated effect on host gene expression may, in part, be due to its ability to promote MIAC.
Overall, significantly upregulated genes in animals inoculated with GBSΔcovR compared to saline controls or GBSΔcovR compared to GBSΔcovRΔcylE included genes encoding factors associated with inflammation, chemokines, cytokines and cell adhesion molecules such as matrix metallopeptidase 8 (MMP8), interleukin-1 alpha (IL-1A), interleukin-1 beta (IL-1B), vanin 2 (VNN2), chemokine receptor (CXCR1), and Selectin-L (SELL). Interestingly, apart from factors associated with preterm labor such as matrix metalloproteinases and pro-inflammatory cytokines and chemokines, GBSΔcovR infected membranes showed increased expression of cell adhesion molecules associated with neutrophil rolling, transmigration and arrest such as CD177, Selectin-L, and ICAM-1 (27, 28) which is consistent with the neutrophil influx observed in these membranes (Fig. 2).
Downregulated genes included genes important for maintenance of cell-cell junctions and cytoskeletal organization such as desmocollin 2 (DSC2), desmoplakin (DSP), integrin alpha 6 (ITGA-6), neuroepithelial cell-transforming gene 1 (NET1), occludin (OCLN) and ADAM metallopeptidase (ADAMTS6). These data indicate that the chorioamniotic membranes responded to infection by hyperpathogenic GBS by increasing expression of factors that directly promote neutrophil recruitment and also diminishing expression of factors that maintain cytoskeletal organization and the adhesive properties of cell layers such as desmosomes and tight junctions. Decreased expression of desmosome and tight junction proteins may facilitate neutrophil transmigration (29). A panel of selected genes was validated by real-time quantitative reverse transcriptase polymerase chain reaction (qRT-PCR). Levels of gene expression obtained with amplified RNA samples were compared to β-actin as a control for each sample. There was a significant correlation between microarray and qRT-PCR for differential gene expression between GBSΔcovR versus saline control animals (Fig. S5B, all P <0.05) and for animals inoculated with the two GBS strains for IL-1β, IL-1α, IL18RAP, CCDC6, NET1, SORBS2 and PIK3C2G (Fig. S5C, all P <0.05). Other genes tested by qRT-PCR showed similar trends of positive or negative gene expression when compared with the microarray analysis (Fig. S5C).
Given that neutrophils were the primary immune cells observed in the chorioamnion of pregnant NHP infected with GBSΔcovR (Fig. 2), we asked if the frequent MIAC observed in GBSΔcovR animals could, in part, be attributed to the ability of the hemolytic GBS pigment to induce neutrophil cell death. While some studies have indicated that hemolytic GBS strains can induce either apoptosis or pyroptosis in macrophages (30, 31), how the GBS hemolytic pigment modulates neutrophil function is not completely understood. Since our discovery that the GBS hemolysin is the ornithine rhamnolipid pigment and not a protein toxin (11), we have shown that the purified hemolytic pigment induces osmotic lysis of red blood cells, is cytotoxic to human amniotic epithelial cells (hAECs) and promotes NLRP3 dependent pyroptosis in macrophages (11, 12). To test the possibility that the purified GBS hemolytic lipid toxin may induce neutrophil cytotoxicity, primary human neutrophils were isolated from fresh adult human blood. Using flow cytometry, we confirmed that ~92.7% of the cells that were isolated were positive for the neutrophil markers CD15 and CD16 (Fig. S6). The isolated neutrophils were then exposed to various concentrations of purified GBS pigment (0.625–5μM) for four hours and cytotoxicity was estimated by measuring the release of lactate dehydrogenase (LDH) as described (32). As controls, we included an equivalent amount of ΔcylE extract (i.e. pigment extraction procedure was performed using the nonpigmented GBSΔcylE strain) or Buffer DTS (DTS: DMSO+0.1%TFA+20% starch; see materials and methods and (11, 12)). The data shown in Fig. 3A indicates that cytotoxicity or percent cell death was significantly higher in neutrophils treated with the hemolytic GBS pigment when compared to control ΔcylE extract or DTS buffer. To confirm that hemolytic pigment mediated neutrophil cell death was also observed with live bacteria, neutrophils were treated with GBS strains namely WT COH1 or isogenic mutants (hyperpigmented ΔcovR or nonpigmented strains i.e. ΔcovRΔcylE or ΔcylE) at various multiplicity of infection (MOI: 1, 10, 100) for four hours and cytotoxicity was measured by LDH release as described (12, 32). Similar to observations with purified pigment, hyperpigmented GBSΔcovR induced neutrophil cell death in a dose dependent manner (Fig. 3B), whereas no significant neutrophil cell death was observed with WT GBS or the nonpigmented GBS strains ΔcovRΔcylE and ΔcylE at any MOI tested. These data indicate that GBS hemolytic pigment induces neutrophil cytotoxicity in a dose dependent manner.
We next performed scanning electron microscopy (SEM) to determine if GBS pigment or hyperpigmented GBS induce morphological changes to neutrophils. To this end, adult neutrophils were treated with 0.5 μM pigment or an equivalent amount of control ΔcylE extract for 10 min. Untreated neutrophils were included as controls. The results shown in Fig. 3C (see top panel) indicate that the morphology of untreated or resting neutrophils (see Media in Fig. 3C) is rounded as shown previously (33). In contrast, neutrophils exposed to the GBS hemolytic pigment exhibit severe morphological changes when compared to untreated neutrophils or those treated with the control ΔcylE extract (Fig. 3C). Consistent with these observations, neutrophils exposed to the nonpigmented GBSΔcovRΔcylE strain also had a greater number of rounded or partially rounded cells when compared to neutrophils exposed to the hyperpigmented GBSΔcovR strain (see bottom panel in Fig. 3C). Collectively, these data indicate that the hemolytic pigment of GBS accelerates morphological changes in neutrophils ultimately leading to neutrophil cell death.
As neutrophils typically undergo cell death via apoptosis (34), we first examined the possibility that the GBS hemolytic pigment may accelerate neutrophil apoptosis. Therefore, we compared the ability of GBS pigment/lipid toxin to induce membrane permeabilization and apoptosis in neutrophils. To this end, neutrophils were either treated with GBS pigment (0.5μM) or controls (ΔcylE extract or buffer DTS). Then, we measured uptake of the membrane impermeable dye propidium iodide (PI) at various times post treatment. A significant increase in the number of fluorescent (PI positive) cells was observed in neutrophils at 30 and 60 min post treatment with GBS pigment indicating loss of membrane integrity, which was not seen in controls (ΔcylE or buffer DTS, see Fig. S7 A–C). To determine if the neutrophil membrane damage induced by pigment was preceded by classical events in apoptosis [extracellular exposure of phosphatidylserine (PS)], we examined binding of fluorescent Annexin V (AV). We observed that uptake of fluorescent PI (Fig. S7 A, B, C) preceded AV staining (Fig. S7 D, E, F) in pigment treated neutrophils. These data suggest that pigment induced neutrophil cell death was independent of apoptosis.
We next examined if the purified hemolytic pigment or hyperpigmented GBS induced neutrophils to undergo an inflammatory form of cell death known as pyroptosis, similar to our observations with macrophages (12). As release of IL-1β is a characteristic feature of pyroptosis (35), we measured cytokine levels using luminex bead assays on supernatants obtained from neutrophils treated with GBS strains (WT, ΔcovR, ΔcovRΔcylE, ΔcylE) or pigment and controls for 4 hours (see materials and methods). Interestingly, levels of cytokines IL1β, IL-6, TNFα, GRO- α, IFNγ or IL-10 were all below the limit of detection. Notably, only WT and nonpigmented/nonhemolytic GBS strains (ΔcovRΔcylE, ΔcylE) triggered the release of IL-8 from neutrophils. IL-8 release was not observed with hyperpigmented GBS (ΔcovR, Fig. S8A) likely due to pigment mediated neutrophil cell death. Also, significant activation of caspase 3/7 was not observed in pigment treated neutrophils when compared to the positive control staurosporine (Fig. S8B). Collectively, these data suggest that GBS pigment induction of neutrophil cell death is not due to apoptosis or pyroptosis but may involve other pathways (i.e. lysis or necrosis), which likely contribute to the adverse outcomes observed during GBS pregnancy associated infection.
Neutrophil activation is associated with generation of reactive oxygen species (ROS) (36). To determine if exposure to the hemolytic GBS pigment induced the generation of ROS, neutrophils were pre-treated with dihydrorhodamine123 (DHR) and exposed to GBS pigment (0.5 μM) or an equivalent amount of control ΔcylE extract or buffer DTS. The conversion of DHR to fluorescent monohydrorhodamine (MHR) indicates generation of ROS and was measured by flow cytometry. Treatment of neutrophils with GBS pigment stimulated the generation of ROS, which was not observed with control ΔcylE extract or buffer DTS (Fig. S9 A, B).
To confirm that GBS strains also induced ROS production, neutrophils were pre-treated with dihydrorhodamine123 (DHR) and exposed to GBS strains (WT, hyperhemolytic ΔcovR or nonhemolytic ΔcovRΔcylE, ΔcylE) at an MOI of 100. Notably, hyperhemolytic GBS accelerated the production of ROS in neutrophils within 15 minutes (Fig. S9C, D). Subsequently, we observed that most neutrophils treated with pigment or hyperpigmented GBS succumbed to cell death and therefore generation of ROS was not sustained.
We next compared the ability of various GBS strains that exhibit differences in hemolytic pigment expression for their ability to resist neutrophil killing. To this end, we compared survival of the GBS strains (WT, hyperpigmented ΔcovR, nonpigmented ΔcovRΔcylE, ΔcylE and capsule deficient ΔcpsK) in the presence of purified human neutrophils. These experiments were performed in the absence of serum as described previously (37), in order to understand the role of the hemolytic pigment, independent of complement opsonization. We observed that hyperpigmented GBSΔcovR exhibit increased survival in the presence of neutrophils when compared to nonhemolytic or acapsular GBS strains (Fig. S9E). These data suggest that despite inducing generation of ROS, increased expression of the hemolytic toxin enables GBS to subvert neutrophil killing mechanisms likely by promoting neutrophil cell death.
Extrusion of neutrophil contents has been associated with formation of neutrophil extracellular traps (NETs, (38, 39)). Formation of NETs is dependent on neutrophil components that include NADPH oxidase, generation of ROS, myeloperoxidase, neutrophil elastase and histone deamination (39–41). While one function of NETs is to trap and promote extracellular killing of bacterial and fungal pathogens (33, 42, 43), certain bacterial toxins such as the Staphylococcus aureus leukotoxin GH have been shown to induce NETs (44). As GBS-infected neutrophils in vitro exhibited DNA extrusion (45), we were interested to determine if the purified hemolytic GBS pigment is sufficient for induction of NETs. To this end, neutrophils in NET assay buffer were treated with either GBS pigment (5μM), equivalent amount of control (ΔcylE extract) or with the GBS (hyperhemolytic ΔcovR, or nonhemolytic ΔcovRΔcylE) at an MOI of ~ 10 for 4 hours at 37°C. Phorbol myristate acetate (PMA, 20 nM) was included as a positive control. SEM was then performed and the results shown in Fig. 4A. The top panel in Fig. 4A shows that the hemolytic GBS pigment induced the formation of NETs similar to PMA, whereas the control ΔcylE extract did not induce significant NET formation. Similarly, the hyperhemolytic GBSΔcovR strain robustly induced NET formation whereas the ΔcovRΔcylE strain showed slightly attenuated NET formation (Fig. 4A, lower panel). As neutrophil elastase activity is associated with NET formation (46), we measured NET associated neutrophil elastase. Briefly, neutrophils in NET assay buffer were treated with either pigment or the GBS strains as described above. After removal of soluble neutrophil elastase and treatment with S7 nuclease, NET associated neutrophil elastase activity was measured as described in Materials and Methods. The results shown in Fig. 4B indicates that, consistent with the SEM images, increased NET associated neutrophil elastase activity was observed with pigment and hyperhemolytic GBSΔcovR compared to controls or nonpigmented GBS ΔcovRΔcylE strain, respectively.
We then examined the ability of the GBS strains to resist killing by NETs. To this end, neutrophils were induced to produce NETs and the ability of the GBS strains (WT COH1, hyperhemolytic ΔcovR, nonhemolytic ΔcovRΔcylE,) to resist killing by NETs was evaluated. As sialic acid deficiency on the GBS polysaccharide capsule was previously described to increase sensitivity to NET killing (47), we included the sialic acid deficient GBSΔcpsK strain (48) as a control in the NET killing assay. The results shown in Fig. 4C indicates that the hyperhemolytic GBSΔcovR strain was resistant to killing by NETs unlike the nonhemolytic ΔcovRΔcylE strain or the control capsule deficient ΔcpsK. Taken together, our observations suggest that the increased virulence of the hyperhemolytic GBSΔcovR strain may primarily be due to its ability to induce neutrophil cell death and may be augmented by resistance to NETs.
To confirm that NET formation occurs during GBS infection in vivo, we stained the chorioamniotic membranes of nonhuman primates (NHP) infected with either GBSΔcovR or GBSΔcovRΔcylE for neutrophil-derived elastase and DNA (DAPI) as described (33, 49). Membranes from saline controls were also included. The results shown in Fig. 5 demonstrate increased NETs (i.e. co-localization of extracellular DNA with neutrophil elastase) in chorioamniotic membranes of NHP infected with GBSΔcovR (animal from Fig. 1B) when compared to membranes infected with GBSΔcovRΔcylE (animal from Fig. 1C) or the saline control. Overall, increased NETs were observed in chorioamniotic membranes of ΔcovR infected samples when compared to GBSΔcovRΔcylE (Fig. S10). Collectively, these data indicate that NETs are formed in the chorioamniotic membranes during ascending GBS infection and that expression of the hemolytic pigment enables GBS to circumvent extracellular killing by NETs.
We also tested the hypothesis that fetal neutrophils may be more susceptible to the hemolytic GBS pigment when compared to maternal neutrophils. To test this hypothesis, we obtained maternal and cord blood from non-laboring pregnant women undergoing elective Cesarean section at term. Neutrophils were isolated from maternal and cord blood pairs and flow cytometry confirmed that the isolated cells were enriched for the neutrophil markers CD15 and CD16 (Fig. S11). We first compared the sensitivity of maternal and fetal neutrophils to the GBS pigment by measuring LDH release (Fig. 6A) or PI uptake (Fig. S12). While both maternal and fetal neutrophils were sensitive to GBS pigment compared to controls (ΔcylE extract and DTS buffer), fetal neutrophils were similar in their sensitivity to the GBS pigment when compared to maternal neutrophils for LDH release (Fig. 6A) and PI uptake (Fig. S12). We next tested the release of ROS from maternal and fetal neutrophils after exposure to GBS pigment or controls. Here, we observed that conversion of DHR to fluorescent MHR indicative of ROS generation was significantly attenuated in fetal neutrophils when compared to maternal neutrophils (compare panels ii and iii to v and vi in Fig. 6B). We then compared the survival of various GBS strains (WT, hyperhemolytic ΔcovR and nonhemolytic ΔcovRΔcylE) in the presence of maternal and fetal neutrophils. Notably, while survival of the nonhemolytic GBSΔcovRΔcylE was significantly attenuated when compared to the hyperhemolytic ΔcovR in maternal neutrophils, this trend was not observed with fetal neutrophils (Fig. 7A). We then compared the release of NET-associated neutrophil elastase between maternal and fetal neutrophils exposed to hemolytic pigment or GBS strains. The results shown in Fig. 7B indicates that fetal neutrophils showed significantly lower NET associated neutrophil elastase activity when compared to maternal neutrophils exposed to GBS pigment or hyperpigmented GBS. These data indicate that decreased ROS generation and NET formation by fetal neutrophils may contribute to increased susceptibility of neonates to many GBS strains.
Collectively, the results presented in this study indicate that GBS utilizes a unique rhamnolipid pigment toxin to circumvent neutrophils and neutrophil extracellular traps in the chorioamnion to facilitate microbial trafficking into the amniotic cavity and fetus during pregnancy.
GBS infections during pregnancy remain a significant public heath concern (8). Although current methods of prevention of GBS transmission from mother to fetus during labor and delivery has made remarkable progress in decreasing early onset disease, intrapartum antibiotic prophylaxis does not address or decrease ascending GBS infection leading to preterm births or stillbirths. Recent reports indicate the increasing prevalence of GBS in pregnant women and preterm delivery (50) and GBS infection associated stillbirths are currently estimated at 12.1% (51). The lack of a mechanistic understanding of events that enable the pathogen to circumvent the maternal and fetal immune system during pregnancy imposes limitations in developing alternate methods of prevention and treatment.
In this study, we used a highly relevant, chronically catheterized pregnant nonhuman primate model that closely emulates human pregnancy. The tethered chronic catheter preparation was used for all in vivo experiments and is ideal for studying maternal-fetal immunologic responses. This model allows us to investigate uterine responses and immune mechanisms during the course of a bacterial infection beginning from the time of inoculation until preterm labor, which would be impossible in humans. Using this model, we demonstrate how increased expression of the hemolytic pigment enables GBS to penetrate the placental chorioamniotic membranes and infect the amniotic fluid and fetal organs. Interestingly, we observed that inoculation of hyperpigmented GBS into the choriodecidual space can result in bacterial invasion of the amniotic cavity and fetus as early as within 15 minutes to within a few hours post infection. GBS invasion of the amniotic activity induced increased expression of neutrophil recruiting cytokines and chemokines in the chorioamniotic membranes and amniotic fluid. Consistent with these findings, we observed significant recruitment of neutrophils in the chorioamnion of NHP infected with hyperpigmented GBS. These results indicate that neutrophils comprise the initial and primary host defense mechanism utilized by the chorioamnion to combat invasive pathogens such as GBS. However, despite the recruitment of neutrophils, the hyperhemolytic GBS strain was able to traffic across the chorioamniotic membranes in 3 of the 5 animals and bacteria were recovered from multiple fetal organs. These observations suggest that although neutrophils are recruited to the site of infection, increased expression of the hemolytic pigment may impair their phagocytic function. Our findings in the pregnant NHP model are consistent with previous reports on the role of hemolysin in exacerbating GBS infections such as meningitis, (52), experimental sepsis (53, 54), lung injury (55, 56), urinary tract infections (57) and fetal demise (12, 13) in murine, rat or rabbit models. Although expression of the GBS hemolytic pigment is typically associated with generation of a proinflammatory response and neutrophil recruitment, this appears ineffective in curtailing bacterial burden as observed by others ((13, 52, 57) and this work).
We then examined how increased expression of the hemolytic pigment may enable GBS subvert neutrophils. Our studies revealed that the hemolytic pigment and hyperpigmented GBS induced neutrophil cell death within four hours, in a dose dependent manner. The neutrophil cell death observed with the GBS pigment is independent of apoptosis or even pyroptosis. While we previously noted that macrophages lacking the NLRP3 inflammasome were able to recover from the membrane permeability induced by the GBS hemolytic pigment (12), the short lived nature of neutrophils may prevent remodeling of neutrophil membranes thereby increasing their susceptibility to the GBS lipid toxin. These data indicate that cell death due to GBS pigment may be due to direct lysis or necrosis. Indeed, SEM of neutrophils exposed to the GBS toxin revealed severe morphological changes and interestingly, the toxin induced NET formation both in vitro and in vivo. To our knowledge, NET formation in placental membranes due to microbial infections has not been shown. Prior to our study, NET formation in placental membranes was only reported in association with a non-infectious condition i.e. preeclampsia (49). Here, we demonstrate that in vivo bacterial infections induce NET formation in placental membranes. Recently, studies using mouse models have shown that live neutrophils form NETs in vivo to limit systemic bacterial infection (58). It is likely that in NHP infected with hyperpigmented GBS, both live and dying neutrophils contribute to NET formation in vivo. Whether NETs themselves contribute to placental dysfunction as suggested (59) is not known.
Typically NET formation is associated with the ability of neutrophils to ensnare bacteria for further antimicrobial action. However, we observed that GBS strains that overexpress the hemolytic pigment were resistant to killing by NETs when compared to the nonhemolytic strain. This may in part be due to the ability of the unsaturated polyene chain in the hemolytic pigment toxin to quench reactive oxygen species (ROS). Previous work by Liu et. al. (30) demonstrated the antioxidant nature of the GBS pigment. They observed that hyperpigmented GBS strains are more resistant to ROS such as hydrogen peroxide, superoxide and singlet oxygen in vitro and in macrophages (30). Although expression of an extracellular nuclease (nuclease A) in GBS was shown to degrade NETs (60), previous work by others and us showed that expression of nuclease A is not under CovR/S regulation in GBS (23, 61). This suggests that differences in resistance to NET killing between GBSΔcovR and GBSΔcovRΔcylE is likely not due to altered endogenous DNAse activity. Our observations suggests that increased expression of the hemolytic toxin enables GBS to induce neutrophil cell death and likely resist killing by NETs, which may promote MIAC during ascending infection. It is also noteworthy that a few reports have indicated that NETs may only entrap bacteria to prevent dissemination and wall off infection, without actually inducing bacterial cell death (62, 63). Addition of DNase at the end of the NET killing assay is thought to relieve the clumping effect and provide accurate results for discrimination between ensnaring of bacteria by NETs versus bacterial killing by NETs (62), While we repeated these experiments with addition of DNase at the end of the experiment to confirm increased sensitivity of the GBSΔcovRΔcylE to NETS, it remains plausible that GBS entrapment by NETs rather than NET associated killing regulates MIAC in vivo.
Interestingly, we observed that two of five animals inoculated with the hypervirulent GBS strain did not develop MIAC despite exhibiting preterm labor. We predict that this may in part be due to early neutrophil recruitment wherein the number of neutrophils recruited may have surpassed the bacterial load to effectively curtail GBS invasion of the amniotic fluid and fetus. Recently, using a mouse vaginal colonization model, we showed that mast cell degranulation can promote neutrophil recruitment and enable eradication of hypervirulent GBS from the vaginal tract (32). While the mast cell response of humans and NHP to vaginal microorganisms remains unknown, we posit that an early and effective neutrophilic inflammatory response is essential for prevention of ascending GBS infection. It is likely that in the absence of neutrophil cell death, the expression of neutrophil recruiting cytokines and chemokines by chorioamniotic membranes and the release of IL-8 from activated neutrophils themselves may support the recruitment of additional neutrophils that assist in GBS phagocytosis. Consistent with this hypothesis, Mohammadi et al. recently reported that murine bone marrow derived neutrophils released TNF and IL-1β when exposed to GBS for 24 hours (64). Taken together, these observations suggest that when neutrophils escape cytotoxic killing by the GBS hemolytic pigment, cytokine responses may promote additional neutrophil recruitment.
Our results also revealed the attenuated ability of human fetal neutrophils to combat GBS. When compared to maternal neutrophils, fetal neutrophils exhibited decreased ROS generation and NET associated neutrophil elastase activity when exposed to GBS pigment or hyperpigmented GBS strains. These data are consistent with previous observations of decreased or delayed NET formation by neonatal neutrophils upon treatment with inflammatory stimuli such as PMA and LPS (65, 66). Furthermore, fetal neutrophils were not as efficient as maternal neutrophils in curtailing nonhemolytic GBS strains. While previous studies reported that pregnancy can been associated with diminished oxidative burst function in maternal neutrophils (67), our data suggests that compared to fetal neutrophils, maternal neutrophils play a critical role in restraining lower genital tract pathogens such as GBS from accessing fetal compartments and tissues.
Although in this study, we describe the role of the hemolytic pigment in MIAC and preterm labor in the NHP model, many virulence factors are likely to play critical roles in the outcome of GBS infections. It is also important to note that differences between clinical strains of GBS are not always limited to differences in hemolytic pigment expression but can include differences in expression of many other virulence factors, which were not examined in this study. Further studies that explore the role of other pathogenic determinants in GBS infection-associated preterm birth in pregnant animal models will provide new insight into the repertoire of virulence factors utilized by this pathogen to cause fetal injury, preterm birth or stillbirth.
In summary, we show that GBS utilizes the hemolytic pigment to invade the amniotic fluid and fetus during pregnancy by circumventing neutrophils and NETs in placental membranes. Understanding how pathogens circumvent host immune responses at the maternal/fetal interface is a critical first step for determining strategies to prevent microbial trafficking across the chorioamniotic membranes during pregnancy.
Fig. S1: Uterine contractions, amniotic fluid cytokines, prostaglandin and bacterial CFU from choriodecidual inoculations of hyperpigmented GBSΔcovR in four chronically catheterized pregnant Macaca nemestrina.
Fig. S2: Uterine contractions, amniotic fluid cytokines, prostaglandin and bacterial CFU from choriodecidual inoculations of nonpigmented GBSΔcovRΔcylE in four chronically catheterized pregnant Macaca nemestrina.
Fig. S3: Choriodecidual inoculation of hyperpigmented GBSΔcovR in pregnant nonhuman primates induced bacterial invasion of fetal tissues and increased levels of TNF-α.
Fig. S4: Heat map demonstrating differential gene expression in chorioamniotic membranes from pregnant nonhuman primates.
Fig. S5: Changes in chorioamniotic gene expression.
Fig. S6: FACS (fluorescence-activated cell sorting) of neutrophils purified from human blood.
Fig. S7: The hemolytic pigment toxin induces membrane permeability in neutrophils.
Fig. S8: Neutrophil cell death due to GBS pigment dampens IL-8 release (A) and is not associated with activation of caspase 3/7 (B).
Fig. S9: The hemolytic pigment and hyperpigmented GBS induces generation of reactive oxygen species (ROS) in neutrophils (A–D). Hyperpigmented GBS are more resistant to neutrophil killing (E).
Fig. S10: Increased NETs in the chorioamnion of NHP infected with hyperpigmented GBSΔcovR.
Fig. S11: FACS (fluorescence-activated cell sorting) of neutrophils purified from human maternal and fetal (cord) blood.
Fig. S12: The hemolytic pigment toxin induces membrane permeability in maternal and fetal neutrophils.
Table S1. Detailed pregnancy outcomes by animal
Table S2: Selected genes showing differential expression in NHP chorioamniotic membranes infected with GBS strains or saline.
Table S3: qPCR probes and primers.
We are grateful to the human subjects who participated in our study. We thank Amy Gest, Jan Hamanishi, Gina Heidel, Danny Power, Joyce Karlinsey and Connie Hughes for their assistance. We acknowledge Drs. Craig Rubens and Michael Gravett for study design related to performance of the original nonhuman primate experiments.
FUNDING: This work was supported by funding from the National Institutes of Health, Grant # R01AI100989 to L.R and K. A. W, R56AI070749, R01AI112619 and R21AI109222 to L. R, and P30HD002274. The NIH training grants T32 HD007233 (PI: Lisa Frenkel) and T32 AI07509 (PI: Lee Ann Campbell) supported E.B and J.V, respectively.
The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or other funders. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.
AUTHOR CONTRIBUTIONS: E. B., C.G., J.V., L.R and K.A.W designed the research. E. B., C.G., L.N., C. B., J. V., M.C., S.M., B.A., C.W., V.A., V. S-U., J.O., M.G., S.S., L.R and K.A.W performed the experiments. E. B., C.G., L.N., C. B., J. V., M.C., S.M., B.A., C.W., V.A., S.S., J. W.M., T.K.B., A.B., H.D.L., L.R., and K.A.W analyzed the results. E.B., C.G., J. V., T.K.B., H.D.L., L.R., and K.A.W wrote the paper.
COMPETING INTERESTS: The authors declare no competing financial interests.
DATA AND MATERIALS AVAILABILITY: The microarray data for this study have been deposited in the GEO database and is available through the GEO accession number GSE80248.