Abstract
Neonates, most notably prematurely born neonates, are particularly vulnerable to enteric pathogens. It is recommended that all pregnant patients receive antibiotics following Preterm Premature Rupture of Membranes (PPROM) to prolong pregnancy, yet not all antibiotic regimens have been studied for efficacy, especially when considering neonatal outcomes. Here we find, in a clinical cohort of at-risk infants, broad spectrum antibiotic treatment, including cephalosporins, increased the risk of LOS in neonates in comparison to penicillin base treatments. We followed those initial clinical findings with a preclinical model, here showing that prenatal cephalosporin antibiotics, but not penicillins, decreased mammary IgA in dams, and limited the availability of IgA within the lumen of nursing offspring. Decreased IgA correlated with an outgrowth of commensal E.coli and increased colonization and translocation of pathogen E. coli. This increased risk for mortality in a model of LOS that was associated with decreased mammary IgA. We propose maternal administration of select antibiotics could limit maternal IgA availability by disrupting mammary IgA, leading to increased risk of LOS in infants by allowing for pathogen colonization in the neonatal intestine.
Introduction
Late-onset neonatal sepsis (LOS) resulting from bloodstream infections occurring 7 days following delivery, are a major contributor to neonatal mortality. Very low birth weight (VLBW, <1500 grams) infants are particularly at risk for LOS, with an incidence of 10%, and a mortality rate of 18% (1). In a substantial portion of LOS, the causative pathogen, including gram-positive or gram-negative bacteria, in can be found residing in the neonatal intestine as part of the intestinal microbial community implying intestinal translocation as the bacterial source (2–6). Accordingly, multiple animal models of LOS show gut-administered enteric pathogens can breech the intestinal barrier and secondarily disseminate causing systemic disease (7–9). Thus, the route of LOS is in stark contrast to early-onset sepsis, neonatal sepsis occurring >72 hours following delivery, thought to be the result of a bacterial infection introduced in utero or during the delivery process. Human studies have also demonstrated that enteral route of infection contributes to morbidity and mortality (10) of LOS. Escherichia coli (E. coli) is the second most common causative pathogen of LOS. Thus, there is a clear need to understand the causative risks for LOS compared to early-onset sepsis. In a substantial portion of LOS-causative bloodstream infections, E. coli has been demonstrated to reside in the neonatal gut microbial community prior to disease (11–14). As E. coli is a common member of the neonatal intestinal microbiota, luminal protection within the neonatal intestinal is key to preventing opportunistic pathogens from disseminating systemically.
Maternal milk contains numerous factors that are protective against the development of LOS, including oligosaccharides, secretory antibodies, growth factors and lactoferrin that contributed to protection against infection in the neonate (15–17). Overall, maternal milk has been linked with lower infant death, hospitalization risk and risk of sepsis (18–21). Immunoglobulin A (IgA) has long been known to be a valuable component of maternal milk in providing protection to the neonate by binding pathogens (22, 23). Maternal IgA cross-reacts to a variety of gut resident pathogens, including viruses, bacteria, and parasites (24–26), and shapes the developing neonatal microbiome (27–29). Maternal IgA originates from plasma cells that migrate from the maternal gastrointestinal tract (30, 31), delineating the “Gut-Mammary Axis” (32).
While neonatal antibiotic use has been shown to increase risk to LOS, perinatal maternal antibiotics can also affect the neonate (33, 34). Antibiotics are prescribed in 25-30% of all pregnancies, making them the most common medication given to pregnant women (35). Indeed, in contemporary management of Preterm Premature Rupture of Membranes (PPROM), it is recommended that all pregnant patients receive antibiotics to prolong pregnancy. Previous studies have demonstrated that broader spectrum antibiotics were associated with increased LOS, NEC and adverse neonatal outcomes in patients who present with PPROM (36). Despite this, there are clinical trends towards use of additional broad-spectrum antibiotics without prospective safety studies, culture guided treatments and without neonatal follow up. This is particularly relevant when using salvage regimens for antimicrobial therapy in the setting of a penicillin allergy. Recent observations in mice has shown maternal antibiotics throughout lactation can decrease mammary IgA through influence of the gut microbiota (31). In this study, we sought to determine if antenatal use of broader spectrum antimicrobials leads to increased susceptibility to LOS and NEC in a high-risk cohort of premature infants who were exposed to maternal breast milk. We found a strong association with alternate, broader spectrum maternal antibiotic use and the development of LOS. Additionally, broad spectrum antibiotics lead to increased susceptibility of LOS in mice that was secondary to significantly reduced maternal IgA passed to the nursing neonate with decreased binding of E. coli in the nursing offspring. Thus, maternal antibiotics may increase risk of LOS in infants through the reduction of pathogen-reactive IgA in maternal milk.
Results
Antenatal cephalosporin and non-beta lactam use is associated with increased rates of LOS
To determine if there was a clinical association of antenatal antibiotic exposure and adverse neonatal outcomes, we generated a cohort of neonates born to mothers with PPROM (Fig. 1). The recommended regimen following PPROM is ampicillin IV for 48 hours followed by amoxicillin orally for 7 days in combination with a macrolide antibiotic (traditionally erythromycin, and not azithromycin). The recommended antibiotics in the setting of a penicillin allergy are unclear and left up to the clinician. We identified patients through ICD9/ICD10 codes in the maternal chart and detailed maternal and neonatal chart review (Fig. 1). There was no difference in pregnancy latency, gestational age at delivery, gestational age at PPROM, days that patient received antibiotics for and breast milk feeding of neonates (Supplementary Table 1). When assessing neonatal outcomes, there was no difference in select neonatal outcomes including early onset sepsis and NEC>2A. However, there was a significant increased risk of late onset sepsis with exposure to broader spectrum antibiotics with an OR of 4.836 (p<0.0008) (Table 1). A subanalysis demonstrated that 2 neonates who developed LOS in the traditional antibiotics cohort were fed donor breast milk, the remainder were fed with maternal breast milk that was fortified as needed. There was incomplete data for one neonate who was treated for LOS in the non-traditional cohort. In this context, a comparison was made between maternal breastmilk exposed infants in the traditional and nontraditional regimens. There remained a substantial increase in the cases of LOS in the nontraditional cohort (13.3% versus 3.1%, OR 4.789 p=0.0084).
Clinical cohort schematic
A sub-analysis of the alternate antibiotic regimen demonstrated that there was no difference between the cephalosporin regimen and non-beta lactam regimens with similar rates of late onset sepsis in both cohorts (Table 2). Positive blood cultures were also more likely in the non-traditional regimen in the work-up for late onset sepsis (6.38 vs 1.27%, p<0.05). Composite neonatal morbidity was higher in the non-penicillin-based regimens, but this did not achieve statistical significance. Taken together, this data demonstrates that non-penicillin and azithromycin-based regimens with broader activity against Gram negative enteric bacteria and E. coli (M. Shaffer and C. Megli, submitted) are associated with increased rates of LOS.
Maternal Cephalosporin is associated with Reduced Mammary IgA
We next moved into a preclinical model to investigate the impact of perinatal antibiotics on infant health. To model antenatal antibiotic use, a variety of clinical-relevant dosages of antibiotics (including amoxicillin, ampicillin, cephalexin, metronidazole, and ceftazidime) were administered to pregnant dams via the drinking water for seven days: from embryonic day 14 until delivery. Mice were then placed on regular drinking water following delivery. On postnatal day 2 (PN2), IgA was significantly reduced in mammary glands following cephalexin and ceftazidime treatment, but not other penicillin-based antibiotics or metronidazole (Fig. 2A). This reduction was specific to the mammary glands, and neither cephalexin nor ceftazidime significantly decreased IgA in the serum (Fig. 2B). In the pups born to cephalexin or ceftazidime-treated dams, there was a significant decrease in intestinal IgA (Fig. 2C) but not serum IgA (Fig. 2D).
Cephalosporins decrease mammary IgA and luminal IgA in nursing pups. Stated antibiotics were administered to pregnant dams in drinking water from ED 14 to delivery of litter, and tissue was analyzed on PND2. Total IgA was measured by ELISA from A) mammary glands of dams, B) serum of dams, C) colon contents of pups, and D) plasma of pups. Cephalexin was administered to pregnant dams in drinking water from ED 14 to delivery of litter, and tissue was analyzed on PND7. Total IgA was measured by ELISA from E) stomach contents (ST), small intestinal contents (SI), stool, and serum of pups, and F) stool, mammary and serum of dams. One way ANOVA was used to calculated statistical significance. N=3 dams and corresponding litters per group.
To understand the duration of decreased IgA in the mammary glands of dams and intestines of nursing pups, we next assessed pups at PN7, and found decreased IgA in both the mammary glands and neonatal intestinal tracts following prenatal cephalexin administration (Fig. 2D,E). In contrast, there was no reduction of IgG in the dam or pups (Supplementary Fig. 1A,B). Thus, maternal use of cephalexin and ceftazidime reduces the IgA in the mammary glands and provided to the nursing pups, without limiting systemic IgA or IgG.
Reduced mammary IgA is associated with increased E. coli colonization in mice
Though the cephalexin was administered to the dams exclusive to pregnancy, we observed a noticeable shift in the microbiota of the pups, rather than the dams. Principal component analysis revealed the microbiota of pups clustered distinctly from the dams, as expected given the age-dependent development of the microbiota (Fig. 3A). However, microbiota from cephalexin-treated dams did not separate distinctly from untreated dams, suggesting any effects on the intestinal microbiota by cephalexin was transient (Supplementary Figure 2). Pups of cephalexin-treated dams had the most observable microbial difference: an increase in Enterobactericeae (Fig. 3B) and a significant increase in the relative abundance of E. coli (Fig. 3C), suggesting the reduction in maternal IgA was associated with an increase in E. coli in the offspring. Antibiotic treatment of pups or pregnant dams increased susceptibility to Klebsiella pneumonia in a model of LOS, which was attributed to a reduction of Lactobacillus spp, including Lactobacillus murinis in the intestinal microbiota of the pups (8). Decreased Lactobacillus was a strong indicator of neonatal dysbiosis, and was associated with increased pathogen burden, suggesting lactobacillus is associated with colonization resistance to enteric pathogens. Cephalexin treatment of the dams did not reduce the abundance of L. murinis in the intestinal contents of the dams or pups following cephalexin treatment (Fig. 3C), suggesting the expansion of E. coli is not the result of reduced protection from L. murinis.
Cephalexin induced decrease mammary IgA is associated with increased E. coli colonization. A) Principal component analysis of 16s analysis of pups and dams indicate microbiota composition is based on phase of life (pups, blue vs dams, pink), and cephalexin (closed circles) had a more observable effect on pups compared to dams. B) Stacked bar plots of relative abundance of families in pup stool. C) Relative abundance of E. coli based on 16S analysis on stool of pups. D) Relative abundance of Lactobacillus murinus in dams and pups.
Mammary IgA cross-reacts and binds to a variety of pathogenic E. coli
Maternal IgG has been shown to target enterotoxigenic E. coli in the neonatal intestine and protect neonates from enteric infection (7), and we next sought to confirm maternal IgA can cross-react with extraintestinal pathogenic E. coli species. Human IgA from donor human milk was able to bind to a variety of E. coli strains (Fig. 4A), including both pathogenic and commensal E. coli from neonates. We also observed IgG cross-reacted to E. coli strains, though overall, there was less cross-reactive IgG compared to IgA. On average, donors had five times more cross-reactive IgA than IgG (Fig. 4B). We observed significantly more binding of pathogenic E. coli when compared to the lab strain K12 E.coli, suggesting maternal IgA may have improved cross-reactivity to intestinal-derived E. coli. Certain samples had higher amounts of crossreactivity to multiple E. coli isolates than others, which was independent of the total IgA concentrations in the donor milk (Fig. 4C). IgA from the stomach contents of PN7 pups was able to bind pathogenic E. coli, and this was decreased in PN7 pups from cephalexin treated dams (Fig. 4D), interestingly, cross-reactivity of another bacterial pathogen, group B streptococcus (GBS), remained unchanged after cephalexin treatment, suggesting an effect specific to E. coli-targeting IgA. Additionally, we did not observe a decrease in IgA-bound bacteria in the stool of pups following cephalexin treatment (Fig. 4E), suggesting a selective decrease in mammary IgA able to bind specific taxa. Thus, maternal IgA can cross-react to a variety of E. coli and has the potential to target E. coli and limit colonization in the nursing offspring.
Human milk cross reacts to E. coli commensals and pathogens. A) Cross-reactive IgA from human donor milk was quantified against indicated E. coli isolates. B) Cross-reactive IgA from human donor milk was quantified against indicated E. coli isolates. C) Cross-reactive profiles for individual donor milk. D) Cross-reactive IgA from the stomach contents of pups following dam cephalexin treatment was measured against group B Streptococcus (GBS), and E. coli. E) IgA+ bacteria from stool were measured by flow cytometry. N=20 donor milk specimens (A-C), N=3 independent litters per group (D). One-way ANOVA was used to calculate the statistical significance (A,B,D).
Reduced mammary IgA is associated with increased E. coli LOS in mice
It has been reported that the concentration of IgA in murine milk differs between murine strains, potentially due to the difference in polyreactive IgA across strains, with BALB/c mice having increased IgA compared to C57BL/6 mice (37, 38). Potentially due to this higher baseline of mammary IgA, we did not observe a decrease in mammary IgA in BALB/c dams following cephalexin treatment (Fig. 5A). Similarly, BALB/c pups from cephalexin-treated dams did not have decreased intestinal IgA (Fig. 5B). When comparing strains, we observed a slight decrease in IgA in the mammary glands and serum of C57BL/6 (mammary gland mean IgA 7.9 ug/ml), without any reduction in IgG, when compared to BALB/c (mammary gland mean IgA 35 ug/ml), similar to previous reports (compared Fig. 2B to Fig. 5A). This difference was more pronounced in the pups, and C57BL/6 pups had significantly less IgA in the stomach and small intestine (6.2 ug/ml and 17.8 ug/ml) compared to BALB/c mice at postnatal day 5 (24.5 ug/ml, and 23.8 ug/ml), without any significant difference in serum IgA, or IgG (compare Fig. 2A to Fig. 5B). Principal component analysis comparing the microbiota of the pups found the cephalexin treatment had distinct effects on BALB/c pups as compared to B/6 pups (Supplementary Figure 2). Notably, in BALB/c pups, Gammaproteobacteria, and E.coli was not a major component of the microbiota either with or without cephalexin (Fig. 5C). Thus, the increased E. coli in C57BL/6 pups of cephalexin treated dams is associated with decreased mammary IgA, and unlikely to be due to other off-target effects of cephalexin.
BALB/c. Cephalexin was administered to pregnant BALB/c dams in drinking water from ED 14 to ED21, tissue was analyzed on PND7. Total IgA (A) was measured by ELISA from stool, mammary glands, and serum of dams and (B) from stomach contents (ST), small intestinal contents (SI), stool, and serum of pups. C) Relative abundance of E. coli based on 16S analysis on stool of pups. 10 ^ 6 CFUs of pathogenic BSI-B E. coli were gavaged to BALB/c and C57BL/6 pups following cephalexin treatment of pregnant dams in drinking water from ED 14 to ED21. 10 ^ 6 CFUs of pathogenic BSI-B E. coli were gavaged to pups. D) Survival of pups following E. coli gavage on PN5. E) CFUs from stool were plated 48 hours following infection. F) CFUs from mesenteric lymph node (mLN) and spleen 48 hours following infection. One way ANOVA was used to calculated statistical significance, A,B,D,E. Kaplan Meier Curve was used in F. N=3 dams and corresponding litters per group.
Minimal shifts in the microbial communities of control and cephalexin treatment C57BL/6 and BALB/c dams was observed (Supplementary Figure 3), again suggesting cephalexin treatment effects on the microbiota were transient, and restored by the time of lactation. We then compared the microbial communities before (Day 0) and during cephalexin treatment (Day 4) and observed the C57B/6 dams and BALB/c dams had different responses to the cephalexin treatment. While both strains had decreased Lactobacillus strains, potentially a direct result of cephalexin in the drinking water, there was an expansion of the Bacteroidia, particularly Muribaculaceae in the C57Bl/6, while there was an expansion of the Clostridia, particularly Lachnospiraceae, in the BALB/c. Interestingly, both strains had Muribaculaceae and Lachnospiraceae resident in the intestinal microbiota prior to cephalexin, though at different abundances, so it remains unclear why cephalexin treatment resulted in to disparate trajectories of the microbiota during antibiotic administration.
To evaluate the functional role of decreased luminal IgA, PN5 pups were orally inoculated with 106 CFUs of BSI-B E. coli, a pathogenic ST131 E.coli isolated from the bloodstream of a neonatal LOS patient. This strain of E. coli has been shown to translocate from the intestine resulting in bacteremia in a model of neonatal sepsis (17). We observed increased mortality in the pups from cephalexin treated C57BL/6 dams (Fig. 5D), recapitulating an increased risk to LOS following cephalosporin use in our clinical cohort. There was no mortality seen in the pups from cephalexin treated BALB/c dams. 48 hours following oral infection, there was a significant increase in CFUs of E. coli in the stool of the infected pups from cephalexin treated C57BL/6 dams compared to the infected pups from untreated C57BL/6 dams or pups from BALB/c dams (Fig. 5E). The increased risk of mortality was associated with presence of E. coli CFUs in the spleen and mesenteric lymph node (MLN) at 48s hours post infection, though this was not significantly different between treatments, potentially due to the low overall mortality (40%) (Fig. 5F).
Taken together, we propose maternal IgA can limit intestinal colonization of E.coli in the nursing offspring and decreased maternal IgA following broad spectrum prenatal antibiotics increases risk of neonates to LOS. Our data have applications in clinical management as well as in understanding the pathophysiology of late onset sepsis in neonates.
Discussion
PPROM is responsible for approximately 1/3 of all preterm birth and is a substantial risk factor for complications associated with prematurity. The landmark Mercer trial on PPROM suggested antibiotic management prolongs gestation and neonatal outcomes. However, there was a lack of extensive neonatal follow up past discharge or transfer and sepsis was included in the primary outcome only if it occurred <72 hours after delivery, also defined as early-onset sepsis (39). This approach has been widely adopted in the United States and in our cohort patients universally received antibiotics and achieved a substantially longer mean latency than was reported in the initial RCT. Alternative antibiotics to the initial regimen have not been thoroughly studied with the exception of the ORACLE trial that exchanged ampicillin for ampicillin/sulbactam and amoxicillin for amoxicillin/sulbactam (36). In this trial the addition of beta-lactamases leads to increased rates of adverse neonatal outcomes but additional studies on other antibiotic regimens are not well studied.
The use of non-traditional antibiotics in the setting of PPROM is often secondary to a perceived penicillin allergy. In a retrospective re-analysis of the MFMU BEAM trial, non-beta lactam antibiotics were associated with increased rates of maternal endometritis and increased rates of adverse neonatal outcomes but did not reach significance. Cephalosporins are currently recommended for use with PPROM in the setting of a low-risk penicillin allergy, and therefore our analysis is different in that we included cephalosporin use in the non-traditional antibiotic cohort and saw increased rates of LOS. Taken together, these data suggest that more studies are needed when using broader spectrum antibiotics in the absence of infection. Our data further support the evidence in the ORACLE trial that broader spectrum antibiotics including those in the beta-lactam class, can have adverse effects on neonates.
Our data is further supported by our preclinical model: antenatal cephalosporin was associated with decreased maternal IgA in the mammary glands and gastrointestinal tract of the nursing neonates. Decreased maternal IgA has been associated with increased Enterobacteriaceae species that preceded necrotizing enterocolitis, another intestinally-related disease commonly seen in preterm infants, showing the potent ability of maternal IgA in limiting potential enteric pathogens in neonates (40). Here we show that in addition to protection from enteric disease by limiting intestinal colonization, maternal IgA also protects against systemic dissemination resulting in sepsis. This protection was strongly associated with maternal IgA and mammary IgA cross-reacted to E. coli more than IgG, and we did not observe a decrease in maternal IgG in mammary glands or the lumen of pups following cephalexin treatment.
Maternal IgA was associated with limiting both commensal E. coli as part of the microbiota, and orally administered pathogenic E. coli, as both were expanded in B/6 pups following cephalexin treatment. Prior observations have shown intestinal IgA can limit the growth of E. coli, and selective IgA deficiency in humans resulted in increased E. coli (41). Maternal IgA responses targeting E. coli have long been appreciated to be transferred to neonates in livestock mammals (22) and orally administered monoclonal IgA protect against enteric bacterial infections, including E. coli (42). E. coli colonization was not expanded in BALB/c mice, which had significantly more IgA compared to B/6 pups. Nor did E. coli expand in BALB/c mice following cephalexin treatment, explained by the lack of a decrease in IgA. There was no significant shift in E. coli within the intestines of dams, either between murine strains or after antibiotics. Thus, the expansion of E. coli in the pups is unlikely due to increased vertical transmission from the dams.
The source of mammary derived IgA has been long connected to gut derived IgA responses in the mother. IgA-committed lymphoblasts from the mesenteric lymph node home to the mammary glands specifically during late pregnancy (30). These IgA responses can be induced gut-associated lymphoid tissue, such as the Peyer’s patches in response to specific bacteria, including Parabacteroides goldsteinii, B. acidifaciens, and Prevotella buccalis (31). Disruption of these bacterial species with vancomycin treatment during pregnancy and throughout lactation reduced mammary IgA (31), however, we did not observe a significant shift in those species in our model. Indeed, the major microbial shift occurred during cephalexin treatment, when we observed different microbial trajectories in response to the antibiotics. In our model, cephalexin treatment, and the antibiotic-induced shift, is occurring during gestation between ED14-ED21. This could coincide with a time in life when IgA producing plasma cells are expanding or trafficking to the mammary gland in preparation for lactation. We only observed decreased maternal IgA with the cephalosporins: cephalexin and ceftazidime, and future work will focus on how these antibiotics are affecting the microbiota specifically during treatment and the larger affect this could have on mammary IgA production.
The basis for the specificity in IgA able to bind E. coli remains unclear. The decreased mammary IgA seemed to be directed to E. coli: we observed increased colonization of E. coli, and decreased binding of E. coli, following cephalexin treatment, and the decreased mammary IgA did not show a universal decrease in binding all bacteria. There was no decrease in cross-reactivity to GBS, and no decrease in overall IgA-bound bacteria in the stool. While dams contained E. coli within the intestinal microbiota at low abundance, they were not immunized against E. coli to induce a specific response. Similarly, the human IgA from donor milk had the ability to cross-react to multiple E. coli species. Thus, naturally induced IgA response can provide maternal protection. Naturally induced IgA have low-affinity but high cross-reactivity to a number of intestinal bacteria (43), and microbial exposure at the mucosal surfaces influence the B cell and IgA repertoire (44). Previous reports using auxotrophic E. coli suggest transient exposure to commensals can result in sustained IgA memory response in the absence of commensals (28). This IgA memory response can be disrupted following shifts to the microbiota (28). This has led to the speculation that IgA memory in response to commensals may be short-lived, with rapid turnover, and impaired expansion upon restimulation (45). However, E. coli in the dams was unchanged after cephalexin treatment, therefore it is unlikely antibiotic ablation of a specific species reduced IgA-producing plasma cells cross-reacting to that species. Further exploration into how this exposure can inform mammary IgA responses, and how disruptions to the maternal intestinal microbiota can disrupt mammary IgA responses remains should address this rapid crosstalk in the gut-mammary axis.
One limitation of our work is the limited translocation and mortality to E. coli, even following cephalexin treatment. The slight increase in translocation of E. coli to the spleen and MLN following cephalexin treatment potentially reflects the increased colonization in the lumen, presenting the pathogen with increased opportunity to translocate. We have previously shown E. coli translocation from the intestine was significantly increased when maternally supplied epidermal growth factor and offspring expressed epidermal growth factor receptor was disrupted, allowing for the formation of goblet-cell associated antigen passages, which allowed E. coli a route across the epithelium (9). This suggests multiple maternal components may work in tandem for the intestinal protection of neonate with IgA limiting pathogen colonization, and EGF limiting access through goblet-cell associated antigen passages.
These data should be considered with the reports advocating for treatment of intrauterine infection using broad spectrum antibiotics and expectant management. Moreover, our data further suggest that the microbiome-immune axis is relevant to neonatal health. Further studies in how antibiotics can cause a perturbation of the microbiome-immune axis and lead to consequences in maternal-child health are warranted.
In conclusion, selective maternal antibiotic treatment perinatally can result in decreased mammary IgA that is important for limiting pathogen colonization and expansion in the intestinal lumen of the nursing offspring. Our work builds on previous data showing the contribution of maternal IgA in influencing the developing microbiota (27) and maternal antibiotic use can decrease mammary IgA (31) by identifying a functional outcome of the decreased mammary IgA in the nursing offspring. IgA binding of bacteria can create a positive feedback loop the promote future IgA responses(37, 38), so disruption of this process may reduce future IgA responses against potential enteric pathogens in the offspring. Future directions should explore the gut-mammary axis to better understand how maternal antibiotics affect mammary plasma cells.
Methods
Patient inclusion
Patients were identified using ICD9/ICD10 codes for PPROM and delivering within the UPMC system between 2017-2020. Maternal and neonatal charts were reviewed, and multiple clinical variables and outcomes were collected and deidentified. Data were reviewed by a maternal fetal medicine specialist (CM) and a neonatologist (DB). Patients were excluded from analysis in the setting of previable delivery less than 23 weeks, stillbirth, incomplete information regarding antibiotic exposure and if they were not eligible for expectant management. Inclusion and exclusion criteria are demonstrated in Figure 1. This study was approved through the University of Pittsburgh IRB board.
Definition of clinical outcomes
Late onset sepsis was defined as diagnosis of sepsis greater than or equal to 7 days after delivery. Necrotizing enterocolitis was included if radiologic findings were consistent with necrotizing enterocolitis greater than 2A based on Bells Staging Criteria. Diagnosis of late onset sepsis and spontaneous intestinal perforation were included if patients were clinically diagnosed and treated. Neonates who underwent a diagnostic evaluation for late onset sepsis but did not complete treatment were not considered to have late onset sepsis.
Mice
C57BL/6J and BALB/cJ mice were purchased from the Jackson Laboratory and bred in-house at the Mayo Clinic. All mice were stored in standard housing. All mouse husbandry and experiments involving mice were approved by the Mayo Clinic Institutional Animal Care and Use Committee (IACUC).
Bacterial Isolates
E. coli strains BSI-A, BSI-B, BSI-C, BSI-D, BSI-E, and BSI-F are bloodstream isolates, along with non-pathogenic strains C1 and C2, which were isolated from stool from infants. (9) Commercial E. coli strain K12 (non-pathogenic) was purchased from the American Type Culture Collection (Manassas, VA, USA). All bacteria were cultured at 37°C in Lennox L (LB) Broth or Agar (Invitrogen, Waltham, MA, USA). Bacterial strains were made nalidixic acid resistant as previously described (9).
E. Coli Infection
Mice were orally gavaged with 50 uL of PBS containing 106 colony-forming units (CFUs) of BSI-B on day of life five. Mice were weighed at 24hrs post infection to assess the impacts of the infection. If mice had lost any weight at the 24hr check, they were euthanized and declared moribund. All surviving mice were euthanized at 48hrs post-infection.
Antibiotic Treatments
14 days after mating (embryonic day 14), the male and female were separated. At this point, antibiotics were given to the female pregnant mouse via drinking water until delivery. This included amoxicillin (0.2mg/mL), ampicillin (0.4mg/mL), cephalexin (0.4mg/mL), metronidazole (0.4mg/mL) or ceftazidime (0.4mg/mL). Mice were then placed on regular drinking water following delivery of the litter.
Quantification of Murine IgA/IgG
Murine IgA and IgG were quantified using the IgA Mouse Uncoated ELISA Kit and the IgG (Total) Mouse Uncoated ELISA Kit (Invitrogen). Samples were diluted in PBS, and assays were performed according to manufacturer’s instructions. The optical density was found using a BioTek 800 TS absorbance reader at 450nm. Concentrations of IgA and IgA were reported as ng/mL based on an intraassay standard curve.
16S Metagenomic Sequencing
Stool from pups and dams were taken from uninfected mice on PN7. Dam stool was sequenced per each individual dam, while stool from pups of the same litter were pooled for a representative litter sequence. Murine stool DNA was isolated using the PowerFecal Pro DNA Kit (QiAmp) and sequenced using the 16S Barcoding Kit 1-24 (SQK-16S024) on a GridION (Nanopore). Sequences were aligned using centrifuge (46, 47). Data was analyzed using RStudio software, and all mouse and human sequence data was removed during analysis. Data analysis included alpha diversity, principal component analysis, and taxonomical distribution.
Tissue Collection
For DOL 5-7 pups, stool/colon contents were homogenized at 1mg/100uL PBS, stomach contents in 200uL PBS, and small intestine in 100uL PBS, and serum was collected via cheek puncture. For DOL 0-1 pups, colon, stomach, and small intestine homogenized in their entirety in 100 uL PBS and serum was collected during euthanasia. For dam samples, stool was homogenized at 1mg/mL PBS, and mammary glands at approximately 0.2729 g/840uL PBS. The portion of mammary gland homogenized was slightly varied among samples, so all data is normalized by weight.
Tissue CFUs
The spleen and mesenteric lymph node were homogenized in 300uL PBS, then serially diluted, and plated for CFUs on LB plates. For stool plating, 0.1 mg of stool was homogenized in 300uL PBS, then serially diluted, and plated for CFUs on LB plates. 20 ug/ml nalidixic acid was added to MacConkey agar for selective growth of nalidixic acid resistant BSI E coli.
Donor Milk
Unpasteurized donor milk flagged for research use was obtained from the Minnesota Milk Bank for Babies.
Quantification of Donor Milk IgA
Total IgA was quantified using Uncoated ELISA Kit (Invitrogen). Assay was performed in triplicate. The optical density was found using a BioTek 800 TS absorbance reader at 450nm. Concentrations of IgA were reported as ng/mL based on an intraassay standard curve (provided by kit manufacturer).
Cross-reactive IgA was measured using the IgA Human Uncoated ELISA Kit (Invitrogen). For the coating step, bacteria were grown up to 0.3 OD in LB Broth, spun down and resuspended in PBS, then heat-killed (via heat block, for at least one hour) and added to the plate (100 uL a well). The rest of the experimental protocol followed the manufacturer’s instructions, data was reported as the absorbance units.
Statistical Analysis
All statistical analysis was done using GraphPad Prism 6.0 (GraphPad Software Inc.). Data is reported as mean +/− standard deviation. Differences between two groups was analyzed using one-way ANOVAs, with Sidak’s posttest. For the clinical cohort, continuous variables were compared using students T test and dichotomous variables were compared using Chi-Squared Analysis.
Data availability
Metagenomic 16S sequences will be deposited and available in the NIH GenBank SRA database.
Author Contributions
KM, KG, KL, and JM performed antibiotic animal experiments, flow cytometry, ELISAs, and data analysis. CM oversaw clinical cohort design, data acquisition and analysis. DO and AS performed detailed chart review of clinical cohort. KK performed 16S analysis. CM and KK wrote initial manuscript, all authors discussed and edited the manuscript and contributed to figure preparation.
References
