Redefining the Role of Intestinal Microbes in the Pathogenesis of Necrotizing Enterocolitis

Clinical Evidence

More than 30 years ago, in 2 classic articles, Sántulli et al^6,7 identified bacterial colonization as 1 of 3 critical factors (in addition to mucosal injury and initiation of enteral feedings) that predispose newborn infants to NEC. This conclusion was based on experience caring for >100 infants with the disease at the Infants' Hospital in New York between 1955 and 1979. Since that time, evidence supporting the notion that NEC is mediated, at least in part, by intestinal microbes has continued to mount. This evidence includes the common findings of bacteremia^8,–,10 and endotoxinemia^11,12 in affected neonates and the pathognomonic imaging finding of pneumatosis intestinalis, which likely represents submucosal gas produced by bacterial fermentation.¹³ Moreover, occasional “clusters” of cases linked to causative pathogens have been reported, although the identity of the responsible organisms has varied among outbreaks and across institutions.¹⁴ The efficacy of broad-spectrum antibiotic therapy in many cases of NEC and the apparent efficacy of probiotic administration in some clinical trials¹⁵ have provided further indirect evidence that manipulation of the intestinal microbial community can affect outcomes in infants with NEC.

Given the strength of the evidence (albeit often circumstantial) that microbes contribute to the pathogenesis of NEC, clinicians and researchers have been searching for more than 3 decades to identify organisms that are causally linked with the disease. Until recently, such efforts have been restricted to culture-based investigations of blood, peritoneal fluid, or intestinal contents of patients with NEC or control populations. The results of these investigations have been highly variable and, taken as a whole, inconclusive. Three fascinating reports in The Lancet 30 years ago demonstrated the presence of clostridial species in samples from infants in NEC outbreaks.^16,–,18 Since that time, results of several other investigations have supported a role for clostridial species (including Clostridium perfringens, Clostridium butyricum, and Clostridium neonatale) in NEC pathogenesis.^19,–,21 It has been suggested (but never proven) that the gas-forming ability of clostridia accounts for pneumatosis intestinalis in affected patients.²² Results of important work from Kosloske and Ulrich²³ indicated that the disease is generally more severe when C perfringens can be cultured from the blood or peritoneal fluid of an affected patient. Nevertheless, others have convincingly demonstrated that the prevalence of colonization with clostridial species (including Clostridium difficile) does not differ between affected infants and healthy controls.^24,–,27

In general, with respect to both clostridial and nonclostridial organisms, microbial species identified in cultures from patients with NEC have not been significantly different from species generally considered “normal” within the gastrointestinal tract of hospitalized patients. A partial list of organisms associated in various reports with NEC (Table 1) includes Klebsiella pneumoniae, Escherichia coli, Enterobacter spp, Pseudomonas spp, C difficile, and Staphylococcus epidermidis.^{22,24,28,–,30} Because these organisms are also commonly found among patients without NEC, none of them fulfill Koch's postulates³¹ that define causal relationships between microbes and disease pathogenesis. The inconsistency of these findings over several decades strongly suggests that no single causative pathogen is responsible for NEC pathogenesis. However, it is still possible that causative organisms do actually exist and that insufficiencies in present cultivation techniques have prevented their proper identification.

Although far more attention has been paid recently to intestinal bacteria, the distinct possibility remains that enteric viruses play a critical role in NEC pathogenesis. Much less is known about the normal viral load or potential viral pathogens in the newborn intestine, partly because viruses are more difficult to study than bacteria. However, outbreaks of the disease over the past 4 decades have provided unique opportunities to make associations between NEC and viruses. More than 2 decades ago, a series of remarkable studies demonstrated the presence of a novel coronavirus within fecal samples and resected intestinal segments from infants with NEC.^32,33 Subsequently, results of other studies that used a variety of experimental approaches have linked the disease at various times and locations to other viruses including coxsackie B2 virus, rotavirus, adenovirus, torovirus, astrovirus, and echovirus 22.^34,–,38 Few if any studies have been conducted to systematically evaluate the contribution of viruses to NEC pathogenesis, partly because of the inherent difficulties in studying viruses in clinical settings. Thus, the role of viruses in NEC pathogenesis remains uncertain at present.

Laboratory Evidence

Extensive work with animal models of NEC has repeatedly confirmed the importance of intestinal bacteria in disease pathogenesis. Barlow et al used a newborn rat model to demonstrate that animals with NEC had a marked overgrowth of Gram-negative organisms in the stool and blood relative to control animals without disease.³⁹ In 1986, Kosloske and co-workers⁴⁰ compared NEC development in germ-free and wild-type newborn rats and determined bacteria to be the most critical factor to determine disease pathogenesis. In 2006, Jilling et al⁴¹ reported that sanitization of feeding catheters between enteral feedings markedly diminished the incidence of disease in their well-validated rodent models of NEC. In numerous additional studies, utilizing quail⁴² and piglet models^43,44 and the more common rodent models,^45,46 the importance of bacteria in NEC pathogenesis has been confirmed. Collectively, these studies defined gut bacteria as necessary for disease, but they did not clarify how and why they are necessary.

More recently, laboratory investigations of NEC have begun to elucidate putative molecular mechanisms by which gut microbes contribute to the development of disease. The most promising advances thus far relate to the importance of Toll-like receptors (TLRs) in NEC pathophysiology. TLRs are integral glycoproteins that allow cells of the innate immune system to recognize highly conserved molecules derived from microbes such as lipopolysaccharides, bacterial flagellin proteins, or microbial nucleic acids.⁴⁷ In recent years, a pair of important publications detailed the relevance of TLR4 activation in NEC pathogenesis. Jilling et al⁴¹ elegantly demonstrated that enterocyte expression of TLR4 increases after birth in animals with risk factors for NEC (ie, formula feeding and environmental stress), whereas TLR4 expression in enterocytes from healthy mother-fed controls gradually decreases during the first 3 days of life. It is important to note that they also demonstrated that TLR4-null mice were protected from disease relative to wild-type animals in the mouse model of NEC. Hackam and co-workers⁴⁸ soon thereafter confirmed some of these findings and documented increased TLR4 expression in intestinal epithelium resected from human infants with NEC. They further demonstrated that the role of TLR4 in disease formation relates to enterocyte injury and repair. It is interesting to note that additional data have suggested that TLR4 expression can be induced by platelet-activating factor, a molecule that has been previously implicated in the pathogenesis of mucosal injury in NEC.⁴⁹

Other notable work has identified the importance of inducible nitric oxide synthetase (iNOS), which is upregulated in the presence of microbes, in the development of experimental NEC.^43,50,51 In the setting of inflammation, increased iNOS activity is believed to generate toxic intermediates of nitric oxide that can compromise mucosal integrity. Future work must elucidate the mechanisms by which bacteria activate TLR4 and iNOS pathways (among others) in the setting of NEC and must determine if the so-called normal commensal bacteria, or invading pathogens, are primarily responsible for activating these pathways during disease pathogenesis.

Metagenomics

Accumulating experience with the molecular characterization of intestinal microbes has clearly demonstrated the value of sequencing all microbial DNA within the intestine and not solely the gene sequences of marker genes such as 16S⁶⁹ (Fig 1). Studies that aim to make taxonomic identifications of gut microbes overlook the reality that 2 strains of a given bacterial species may differ in DNA content by as much as 25%.⁷⁰ For example, a 16S-based study may identify the presence of an E coli species but would not provide any information about whether that particular strain of E coli is predicted (on the basis of DNA content) to be a harmless commensal or harmful pathogen. Such DNA variation, termed genome plasticity, can have important consequences for the clinical phenotype of an organism. Therefore, 16S-based studies fail to provide predictions of bacterial community function that are made possible by sequencing the entire intestinal “metagenome.”

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 1

DNA-based investigations of microbial community structure.

The metagenome (or “community genome”) is defined as the combined genomes of all species within a mixed microbial population, and sequencing metagenomes was not technically feasible before the recent advances in sequencing technology.⁶⁹ The initial metagenomic studies of the distal human intestine were path-breaking because they proved that the gut microbiome encodes for a surprisingly diverse array of metabolic processes^71,72; in fact, these studies proved that gut microbes are essential for degrading specific components of the human diet. In addition, metagenomic sequencing has identified the importance of DNA sequences related to microbial virulence and horizontal gene transfer,^73,74 both of which could be important to NEC pathogenesis. Successfully characterizing all microbial DNA within biological samples is substantially more difficult when samples are complex. Because the neonatal intestinal tract contains far fewer species than the adult intestinal tract, a metagenomic approach offers the possibility of more fully characterizing the gut microbial metagenomes of individual patients at risk for NEC. As the cost of high-throughput sequencing continues to decrease, it should become feasible to conduct an adequately powered comparison of metagenomes from infants with and without NEC.

Metagenomics offers particular promise for investigations of viruses within the gastrointestinal tract. In contrast to marker genes such as 16S that are present in all bacterial species, there are no RNA or DNA sequences shared by all viruses that can be used for easy culture-independent species identification.⁷⁵ However, if all viral DNAs and RNAs within a biological sample are sequenced, then the diversity and complexity of the viral community can be characterized accurately. Such an approach has been used in limited instances to demonstrate that a large number of viruses within the gastrointestinal tract are actually bacteriophages that infect gut bacteria.⁷⁶ A recent study demonstrated that, whereas meconium from a newborn infant did not contain viral particles, samples obtained 1 week later from the same infant contained over 10⁸ viral particles per gram of feces.⁷⁷ Another study used a metagenomic approach to identify novel viruses present within stool samples from pediatric patients with diarrhea.⁷⁸ The clinical relevance of these findings is not yet known.

Functional Characterizations of Intestinal Microbial Communities

DNA-based investigations can delineate the genetic potential of a gut microbial community. However, it is well recognized that the presence of a particular microbial gene within the gut microbiome by no means indicates that the gene in question will be expressed. For example, documenting the presence of a gene encoding a protein involved in virulence or antibiotic resistance indicates that the gene might be expressed without indicating whether the gene actually is expressed in a given biological sample. As such, there is great interest in developing technologies to assess the phenotype of microbial communities.

Three general lines of inquiry are now being applied to evaluate the function of naturally occurring microbial communities such as those found in the gastrointestinal tract (Fig 2). Each of these approaches is routinely used (eg, measurement of gene expression) to characterize behavior of bacterial monocultures under tightly controlled conditions in the laboratory. However, these tasks are made considerably more difficult when applying such technologies to study naturally occurring biological samples that contain unknown and often complex mixtures of microbial species. Metatranscriptomics, or “community transcriptomics,” seeks to identify the genes that are expressed by all species present within a complex specimen such as a fecal sample. This approach has recently been used to detail community gene expression in samples from ocean water and soil^79,–,81; however, no published study has thus far detailed community gene expression within the intestinal tract. Metaproteomics, or “community proteomics,” is similarly the study of proteins collectively expressed within microbial communities. Particularly when corresponding metagenomic data sets are available, mass-spectrometry–based proteomic analysis has enabled the generation and subsequent validation of fascinating hypotheses regarding function of previously unrecognized proteins.^82,83 Recently, 2 publications indicated that this approach is feasible in the study of gut microbes in humans.^84,85 Finally, metabolomics is the study of low molecular weight microbial metabolites within biological samples. Identification of metabolites within a gut microbial community arguably serves as the most accurate measure of microbial function by unambiguously measuring the end products of gene and protein expression.^86,87 Each of these high-throughput approaches to characterizing gut microbial communities has technical and theoretical advantages and disadvantages, and it has not yet been proven that functional studies of fecal microbes are informative with regard to microbial function elsewhere in the gastrointestinal tract. Although these approaches have not yet been exploited to study the microbiology of NEC, the simplicity of the juvenile intestinal tract may allow for successful completion of such studies in coming years.

• Download figure
• Open in new tab
• Download powerpoint

FIGURE 2

High-throughput investigations of microbial community function.

Diet and the Intestinal Microbial Community

An important lesson from the path-breaking research over the past decade has been that nutrition, gut microbial function, and energy metabolism are tightly intertwined.^4,88 This has particular relevance to VLBW infants, because the incidence of microbe-mediated disease such as NEC and late-onset sepsis is significantly higher in formula-fed infants.⁸⁹ A critical and unresolved issue is how gut microbial communities differ in infants who receive artificial formula and those who receive maternal milk. Because different dietary sources can be viewed as distinct “growth media” for intestinal microbes, it makes sense intuitively that differences must exist between the gut microbial communities of milk-fed and formula-fed infants. To date, however, the results of studies that used culture-based techniques and 16S-based molecular profiling do not support a firm conclusion regarding differences between these 2 groups.^27,63,90 However, high-throughput molecular investigations of this subject are only beginning to emerge and are likely to identify such differences.

We recently completed a metabolomic analysis of fecal samples from milk-fed and formula-fed premature infants by using a mass-spectrometry–based platform (V.P. and M.J.M., unpublished data). We identified a total of 174 chemical compounds within these samples, and the distribution of chemicals within these samples was sharply affected by diet. Principal component analysis of the data sets demonstrated that samples from milk-fed and formula-fed infants clustered tightly together. These results indicate that the active metabolic pathways used by microbes within the intestinal tracts of milk-fed and formula-fed infants are distinct. The application of other high-throughput technologies will help further define the impact of diet on the gut microbial communities of newborn infants.

Probiotics and the Intestinal Microbial Community

Multiple clinical studies have been performed to determine if the incidence of NEC can be reduced by the administration of probiotics (defined as dietary supplements of live microorganisms that are derived from the human intestinal tract and that have putative health benefits). The hypothesis that probiotics might help prevent NEC stems from the observation that the prevalence of probiotic genera such as Bifidobacterium and Lactobacillus is lower within the gastrointestinal tracts of hospitalized preterm infants than in otherwise healthy term infants.¹⁵ A recent meta-analysis concluded that enteral supplementation of probiotics does reduce the risk of developing NEC for low birth weight infants.⁹¹ Although numerous questions remain regarding safety, dosage, and the specific species used in different formulations, the prospect of a new strategy to prevent onset of NEC is exciting.

Nonetheless, there are several important points to be made about probiotics and NEC in the context of emerging technologies to characterize intestinal microbes. First, the purported deficiency of “beneficial” gut bacteria in premature infants is based on older, culture-based studies and, thus, should be viewed with caution. Ongoing studies to document colonization patterns with molecular techniques will be particularly valuable in confirming or refuting such long-held beliefs. Second, no culture-independent studies have demonstrated that infants with NEC have lower counts of probiotic species than matched controls without NEC. Proving such a conclusion (which may be possible with studies now in progress) would greatly strengthen the rationale for using probiotics. However, if the pathogenesis of NEC is truly related to a deficiency of favorable bacteria, and if this deficiency is shared by all premature infants, one must wonder why more than 90% of preterm infants lack those desirable bugs but do not develop the disease. Likewise, it must be considered whether the benefits outweigh the risks in administering live bacterial supplements to all patients at risk. The difficulty, of course, lies in prospectively identifying which patients are at highest risk for developing the disease. At present, there is no effective method to ascertain which 5% to 10% of VLBW infants will develop the disease. Theoretically, there may be some characteristics that distinguish the gut microbes of patients who eventually develop NEC from those who do not. If this is true, it may be feasible to give probiotics to a select group of patients at high risk rather than to all patients.