Impaired Cytokine Responses to Live Staphylococcus epidermidis in Preterm Infants Precede Gram-positive, Late-onset Sepsis

Abstract

Background

Late-onset sepsis (LOS) with Staphylococcus epidermidis is common in preterm infants, but the immunological mechanisms underlying heightened susceptibility are poorly understood. Our aim is to characterize the ontogeny of cytokine responses to live S. epidermidis in preterm infants with and without subsequent Gram-positive LOS.

Methods

We conducted a prospective, observational cohort study of preterm infants (<30 weeks gestational age [GA]) with blood sampling on Days 1, 7, 14, 21, and 28 of life. Cytokine responses in peripheral whole blood stimulated with live S. epidermidis were analyzed by 11-plex immunoassay.

Results

Of 129 infants (mean GA, 26.2 weeks; mean birth weight, 887g), 23 (17.8%) had confirmed LOS with Gram-positive organisms and 15 (11.6%) had clinical sepsis, with median onsets at 13 and 15 days, respectively. Blood cytokine responses to an in vitro S. epidermidis challenge were similar between infected and uninfected infants on Day 1, but diverged thereafter. Infants with subsequent LOS displayed broadly reduced S. epidermidis–induced responses from Day 7 onwards, compared to those who did not develop LOS. This pattern was observed with chemokines (interleukin [IL]-8, monocyte chemotactic protein–1, and macrophage inflammatory protein–1α), pro-inflammatory cytokines (IL-1, IL-6, and tumor necrosis factor–α) and the regulatory cytokine IL-10.

Conclusions

Cytokine responses to a live S. epidermidis challenge are impaired in infants with LOS and precede the onset of clinical illness. Quantifying pathogen-specific cytokine responses at Day 7 may identify those high-risk preterm infants at the greatest risk of LOS, and prospective replication is warranted.

In high-income countries, 8–13% of deliveries occur preterm, and infants born before 30 weeks gestational age (GA) contribute disproportionately to neonatal morbidity and mortality [1]. These infants are at high risk of neonatal sepsis and remain susceptible to severe infection into adolescence [2]. Late-onset sepsis (LOS) is a frequent complication of prematurity, affecting 25–50% of the most preterm infants [3]. The sepsis incidence is highest in the second postnatal week, predominantly with coagulase-negative staphylococci (particularly Staphylococcus epidermidis [SE]), the causative organisms in 75–80% of LOS in many countries [4, 5]. In this high-risk population, LOS is associated with increased risks of adverse neonatal and developmental outcomes, as well as substantial healthcare and societal costs [6].

The immunological mechanisms underlying the susceptibility of preterm infants to LOS are incompletely understood. To date, it is not possible to identify those infants at the highest infection risk prior to the onset of sepsis, yet they may benefit most from prophylactic immunological interventions. Furthermore, most analyses of preterm infant immune functions are of cord blood, which is readily available in relatively large volumes. Extrapolating findings from cord blood to estimate functional capacity at the age of the highest LOS incidence (10–14 days) is problematic, as immune responses change markedly in the immediate postnatal period and cord blood responses may not be broadly representative of those in postnatal peripheral blood [7–9].

Here, we utilize unique serial, peripheral blood samples from a prospective cohort of very preterm infants with and without Gram-positive LOS during the first month of life to characterize the cytokine responsiveness to live SE. We hypothesized that infants who subsequently developed LOS would show differential immune responses prior to clinical sepsis, compared to those who did not develop LOS.

PATIENTS AND METHODS

Blood Sampling and Preparation

This prospective, observational cohort study was approved by the institutional ethics committee at King Edward Memorial Hospital, Perth (814/EW), and written informed consent was obtained from parents. We recruited 129 infants born at <30 weeks GA without major congenital abnormalities (Table 1).

Infant peripheral blood was collected either from existing umbilical catheters (Day 1 only) or by venipuncture on Days 1, 7, 14, 21, and 28 into lithium-heparin tubes and was processed within 4 hours.

Confirmed LOS (>72h of age) was defined as the combination of a positive blood culture and a maximum C-reactive protein (CRP) within 48 hours of >3 times the limit of normal in our laboratory: that is, >15 mg/L (n = 23). Clinical LOS was defined as a negative blood culture and a maximum CRP of >15 mg/L (n = 15). Positive blood cultures associated with a CRP <15mg/L were classified as contaminants (n = 1 each for coagulase-negative staphylococcus, SE, and Enterococcus faecalis) and analyzed with the non-LOS group. Infants were assigned an overall LOS classification based on the most severe outcome of all LOS episodes within the first month of life.

Live Staphylococcus epidermidis Preparation

An SE isolate (wild-type strain 1457) was originally isolated from a patient with an infected central venous catheter and was kindly supplied by Dr Michael Otto (National Institutes of Health, Bethesda, MD). Stocks were grown to mid-log phase (optical density [OD] 600 nm: 0.7–0.8) in heart infusion broth (Oxoid, SA, Australia) and stored frozen in tryptone soya broth containing 15% (v/v) glycerol. Bacteria were streaked from the frozen tryptone soya broth glycerol stock onto sheep blood agar plates and incubated overnight at 37°C. The following day, 3 separate colonies were extracted from the blood agar plate and resuspended in phosphate-buffered saline (PBS; Gibco by Life Technologies, Victoria [VIC], Australia). The bacterial suspension was counted using a Helber Counting Chamber and the stock concentration was adjusted to a final concentration of 2.5 x 10⁷ colony-forming units/mL with PBS, and stored on ice until use.

Whole Blood Stimulation with Live Staphylococcus epidermidis

25 μL of peripheral whole blood was cultured in 75 μl Roswell Park Memorial Institute (RPMI) 1640 L-alanine-L-glutamine (GlutaMAX), which was supplemented with 0.01 mol (M) 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), 1 mM sodium pyruvate, and 5.5 μM beta-mercaptoethanol (Gibco Life Technologies, VIC, Australia), for 24 hours at 37°C and with 5% CO₂. Blood cultures were challenged with either 1 x 10⁶ colony-forming units/mL of live SE (based on empirical optimization studies in adult and term infant cord blood [10]) or an equal volume of PBS alone. Following the overnight incubation, an additional 100 μl of supplemented RPMI 1640 GlutaMAX with the addition of 5% (w/v) fetal bovine serum (SAFC Biosciences) which was added and supernatants were harvested and stored at −80°C.

Multiplex Assay

Cytokine and chemokine production was quantified using a validated, in-house, multiplex, bead-based assay [11]. Primary antibodies against interleukin [IL]-1β, IL-12p70, macrophage inflammatory protein–1α (MIP-1α), monocyte chemotactic protein–1 (MCP-1), triggering receptor expressed on myeloid cells–1 (TREM-1; R&D Systems), IL-6, IL-8, IL-10, IL-13, IP-10, and tumor necrosis factor–α (TNF-α; Becton Dickinson [BD] Biosciences, New South Wales [NSW], Australia) were covalently conjugated to carboxylated microspheres (Luminex Corporation, Austin, TX). Supernatants were diluted in PBS with 0.05% (v/v) Tween20 and 2% (v/v) newborn bovine serum (all from Sigma-Aldrich, NSW, Australia). Microspheres, at a concentration of 3500 per bead region, and diluted samples were transferred to manifold vacuum Multiscreen plates (Merck-Millipore, VIC, Australia) and incubated at room temperature for 30 minutes on an orbital shaker (500 revolutions per minute [rpm]), protected from light. Biotinylated secondary antibodies were then added, and the plate was incubated for another 30 minutes at room temperature (500 rpm, in the dark). Wells were washed with PBS containing 1% (w/v) bovine serum albumin, 0.25% (v/v) Tween20, and 0.001% (w/v) sodium azide (NaN₃) vacuum filtration before a streptavidin-phycoerythrin (PE) conjugate (BD Biosciences, NSW, Australia) was added for 15 minutes with shaking. Wells were washed twice as above, and the fluorescence in each bead region was measured on a BioPlex 200 System (Bio-Rad, NSW, Australia). Data were acquired electronically in real time and analyzed using BioPlex Manager 5.0 software. Data in pg/mL were generated from a 7-point, 4- or 5-parameter logistic standard curve. All values below the lowest standard were assigned an arbitrary cut-off value of half the lowest standard for analysis.

Immunostaining of Whole Blood for Flow Cytometry

Using a validated flow cytometry protocol for extended differential white cell counts [12], 25 μL of whole blood was surface-stained with the following monoclonal antibodies: PE-conjugated cluster of differentiation (CD)36 (CD36-PE; clone CD38, mouse antihuman), allophycocyanin (APC)-conjugated CD2 (CD2-APC; clone S5.2⁹, mouse), Alexa Fluor 647–conjugated CD294 (AF647-CD294; clone BM16, rat antihuman), APC-H7–conjugated CD16 (APC-H7-CD16; clone 3G8, mouse antihuman), BD Horizon V450–conjugated CD19 (V450-CD19; clone HIB19, mouse antihuman), and AmCyan-conjugated CD45 (AmCyan-CD45; clone 2D15, mouse antihuman) from BD Biosciences (NSW, Australia). Following 15 minutes of incubation, red blood cells were lysed with 1 mL BD fluorescence activated cell sorter [FACS] Lysing Solution (BD Biosciences) for 10 minutes in the dark at room temperature. Cells were then washed with FACS buffer (PBS, 2% v/v fetal calf serum, 2% w/v bovine serum albumin, 0.01% w/v sodium azide) and resuspended in BD Stabilizing Fixative (BD Biosciences), with 300 μL transferred to BD TruCOUNT Tubes (BD Biosciences) to allow for absolute counts of each cell type, and stored at 4°C in the dark prior to analysis. Spectral compensation for each antibody was achieved using BD CompBeads (BD Biosciences).

Flow Cytometry Analysis

Cells were analyzed using a BD FACSCanto II flow cytometer and BD FACSDiva software (BD Biosciences). Prior to acquisition, a flow cytometer calibration was performed using SPHERO Ultra Rainbow Calibration beads (Spherotech, IL), in addition to BD Cytometer Setup and Tracking beads (BD Biosciences), to monitor the cytometer performance. Postacquisition, a cell population analysis was performed using FlowJo software (Tree Star Inc.). Various leukocyte populations were identified using a validated gating strategy [12], identifying the following cell types from the white blood cell (CD45⁺) population: neutrophils (side scatter [SSC]_high/CD16⁺), inflammatory monocytes (SSC_high/CD16⁺→CD2&CD294^-/CD36⁺), immature granulocytes (SSC_low/CD16^-), non-cytotoxic T cells (SSC_low/CD16^-→CD2&CD294⁺/SSC_low), classic monocytes (SSC_low/CD16^- → CD2&CD294^- → CD19^-/CD36⁺), and B cells (SSC_low/CD16^- → CD2&CD294^- → CD19⁺/ SSC_low).

Data and Statistical Considerations

Continuous data were summarized with medians, interquartile ranges (IQR), and ranges and categorical data with frequency distributions. Univariate comparisons between groups (no LOS, confirmed LOS, and clinical LOS) were made using the Kruskal-Wallis test for continuous outcomes and the Chi-square or exact tests for categorical outcomes. The duration of mechanical ventilation was summarized using Kaplan-Meier survival estimates, with deaths censored in the analysis, and compared between sepsis groups using the log-rank test. Cytokine data were transformed to the natural logarithm for analysis to fulfil the model assumption of residual normality. Mixed linear regression analysis using a random effects model and a group-by-time interaction was performed to assess the effect of LOS on cytokine levels, while accounting for the repeated measures at 1, 7, 14, 21, and 28 days. All models were adjusted for gestational age, birth-weight z-score, sex, and the total counts of cell types relevant to each cytokine. Absolute counts were calculated by dividing the number of events in the population of interest by the total bead count and multiplying by the number of beads per test, divided by the sample volume with dilutions taken into account (per BD Trucount instructions).

Unstimulated cytokine counts were assessed in all models. Comparisons of adjusted means between groups (no LOS vs confirmed LOS and no LOS vs clinical LOS) at each time point were corrected for multiple testing using the Bonferroni method to maintain an overall alpha error rate of 0.05. Parameters were back-transformed to the original scale of measurement and presented as geometric means and 95% confidence intervals.

Statistical Package for Social Sciences (SPSS) (version 22.0, IBM SPSS, Armonk, NY) and Stata (version 12, StataCorp, College Station, TX) statistical software were used for the data analysis. Any P values <.05 were considered statistically significant.

RESULTS

Clinical Characteristics of Study Cohort

A total of 129 very preterm infants were recruited (Table 1). Infants diagnosed with early-onset sepsis (n = 4), Gram-negative LOS (n = 5), or necrotizing enterocolitis (n = 5) during the study period were excluded to reduce confounding. There were 3 infant deaths: 2 from pulmonary hemorrhage and 1 from complications of extreme prematurity. Data obtained prior to death were included in the analysis.

The final analysis included 23 infants with Gram-positive LOS. Of these, 19 (82%) had LOS with coagulase-negative staphylococci; other Gram-positive LOS pathogens were Staphylococcus aureus, Bacillus sphaericus, Streptococcus mitis, and Enterococcus faecalis. The median age at the onset of a confirmed, Gram-positive sepsis diagnosis was Postnatal Day 13.2 (IQR, 9–16). Of the infants with confirmed LOS, 19 infants had 1 confirmed LOS episode and 1 infants had 3 confirmed episodes, while 3 infants had an episode of clinical LOS prior to an episode of confirmed LOS. A further 15 infants developed clinical LOS alone (median age at onset, Postnatal Day 15.3; IQR, 10–20). Of the infants with clinical LOS, 12 infants had 1 episode and 3 infants had 2 episodes of clinical LOS. During the study period, 77 infants remained free of confirmed or clinical LOS.

The infants with and without LOS were similar in most clinical characteristics; however, consistent with known LOS risk factors, infants with confirmed or clinical LOS were more preterm and were smaller at birth than those without LOS (Table 1).

Cytokine Responses to In Vitro Stimulation with Live Staphylococcus epidermidis

Uninfected preterm infants’ cytokine responses to an in vitro challenge with live SE increased rapidly during the first month of life. In contrast, infants who developed clinical or confirmed LOS (ie, any LOS) produced significantly lower levels of the majority of cytokines and chemokines (IL-1β, IL-6, IL-8, TNF-α, MIP-1α, and MCP-1) on Days 7, 14, and/or 21 and, importantly, this impairment was evident prior to the onset of sepsis (Figures 1–3). Further, these findings remained evident following adjustments for GA, birth-weight z-score, sex, and absolute numbers of specific leukocyte subsets (major cell types for each cytokine: [IL 1β, IL-6, IL-12p70, TNF-α, MIP-1α, MCP-1: CD16+ and CD16- monocytes]; [IL-8: neutrophils and immature granulocytes]; [IL-10, IL-13 and IP10: noncytotoxic T-cells and B-cells]; [TREM1: neutrophils]). Levels of IL-10 were lower in the LOS group on Day 7 only, while concentrations of IL-12p70 and TREM1 were lower in the LOS group on Day 14 only.

Comparing infants with confirmed LOS to infants without LOS, and using the same adjustments, we found similar evidence of an impaired capacity to produce several cytokines in infants with confirmed LOS: IL-10 levels were reduced on Day 7. In contrast, IL-1β, IL-6, IL-8, TNF-α, MIP1α, and MCP-1 levels were lower on both Days 14 and 21, while IL-12p70 and TREM1 were lower only on Day 14 (Figures 1–3). The cytokine levels in the clinical-LOS group tended to track between the no-LOS and confirmed-LOS groups, with levels lower for IL-8, MCP-1, and MIP-1α on Day 7.

Next, we analyzed commonly utilized ratios of cytokines (IL-6/TNF-α, IP-10/IL-6, and TNF-α/IL-10) to determine whether this approach may highlight further differences between groups. The results, adjusted for GA, birth-weight z-score, sex, and relevant leukocyte subset counts, showed that compared to infants without LOS, infants with confirmed LOS had higher IP-10/IL-6 ratios on Days 14 and 21, whereas no differences were found between clinical-LOS and uninfected infants. Infants with any LOS had increased IP-10/IL-6 ratios on Days 7 and 14 and a reduced TNF-α/IL-10 ratio on Day 14 (Supplementary Figure 1).

Although a priori adjustments were made for GA, birth-weight z-score, and sex in all cytokine analyses, sex was not associated with any cytokines and only 1 significant association (negative) was observed, between birth-weight z-score and MCP-1. Gestational age was positively associated with all inflammatory cytokines (except TREM) and with IL-8 and MIP-1α. As expected, meaningful associations were found between cell counts of specific leukocyte subsets and cytokine levels when relating absolute numbers of those leukocytes known to be principle producers of these cytokines. For every 10% increase in respective leukocyte numbers, there was a corresponding 1.5–1.7% increase in all inflammatory cytokine levels (except TNF), a 3.6% increase in the IL-13 level, and a corresponding 0.9% decrease in the IL-8 level (Supplementary Table 1).

DISCUSSION

To our knowledge, this is the first description of longitudinal peripheral blood cytokine responses to the most common neonatal pathogen, S. epidermidis, in very preterm infants in the first month of life and prior to the onset of LOS. In addition to analyzing fresh peripheral whole blood from a large cohort of prospectively recruited infants, we used live pathogen stimulation to better reflect in vivo parameters.

Our results demonstrate rapid development of the capacity of very preterm peripheral blood leukocytes to respond to live SE and are in line with the rapid postnatal immune maturation reported recently by Lee et al [9]. In uninfected preterm infants, this may reflect the physiological maturation induced by the immediate postnatal colonization with microbial organisms, especially SE, and suggest an important survival mechanism [13]. In contrast, infants who developed Gram-positive LOS displayed a distinct trajectory over the same period, with a markedly reduced capacity to produce pro-inflammatory and regulatory cytokines and chemokines in response to SE. Importantly, this impairment was evident on Postnatal Day 7 (but not on Day 1) and prior to the onset of LOS (mean onset, ~Day 14). In addition, this impairment was not correlated with leukocyte numbers, was persistent, and became more pronounced following the LOS episode, consistent with the concept of transient postsepsis immune paralysis [14–16].

Similar, albeit smaller differences, were observed between uninfected infants and those with clinical LOS, in whom blood cultures were sterile but a clinical diagnosis of sepsis was accompanied by increased inflammatory markers. This is consistent with the less stringent definition that likely encompasses heterogeneous pathophysiologies, including episodes of LOS with false-negative blood cultures (likely due to inadequate blood volumes), viral infection, and noninfectious inflammatory conditions [17, 18].

Our analyses were adjusted for known relevant clinical determinants of sepsis risk, such as GA, birth-weight z-score, and sex, and confirm the dominant role of GA on cytokine responsiveness. We also assessed whether commonly used cytokine ratios might reveal further differences between the groups, but no differences were evident.

The mechanisms that shape early life immune responses in high-risk preterm infants are unknown, but likely include the establishment of the cutaneous and intestinal microbiome [19]. The composition of the early microbiome is highly dynamic and readily perturbed by common neonatal intensive care interventions, such as antibiotic therapy; the timing of the initiation, progression, and type of enteral feeding; probiotic supplementation; skin care; and exposure to the neonatal intensive care unit environment. Adjustment for all these factors necessitates a far larger sample size and is beyond the scope of this project. Notwithstanding, we adjusted cytokine levels for principal leukocyte subsets by flow cytometry, which had a limited effect, suggesting that the majority of the variances observed are due to intrinsic, pathogen-specific, functional differences rather than overall leukocytes numbers, consistent with our previous investigations [20, 21].

Our study has a number of strengths, including the prospective, longitudinal design; a reasonable sample size given the repeated sampling in a very preterm population; the real-time processing of fresh peripheral blood; the use of the most common live neonatal pathogen; and an 11-plex cytokine assay combined with the flow cytometry determination of principal leukocyte subsets [10, 22]. However, we acknowledge some unavoidable limitations inherent to research in very preterm infants; particularly, there are ethical and clinical constraints on the frequency and volume of blood sampling, which results in inoculation of small blood volumes for culture and potential false-negative results in the setting of low colony count bacteremia. We therefore employed the best routinely available evidence for categorization: the presence of clinical signs, blood culture positivity, CRP levels, and the duration of antibiotic use. The overall sample size did not allow us to specifically characterize the comparatively rare (in our setting) Gram-negative LOS episodes. Further, the limited blood volumes did not permit the determination of all cytokines of interest, but rather a subset of predefined cytokines.

CONCLUSIONS

This study adds to our knowledge of the rapid changes in early life immune ontogeny in very preterm infants. Our study reveals that functional differences in the immune response capacity are evident prior to the clinical onset of LOS, and this approach may have potential for predictive markers following the replication of the findings and further refinement. A better understanding of the susceptibility mechanisms may facilitate the monitoring of those infants at the highest risk, potentially allowing for targeted prophylactic interventions and earlier diagnoses: an important step towards “precision neonatology.”

Supplementary Data

Supplementary materials are available at Clinical Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.

Notes

Disclaimer. The funders had no role in the study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Financial support. This work was supported by the National Health and Medical Research Council of Australia (NHMRC) (grant number 572548); the Department of Health, Western Australia; and Telethon Channel 7 Trust, Western Australia. Research at Murdoch Children’s Research Institute is supported by Victorian Government’s Operational Infrastructure Support Program.

Potential conflicts of interest. D. B. is supported by a NHMRC Senior Research Fellowship (grant number 1064629) and an Investigator Grant (grant number 1175744). T. S. is supported by a Raine Foundation and Western Australia Department of Health Clinician Research Fellowship. All other authors report no potential conflicts. All authors have submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest. Conflicts that the editors consider relevant to the content of the manuscript have been disclosed.

References

Author notes