An Immunological Perspective on Neonatal Sepsis

Immune development during fetal ontogeny

Important progress has been made over the past few years in our understanding of hematopoiesis in humans. In comparison with adults or even infants born at term, premature infants are at a much higher risk of infection (Box 1). This knowledge helps to situate these premature newborns in terms of their immune development (Figure 1). About a quarter of all infants born below 32 weeks of gestation will develop a serious infection during the neonatal period [2]. This early period of prematurity also coincides with major immunological vulnerability. Indeed, infants at this stage of development are considerably immunocompromised because of the incomplete maturation of both innate and adaptive immune defense mechanisms (Box 2).

Open in a separate window

Figure 1

Immune Development during Gestation

The development of hematopoiesis begins early in embryonic life in the yolk sac, followed by the liver, bone marrow, and other secondary hematopoietic organs. Lymphopoiesis begins after erythropoiesis and myelopoiesis. The blue arrows illustrate hematopoietic cells and their progenitors (and precursors) migrating between hematopoietic organs at different stages of embryonic and fetal life. The diagram also represents the aorta–gonad–mesonephric (AGM) region which is an important structure supporting the early colonization of hema-topoietic organs by hematopoietic cells and their progenitors. Infants born very prematurely, between 22 (defining the absolute gestational limit of viability in humans) and 32 weeks, are at high risk of infection. This period also corresponds to a critical period where substantial maturation of antimicrobial PRRs occurs, beginning with endosomal/cytoplasmic, followed by extracellular PRRs. This period also likely overlaps with the beginning of the development of mature, adult-like fetal T cells from an earlier wave of fetal T cells. Concomitant to this is the passive immunization of fetuses through trans-placental maternal antibody transfer. All these events contribute to the immaturity of the neonatal immune system in the context of a preterm birth.

As illustrated in Figure 1, myelopoiesis begins early in the yolk sac around the 3rd week of gestation, followed by the liver after the 5th week. Until about the 20th to 22nd weeks the liver assumes most hematopoietic cell production, after which production of blood cells mainly takes place in the bone marrow [4]. The composition of myeloid cells has been well studied throughout gestation [5]. Neutrophils are very important in protecting newborns against infections. However, they form a very low proportion of blood cells until about 32 weeks [5]. Consequently, these cells have been the most difficult to study during this period. In preterm newborns, neutrophils lack the ability to form neutrophil extracellular traps, which are lattices of extracellular DNA, chromatin, and antibacterial proteins that mediate extracellular killing of microorganisms via the generation of potent reactive oxygen species [6].

In comparison to other myeloid cells, monocytes have been better studied, partly for historical reasons and partly because of their relatively high abundance in cord blood. Like other immune cell types, monocytes recognize pathogens through PRRs designed to recognize conserved microbial structures. These comprise endosomal, cytoplasmic, and extracellular antimicrobial detectors such as Toll-like receptors (TLRs), nucleotide-binding oligomerization domain-containing protein (NOD) and NOD-like receptors (NLRs), C-type lectin receptors (CLRs), and the retinoic acid-inducible gene I (RIG-I) and RIG-I-like receptors (RLRs) [7]. Three main monocyte subsets have been defined in adult blood, according to their expression of the CD14 and CD16 cell surface markers [8]. Before 29 weeks of gestation, a majority of fetal monocytes display an immature phenotype characterized by low CD14 expression [8]. These monocytes are likely to play a more predominant role in tissue remodeling than in aggressive immune responses [4]. Anti-microbial immune recognition and antigen presentation are also significantly impaired at this gestation even in “classical” high CD14-expressing monocytes, because of reduced receptor expression and intracellular signaling [9, 10]. In fact, below 20–24 weeks of gestation, our data suggest little activity in extracellular PRR although this gestational age group was not included in studies [11–15]. By comparison, PRR activity in mononuclear cells increases until about 33 weeks of gestation, with the earliest activity detected in endosomal (TLR7, 8 and 9) and intracellular (e.g. NLRs), followed by extracellular PRRs such as the TLR1, 2, 4 and 5 and the dectin-1 receptor [9, 16].

The significance of this “inside-out” hierarchical development of PRR function during fetal ontogeny is unclear [14]. However, it may important clinical implications. It corroborates well with some of the microbial vulnerabilities clinically observable in early premature infants. For example, activity in TLR2, which plays a predominant role in the recognition of the coagulase-negative staphylococci (CoNS), develops late in gestation [13]. Correspondingly, infections with this pathogen are most common below 30 weeks of gestation [17]. Knowledge of the maturation in activity of PRRs across stages of development is also important because of the adjuvant role activation of these receptors play during immunizations. Accordingly, premature infants do tend to display significantly reduced antibody responses to routine vaccines [18], a pattern that is also mirrored in the elderly [19]. Further, the complement system is also functionally impaired in neonates born prematurely, as evidenced in a decrease in complement proteins C3a, factor H, factor I [20], also limiting pathogen recognition and clearance.

Lymphopoiesis begins after the initiation of myelopoiesis during ontogeny. Production of B and T cells commences between 8 and 10 weeks of gestation in the bone marrow and thymus. Recently, the use of next-generation sequencing has allowed an unprecedented view of the structural diversity of the antigen receptors of lymphocytes within human individuals at the single-cell resolution level. The occurrence of somatic hypermutation, which is a sign of the finely-tuned cognate encounter of a lymphocyte with its antigen, indicates that these interactions do occur during the fetal period [21]. A recent and more in-depth phenotypic characterization of neonatal CD4+ T cells obtained from newborn cord blood also showed that up to 5–10% of cells are differentiated into either memory or effector cells, confirming their activation in utero [22]. The antigenic source of this T and B cell activation remains unclear, although this may not be entirely surprising given that several studies in humans indicate that microorganisms are naturally present well before birth within the intra-amniotic environment [23, 24].

During gestation, the placenta assumes a major dual role protecting the fetus against infection while also preventing immune activation directed towards maternal antigens [25]. While microbial organisms may be the source of T and B cell activation, it is also possible that some of these cells may have been triggered by maternal antigens [26]. In this context, one vital requirement for the fetus is to maintain tolerance not only to self antigens but also to the allogeneic maternal antigens, while retaining the ability to become activated upon encountering infections. Recent data have provided important insights into how this may be achieved. According to current models, the production of B and T cells is not linear during ontogeny, but instead occur in waves which likely originate from distinct hematopoietic stem cell progenitors [27, 28]. In mice, two functionally distinct populations of B cells (B1 and B2) have been described. Recent data in humans identified similar innate-like B1 cells that predominate in cord blood, spontaneously produce IgM antibodies, and express immunoglobulins targeting a narrower repertoire of antigens such as phosphorylcholine and DNA [29]. Like B cells, early fetal T cells also appear functionally distinct from the late-gestation, more resembling “adult-like” T cells in that the former are prone to tolerance [26]. During the second trimester of pregnancy, maternal cells can be detected in fetal lymph nodes [26]. Instead of provoking a classic allogeneic response from fetal T cells, these maternal cells induce the development of suppressive fetal regulatory T cells, as shown by evidence of tolerance induction against non-inherited maternal allo-antigens (NIMAs), and this tolerance can persist at least until early adulthood [26]. Using transgenic mice, researchers have been able to show a systemic accumulation of NIMA-specific regulatory T cells in female offspring. Furthermore, micro-chimerism resulting from the persistence of maternal cells in the fetus not only increased fetal tolerance but also improved reproductive fitness in pregnancies sired by males with shared NIMA specificity [30]. These data potentially highlight a mechanism ensuring a successful pregnancy, although it is unclear how faulty tolerance mechanisms might play a role in the initiation of preterm labor in humans.

The immune system of a full-term infant is thought to be well-adapted to its developmental period [19, 31]. Most recently, studies have demonstrated other unique functional features of the newborn immune system, potentially allowing age-specific interactions between innate and adaptive immune cells during the course of an infection. For example, neonatal immune cells can produce developmentally unique cytokine isoforms, and two examples have been reported in humans with IL-4 and CXCL8 (also known as interleukin-8 (IL-8) [32, 33]. In addition, a large proportion of human naïve neonatal T cells are able to produce the CXCL8 chemokine upon stimulation [34]. The production of CXCL8 by neonatal T cells is unique in that it occurs in absence of effector differentiation, specifically in CD31 T cells that have recently emerged from the thymus [35]. Consistent with an innate-like phenotype of these, production of CXCL8 is enhanced by co-activation of extracellular PRRs. CXCL8 is a powerful neutrophil chemoattractant and is a chemokine that has no direct counterpart in mice [36]. Given the important role of neutrophils in neonatal sepsis, production of this chemokine may play a crucial role during an early response to infection as suggested by data showing an accumulation of CXCL8-producing CD4+ T cells during sepsis or necrotizing enterocolitis (NEC) [34]. In mice, innate γδ T cells have been shown to activate neutrophils directly during infections through a production of large amounts of IL-17, an important cytokine which facilitates the development of mucosal immunity [37]. However, human neutrophils generally respond poorly to IL-17 [34]. Presumably, CXCL8-producing naïve T cells could provide immediate help by sanctioning the activation of neutrophils in humans, in a PRR-dependent manner during an infection. Production of CXCL8 by naïve T cells may be particularly important in newborns who lack a rapidly responding immunological memory to most pathogens at birth. Interestingly, CXCL8 may also have broader trophic mucosal effects in the human fetus as suggested by the high constitutive expression of this cytokine by intestinal epithelial cells (IEC) resulting in protective effect against TNF-α-mediated apoptotic cell death [38].

Development of Mucosal Immune Defenses

Colonization of the fetus marks the beginning of an extraordinarily complex lifelong relationship with microbes (Box 3). Mouse models have been extremely useful in understanding how micro-organisms at mucosal interfaces are essential in shaping immune functions during the post-natal period. However, the existence of major functional differences across species has warranted human studies to confirm the relevance of findings from mouse models. Detailed reviews on the development of mucosal immune defenses during ontogeny have been published elsewhere [39–41]. However, additional relevant studies performed in the context of neonatal sepsis have been undertaken and are discussed here.

The role of the intestine as a functionally distinct immunological organ is increasingly recognized [42]. Traditionally, our knowledge of the developing human intestinal immune system has been greatly complicated by tenuous immune phenotyping protocols coupled to a lack of accessibility to fetal samples. In the last couple of years, advances in the use of multi-parameter flow cytometry reagents and equipment, allowing more extensive, high-throughput characterization of cell populations using minuscule amount of biological samples obtainable from a human neonate has greatly facilitated discoveries in this area. For instance, invariant natural Killer (iNKT) cells are a relatively rare innate-like T cell type particularly important in protecting the host against colonizing micro-organisms at mucosal interfaces. These cells can be detected using fluorescence-labeled antigen-loaded tetramerized major histocompatibility complex molecules [43]. Recent data has shown that intestinal neonatal iNKT cells share unique functional properties in contrast to conventional T cells, intestinal neonatal iNKT cells produce robust Th1 and Th17 responses [44]. In humans, iNKT cells are abundant early on in fetal blood [45], although their tissue of origin had remained unclear until recent data showed that these cells accumulate in the small intestine during the second trimester of gestation [44].

Studies in mice have also revealed the crucial role of the small intestine in regulating both local and systemic Th17 responses [46]. Particularly, IL-17-mediated Th and γδ T cell responses appear very important to prevent invasion of micro-organisms at mucosal interfaces [47]. Earlier studies showed that neonatal CD4⁺ T cells are intrinsically biased towards a “default” Th2 response with an abundant, an epigenetically regulated production of IL-4 and IL-10 (reviewed in [48]). Recently, human neonatal CD4⁺ T cells were also shown to intrinsically differentiate poorly into IL-17-producing cells upon anti-CD3/CD28 activation in vitro [49, 50]. There are six homologs of IL-17 (IL-17A to F); IL-17A is the best characterized and predominant member of this family, whereas IL-17F also shares partly overlapping functions with IL-17A [51]. IL-17A/F-deficient mice exhibit increased susceptibility to a broad range of bacterial and fungal mucosal infections [51, 52]. In humans, the clinical impact of IL-17 deficiency is more limited, highlighting another important difference across species [47]. Loss of Th17 cells in humans with chronic HIV infection results in a loss of gut-barrier function and subsequent microbial translocation [53]. Notably, adult humans lacking production of IL-17 appear specifically vulnerable to neonatal pathogens such as Candida and Staphylococcus aureus [54]. To help promote systemic Th17 responses in the full-term neonate, neonatal antigen-presenting cells produce high levels of IL-1β, IL-6, but also high amounts of IL-23 [55]. By contrast, monocytes and dendritic cells from preterm neonates below 29 weeks of gestation produce low amounts of IL-1β and IL-23 [9, 15]. This lack of Th17-polarizing cytokine production by neonatal antigen presenting cells of infants born very prematurely may partly explain their increased vulnerability to mucosal infections caused by Staphylococcal and Candida species.

Increasing evidence suggest an important influence of the neonatal microbiome on health outcomes. In healthy term newborns, a diverse intestinal microbial flora promotes the development of mucosal barriers and prevents the outgrowth of pathogenic micro-organisms [56, 57]. Transient intestinal dysbiosis in the early post-natal period has recently been associated with an increased risk of asthma during early childhood [58]. Anomalies in the placental microbiome have also been associated with premature birth (reviewed in [59]). Moreover, accumulating evidence suggests that the composition of the microbiome plays an important role in protecting the newborn against infections. For example, disruption of the normal flora with broad spectrum antibiotics increases the risk of neonatal sepsis [60–62]. In preterm neonates, distinct microbial colonization of the gut linked to increased levels of CoNS and decreased colonization by protective commensals such as Bacteroides, Bifidobacterium, and Lactobacillus has also been associated to increased risks of neonatal sepsis [63, 64] and NEC [65, 66]. Data suggest that abnormal PRR signaling at mucosal surfaces in preterm infants may limit microbial diversity, thereby favoring the colonization of pathogenic micro-organisms [62, 64, 67]. These findings have direct clinical relevance. Indeed, administration of probiotics therapeutically reduces the risk of neonatal sepsis and NEC in premature infants [68]. Large clinical trials are also ongoing in developing countries; results so far indicate that administration of probiotics reduces neonatal mortality due to sepsis [69].

Most recently, other important observations with regards to the cellular composition [70] and presence of anti-microbial peptides [71]in breastmilk have been recently published. These data shed light on how the immune properties of human milk vary across periods of lactation and on the protective impact of breastfeeding on immune maturation [70, 71]. A main clinical relevance of these findings is that breastmilk can be used to protect against neonatal sepsis. For example, in the neonatal intensive care unit administration of human milk (as opposed to cow protein-based milk formula) has been repeatedly shown to protect against NEC in premature infants.

Mechanisms of Immune Cell Maturation and Function during Ontogeny

Despite major progress in the characterization of the neonatal immune system at various developmental stages, the fundamental mechanisms regulating the maturation of immune functions in humans during ontogeny remain completely unknown. Recent data has revealed an important immune regulatory role for erythrocytes. These authors suggested that CD71+ nucleated red blood cells (nRBCs) actively suppress both innate and adaptive immune responses through the expression of arginase, which inhibits nitric oxide synthesis [72]. However, this conclusion was disputed by others whose data suggested that the heightened intestinal immune activation in mice treated with anti-CD71 to deplete the nRBCs was unaffected by replenishment of these cells [73]. Of note, significant cross-contamination can occur between nRBCs and other white blood cells types during cell purification for ex vivo immune studies [74]. This type of technical problem therefore raises concern that previous conclusions from functional and genomic analyses of cord blood immune cells may have been biased because of the heterotopic cellular interactions with DNA/high rRNA-containing an potentially suppressing nRBCs that go undetected unless formally excluded by FACS [74].

Another recent contribution to our understanding of the regulation of immune responses has been the demonstration that a major metabolic switch is required in immune cells, in order to accommodate the massive induction of genes following an activation of these cells [75–78]. The essence of these changes is depicted in Figure 2. At rest, immune cells largely utilize oxidative phosphorylation to generate energy. Upon encountering a pathogen, cells turn off oxidative phosphorylation to rely almost exclusively on glycolysis for energy production; these metabolic changes are followed by activation of inflammatory immune pathways. Mitochondria play a central role in operating this metabolic switch through controlling the metabolic fate of glucose, as well as the utilization of fatty acids. By contrast, upon encountering an anti-inflammatory stimulus, enhanced lipid metabolism results in increase production of anti-inflammatory cytokines (IL-10, TGF-β), as well as pro-resolving lipid mediators (e.g. resolvins and protectins) [79].

Open in a separate window

Figure 2

Metabolic Adaptation Occurring during Immune Activation

Resting immune cells, such as monocytes or T cells primarily produce energy (ATP) from breaking down glucose into pyruvate, and through mitochondrial oxidative phosphorylation. The activation of pattern recognition receptors (PRRs) by microbes leads to the activation of the master transcription regulator NF-κB, which in turn results in inflammatory cytokine gene expression and production (e.g. IL-1β, TNF-α). Concurrently, cells also undergo a metabolic shift to aerobic glycolysis (a phenomenon also known as the Warburg effect), where energy production occurring through glycolysis breaks down glucose into succinate. Although less energy is produced for each glucose molecule, this pathway does provide large quantities of anabolic substrates necessary to feed into the high gene transcriptional activity occurring during immune activation. Succinate and NF-κB also stabilize the transcription factor HIF-1α, resulting in further upregulation of inflammatory cytokine gene expression. The shift away from oxidative phosphorylation also allows these cells to produce reactive oxygen species (ROS) more efficiently. By contrast, anti-inflammatory signaling (via STAT6) causes an increase in fatty acid uptake, mitochondrial biogenesis, and increased mitochondrial function. This metabolic shift towards fatty acid oxidation leads to upregulation of anti-inflammatory cytokines (e.g., IL-10, TGF-β) and pro-resolving lipid mediators such as resolvins and protectins.

Metabolic control over the pro or anti-inflammatory nature of an immune response appears to be a general requirement of immune cells, as suggested by similar changes observed during T cell [80], monocyte/macrophage [81] and dendritic cells activation [82]. Corresponding metabolic shifts to glycolysis in pro-inflammatory vs fatty acid oxidation in anti-inflammatory responses have been observed in both human and mouse cells [76, 77, 83]. As another example, activation of effector CD4⁺ T cells requires in a shift from oxidative phosphorylation to glycolysis while activation of regulatory T cells results in a shift in cell energy metabolism toward fatty acid oxidation [80]. The broad and conserved extent of these changes and the unique metabolic state of the fetus suggests that a lack of similar metabolic changes may contribute to immune impairments in preterm neonates. This hypothesis is supported by several primary metabolic differences that have been described between neonates born prematurely, at term, and adults. For instance, during fetal life, mitochondria have a fragmented appearance [84], reduced mass [85], and display a more limited capacity for energy production [86]. In addition, preterm neonates exhibit increased lipid metabolism relative to their glycolytic activity [87]; this is linked to an altered metabolic response during sepsis [88]. Also, oxidative phosphorylation pathways are down-regulated in preterm neonates [86, 89, 90], which compensate by increasing fatty acid oxidation [91, 92]. Hypoxia-Inducible Factor 1-alpha (HIF-1α), and mechanistic Target of Rapamycin (mTOR) have been shown to play an important role in regulating this metabolic switch [75]. HIF-1α is a transcription factor largely responsible for the switch from oxidative phosphorylation to glycolytic pathways; in addition, HIF-1α activity can be induced by inflammatory stimuli through via mTOR signaling [75]. Whereas the role of HIF-1α and mTOR has been relatively well defined in mature organisms, little is known about the regulation of these molecules during ontogeny. The fetus naturally evolves in a relatively hypoxic environment that is due to its unique blood circulation. Regulation through HIF-1α is influenced by a variety of stimuli, including inflammatory signaling [93] and oxygen levels [94], thereby providing a potential link between low oxygen concentration in utero and decreased immune responsiveness [78].

Another important question is whether observations of attenuated immune responses made in preterm neonates at birth can be extrapolated to explain the high risk of infection observed during the neonatal period. Indeed, our knowledge of the developing newborn immune system has been largely derived from cord blood. Nonetheless, recent studies have examined the maturation of immune cells during the neonatal period, at the time infants are most affected by neonatal pathogens (i.e. around 1–3 weeks of post-natal age). Indeed, using both whole blood and single-cell flow cytometry analyses, Marchant et al. have confirmed that PRR responses (e.g. TLR4, TLR7/8), remain attenuated in post-natal blood obtained from neonates born below 30 weeks of gestation [95]. These results provide important support to the premise that reduced immune activity is responsible for the high vulnerability to infections.

Premature infants are often born in pathological conditions such as following chorioamnionitis, which may potentially impact on functional measurements of immune responses in cord blood (Box 4). A vast majority of these infants are also exposed to corticosteroids (an immunosuppressive drug), which are often administered to pregnant mothers in order to promote fetal lung maturation. However, multivariable statistical analyses of immune responses in these infants following rigorous clinical definitions of antenatal exposures suggest that the functional immunological differences measured from cord blood are probably developmentally related, rather than due to clinical exogenous perinatal factors linked to their premature birth [14].

Lessons from Neonatal Mouse Models

Studying the role of immune components in sepsis experimentally in newborn infants is virtually impossible for obvious ethical reasons. The miniaturization of immunological methods has greatly helped the development of neonatal mouse models of sepsis. These models have confirmed the limited myelopoietic capacity of the newborn and its impact on survival of newborn mice during sepsis [96]. Using a poly-microbial infection model, Wynn and Moldawer demonstrated that rodent neonates rely heavily on innate immune defense during sepsis [97]. Moreover, the brain of a developing newborn is also particularly vulnerable to inflammation, and neonatal infections carry an important risk of major long-term neurological disability in premature infants [98]. To study the mechanisms of brain injury during neonatal infections, Levy and Mallard developed an interesting mouse model [99] in which they demonstrated that inflammation due to Staphylococcus epidermidis caused brain injury even in the absence of bacterial entry into the cerebrospinal fluid space. These results suggest that low-grade inflammation produced during transient bacteremia could be sufficient to cause damage to a developing neonatal brain. The findings have important clinical implications, since earlier studies have indicated that transient bacteremia commonly occurs in infants with sepsis [100, 101]. Another study has demonstrated persistent sub-clinical inflammation in the absence of overt clinical signs of infection in these infants throughout the neonatal period, possibly due to an ongoing exposure to intensive care therapies [102]. In light of these findings, better strategies will be necessary to both limit the sustained exposure of premature infants to any potential triggers of inflammation or sources of transient bacteremia, and to prevent infections [103].

The coupling of neonatal mouse models to genome-wide transcriptome studies has also pointed to important functional differences between neonatal and adult responses [104]. Indeed, the nature of innate immune PRR pathways involved during sepsis differs between neonates and adults, highlighting the importance of age-appropriate models [96]. However, major differences in the nature of the transcriptome responses depend on the model used [105], and well-recognized differences between humans and mice may limit the application of findings across species. This justifies direct observations in humans [106]. Most recently, two important genome-wide gene expression studies have provided insights into the molecular pathways activated in newborns with severe infections [88, 107]. For instance, on the one hand marked transcriptional alterations in metabolic pathways associated with glucose transport, glycolysis, and cholesterol homeostasis, transport, and metabolism have been detected in newborns during active sepsis [88]. On the other hand, transcriptional pathways related to antigen presentation processes via MHC II were down-regulated in septic infants [88]. The specific up-regulation of immune suppressor signaling pathways, such as TNFAIP3, CD71, and SOCS1/3 in infants with sepsis and which have been reported might help to explain why some infants fail to clear infections, although it is possible that the lower gestational age of infants from the septic group in this study may have confounded some of these results [88]. The main clinical importance of these results is that they provide proof-of-concept validation that measuring genome-wide host immune responses can greatly help discriminate infants with or without bacterial infection more accurately than traditional clinical laboratory markers. These data should drive the development of a new generation of diagnostic tests that will inform clinical decision-making regarding antibiotic treatment in neonatal sepsis in a timely and pathogen-oriented manner, thereby reducing missed diagnoses and unnecessary interventions.

We are grateful to our colleague Dr Tobias Kollmann for providing insightful comments on this manuscript. H.R.R. is supported by a Mitacs Accelerate Scholarship, in partnership with STEMCELL Technologies Inc. and the British Columbia Children’s Hospital Foundation. P.M.L. is supported by a Career Investigator Award from the Michael Smith Foundation for Health Research, and a Clinician-Scientist Award from the Child and Family Research Institute. We thank Kelsey Lee for help in revising this manuscript. Part of the research presented in this article has been funded by the Hospital for Sick Children, the Canadian Institutes of Health Research (operating grants MOP-110938; MOP-123478), and the US March of Dimes Foundation.