key: cord-0004824-ym8k2y95 authors: Panigrahi, Pinaki; Gupta, Sunil; Gewolb, Ira H; Morris, J Glenn title: Occurrence of Necrotizing Enterocolitis May Be Dependent on Patterns of Bacterial Adherence and Intestinal Colonization: Studies in Caco-2 Tissue Culture and Weanling Rabbit Models date: 1994 journal: Pediatr Res DOI: 10.1203/00006450-199407001-00021 sha: 8154bc87acc917fb9801492efc247f96f51ae6e8 doc_id: 4824 cord_uid: ym8k2y95 ABSTRACT: Necrotizing enterocolitis (NEC) is one of the leading causes of death in neonatal intensive care units. The underlying pathophysiology of NEC is poorly defined, although there is a suggestion that bacterial agents play an important role in the process. In this study, we evaluated bacterial isolates from 17 NEC cases and matched asymptomatic control infants. Isolates from NEC patients were no more likely than control isolates to be adherent to enterocytes, as assessed by a Caco-2 cell tissue culture model. Adherent Escherichia coli isolates, from both NEC cases and controls, were able to cause pathologic changes typical of NEC in a weanling rabbit ileal loop model. Adherence of E. coli strains to Caco-2 cells, and subsequent production of disease in weanling rabbits, could be blocked by coinfection with Gram-positive isolates from control children. In contrast, in three of four instances, adherent E. coli from NEC cases retained their adherence and caused illness in rabbits when coinfected with Gram-positive isolates from the homologous child. Our data suggest that patterns of intestinal adherence, as influenced by the underlying intestinal microbial ecology, play a role in the pathophysiology of NEC. NEC is a serious gastrointestinal disorder in newborns affecting predominantly premature or low-birth-weight infants. NEC is one of the leading causes of death in NICU, with an incidence of 2-3% in premature infants and a mortality of 10-55% (1) (2) (3) (4) . The etiology of NEC is unknown. Although there have been outbreaks of NEC associated with specific infectious agents (2, (5) (6) (7) (8) (9) (10) , the range of agents isolated is broad, and in studies of endemic disease it has not been possible to consistently link a single infectious agent with illness (11) . During the past decade, there has been increasing recognition that the environment or milieu in which a microorganism is placed can have a profound effect on its viru-quent production of disease in weanling rabbits, could be blocked by coinfection with Gram-positive isolates from control children. In contrast, in three of four instances, adherent E. coli from NEC cases retained their adherence and caused illness in rabbits when coinfected with Grampositive isolates from the homologous child. Our data suggest that patterns of intestinal adherence, as influenced by the underlying intestinal microbial ecology, play a role in the pathophysiology of NEC. (Pediah. Res 36: 115-121,1994) Abbreviations NEC, necrotizing enterocolitis NICU, neonatal intensive care unit lence (12) . In experimental animals, for example, severe combined immunodeficient mice can be protected against lethal cryptosporidium infection by modification of intestinal microflora with a combination of avirulent bacteria (13) . In humans, Closfridium difficile often colonizes the intestine without any apparent consequence (14, 15) , with colitis occurring in settings in which there are perturbations of the intestinal flora due to administration of antibiotics or chemotherapeutic agents. There are also suggestions from rabbit studies that with some bacteria the simple process of intestinal adherence or colonization can lead to illness, without expression of specific toxins or other virulence factors (16) . "Schaedler's Cocktail," a combination of harmless bacteria, is routinely used while raising specific pathogen-free rodents; when animals are raised germfree without such treatment, normal physiologic development of the gut does not occur, and they become susceptible to pathogens that would be normal flora for the same strain of rodent (13) . We hypothesized that similar processes were important in NEC-that normal flora bacterial isolates might be able to cause illness based on patterns of adherence to the intestinal mumsa and on the underlying microbial ecology of the neonatal gut. To test this hypothesis, we examined a collection of bacterial isolates from 17 NEC cases and matched controls. As a marker for adherence and colonizing ability, we evaluated adherence of each bacterial strain in a Cam-2 cell culture assay (17) (18) (19) (20) ; we also looked at the effect of combinations of bacteria on colonization in this model. To assess the in vivo relevance of these observations, selected strains and strain combinations were inoculated into ileal loops in weanling rabbits, a model that has been used in prior NEC studies (21, 22) . Collection of bacterial isolates and clinical characteristics of cases have been previously reported (1 1). Briefly, stool samples were collected from 17 infants diagnosed with medical or surgical NEC and from matched control subjects. Children were hospitalized in the NICU at the University of Maryland Hospital between June 1990 and February 1992. Case patients were defined as those infants who fulfilled the modified Bell's criteria for staging of NEC (stage 1 to 3) (23). Matched controls were selected from infants who met the following criteria: 1 ) no evidence of NEC; 2) birth weight within 25% or 250 g of the birth weight of the case patient; 3) presence in the NICU at the time of diagnosis of NEC in the matching case subject; 4) not on antibiotics at time of enrollment; 5) postnatal age within 1 wk of the matched case patient. Studies were reviewed and approved by the Human Volunteer Research Committee, University of Maryland at Baltimore. A single stool sample was collected as soon as the definition of NEC was satisfied. Case patients were enrolled only if a stool was passed no later than 5 h after the first antibiotic dose. A single stool specimen was obtained in controls as soon as possible after enrollment. All representative colony types were picked and identified to the species level by using standard microbiological techniques, with confirmation by API20E or API-Staph Ident (Analytab Products, Plainview, NY). Isolates were frozen in Luria broth (L-broth) containing 25% glycerol and stored at -70°C. For each experiment, fresh bacteria regrown from frozen stock were used. A listing of all isolates in the collection is given in Table 1 . In seven of eight instances in which infants subsequently had a positive blood culture, the organism isolated from blood was also present in stool; in five instances, Staphylococcus epidemidis was the blood isolate. Isolates within the same species were further evaluated by plasmid profile analysis. As previously reported ( l l ) , these data suggested that each infant was colonized with a limited number of unique strains: isolates of the same species from different infants had different plasmid profiles, whereas almost all isolates of the same species from the same infant had the same plasmid profile. For the purposes of the current study, a single isolate of each species from each child was selected for evaluation. In the few instances in which more than one plasmid profile was present among isolates of the same species from the same infant, an isolate of the most common plasmid profile was selected. All Escherichia coli isolates were screened by DNA probes to confirm that they did not belong to previously identified pathogenic groups, including enterotoxigenic, enteropathogenic, enteroinvasive, enterohemorrhagic, or enteroaggregative E. coli (24) . Only two isolates from control infants were positive for the gene encoding diffuse adherence. In vitro adherence assay using Caco-2 cell.. Ten-d-old Caco-2 cell monolayers grown on chamber slides (Nunc, Naperville, IL) were infected with lo8 bacterialml in antibiotic-free medium at 37"C with gentle rocking for 1.5 h. The inoculum was then removed, and monolayers were washed, stained with Giemsa, and evaluated by light microscopy. Coinfection studies were conducted using a combination of Gram-negative and Gram-positive organisms. Adherence was graded on a scale of 0 to 4 (0: no adherent bacteria; 1: one to five bacteria; 2: six to 25 bacteria; 3: 25 to 50 bacteria; and 4: more than 50 bacteria per cell). Each experiment was repeated at least three times before assigning any adherence grade to a strain. Ileal loop model in weanling rabbit. Ileal loops were prepared in weanling New Zealand White rabbits weighing between 392 and 430 g by following standard methods (21, 22, 25) . Each bacterial isolate was inoculated in duplicate animals. Rabbits were killed after 16-18 h, gross changes in the loops were noted, fluid accumulation was measured, and tissue samples were fixed in formalin for histopathology. Only the center portion of the loop sufficiently away from the ligature sites was collected to avoid any local inflammatory changes due to physical trauma. All studies were approved by the Institutional Animal Care and Use Committee of the University of Maryland at Baltimore. Representative adherent and nonadherent strains of E. coli, Klebsiella, Staphylococcus, and Enterococcus species from NEC patients and controls that had been already evaluated in the in vitro Caco-2 cell culture model were selected for study. A total of 5 x lo7 organisms were used per loop (5 cm) for monocontamination exper-iments. In coinfection (bicontamination) studies, Gram- Table 3 . Results of adherence assays performed on E. coli, Klebsiella pneumoniae, enterococci, and S. epidennidis are shown in Table 2 . Strains of E. coli were most likely to have high grade (grade 4) adherence. Adherence of Klebsiella species was much less pronounced in our Caco-2 cell model. Enterococcus and Staphylococcus species showed moderate adherence (grade 1-3). All other strains isolated from case patients as well as from controls showed minimal (I+) or no adherence, with the exception of one strain of Citrobacterfreundii that showed a 4+ adherence (data not shown). There was no clear association between the adherence grade of the colonizing strain and the occurrence or severity of NEC. In vih.0 adherence ajler coinfection. Because there were no obvious differences in the adherence grade of isolates from case or control infants, we examined the adherence pattern of selected Gram-negative species in combination with Gram-positive bacteria isolated from the same infants. Studies were limited to strains from infants from Control 0 + 0 + Control 0 + 0 + Control 1 + Control O+ 0+ Control 1 + 1 + whom both a Gram-negative and a Gram-positive species adherence, two had coinfections with enterococci and had been isolated. one with staphylococci; the E. coli strain from the fourth As shown in 39 three of three adherent NEC case patient (which became nonadherent on coin-(grade 4) E. coli from control infants became nonadherent fection) was coinfected with an enterococcal strain. (grade 0) upon coinfection with a Gram-positive isolate Coinfection studies were repeated by crossing the from the homologous infant. The coinfecting Gramstrains, i-e. combining an enterococcus from a control positive strains retained adherence, albeit at levels comchild with an adherent E. coli from an NEC case patient parable to Or 'lightly that seen when the Gramand enterococci from an NEC case patient with adherent positive strain was inoculated in pure culture. In two E. coli from a control infant. When E. faecium strain 20-2 instances (for E. coli strains 6-1 and 32-I), Enterococcus from a control patient was used in coinfection studies faecalis was the Gram-positive partner; in the third, it with E. coli strains 21-1,25-1,27-1, and 33-1 (from NEC was Enterococcus faecium . case patients), all became nonadherent. In contrast, E. In three of four adherent strains from fecolis strain 21-2 (isolated from an NEC case patient) NEC case patients remained adherent when combined could not abrogate the adherence of E. coli strains 6 1 , with Gram-positive isolates from the homologous infant 20-1, and 32-1 isolated from control infants. Represen-( Table 3) . Of the three case E. coli strains that retained tative results are shown in Experimental NEC in rabbit ileal loops. All control loops had a healthy gross and microscopic appearance, and there was minimal (<0.5 mLlloop) or no fluid accumulation. Fluid accumulation in loops infected with bacteria ranged from 1.6 to 2.4 mL12.5 cm of loop. Adherent strains of E. coli (isolated from NEC case patients and controls) produced grossly apparent damage to the loop, with purulent necrotic spots and hemorrhagic fluid accumulation (Table 4 ). Histopathology showed massive necrosis of luminal epithelium extending to the muscle layers, with edema, hemorrhage, and polymorphonuclear cell infiltration. Bacteria were seen that were adherent to necrosed epithelial cells. In keeping with reports of experimental NEC from other investigators, pneumatosis intestinalis was not observed in our model. Strain invasion was confirmed by visualization of large numbers of bacteria in the lamina propria and submucosa of the infected loops (Fig. 2) . All of these rabbits appeared sick, and one died between 12 and 16 h after inoculation due to rupture of the loop. Only the inoculated organisms were cultured from the infected loops at necropsy. One Klebsiella strain caused pathologic changes comparable to those seen with adherent E. coli. The remaining two Klebsiella strains tested showed mild inflamma-tory reaction, with infiltration of neutrophils, congested vessels, and some blunted villi with hyperplastic epithelium. Villus necrosis was not produced, and lamina propria, submucosa, and the muscle layers were healthy in the loops infected by these Klebsiella strains. Ileal loop coinfection with Gram-positive bacteria. Results of coinfection studies in rabbit loops correlated with results of adherence assays in the Caco-2 model (Table 4 ). In instances in which coinfection resulted in a loss of adherence, typical NEC-like pathology was not seen. There were signs of mild inflammation in a few focal areas showing flattened villi where the single-layered simple columnar epithelium was replaced by more basophilic, hyperplastic cuboidal epithelium, suggesting active regeneration after mild injury (Fig. 2) . In contrast, strains that retained adherence in the Caco-2 coinfection assay produced typical NEC-like injury in the ileal loops in the presence of the Gram-positive partner. This study was made possible by the availability of a unique collection of bacterial isolates from NEC case patients and matched controls. These strains were originally collected in an effort to identify bacterial strains or species that might be associated with NEC cases; as such, quantitative cultures were not obtained. However, the observation that in most instances all picks of a common species from a single patient had a common plasmid profile, and the intentional selection of the most common plasmid profile when multiple profiles were identified, suggests that the isolates studied represented the predominant strain of each species in each infant. In this study, seven of eight blood isolates were also present in the stool. It is possible that in some instances blood isolates reflect entry of intraluminal organisms through damaged mucosa, and may not be indicative of the organism that triggered the underlying pathologic process. Epidemics of NEC have been associated with single pathogens such as E. coli, Enterobacter cloacae, K. pneumoniae, Salmonella spp, S. epidemidis, Clostridium butyricum, C. dificile, corona virus, enterovirus, and rotavirus (2, (5) (6) (7) (8) (9) (10) . Although these outbreak reports have generated interest in the role of specific microorganisms in NEC, they may not be representative of "normal" endemic NEC in neonatal nurseries. In our studies here at the University of Maryland, we have found no evidence that a single agent is responsible for all NEC cases or even for isolated clusters of cases (11) . We have observed no association between NEC and production of toxins such as the delta hemolysin of S. epidemidis (7, 8) . We have also failed to find a correlation between disease and the ability of strains to ferment carbohydrate, as measured by P-galactosidase activity (21, 22, 26) . These observations led us to investigate other possible pathogenic mechanisms that might underlie NEC. Bacterial adherence to host cells is a well-studied phenomenon in almost all bacterial mucosal diseases (27) (28) (29) (30) . The process is frequently complex, and there is a wide range of host cell and bacterial responses that are evoked as a result of bacterial adherence that can culminate in tissue injury (31) . This study represents the first effort to assess the possible role of bacterial adherence in the pathogenesis of NEC. We screened all isolates from NEC case patients and controls for adherence using a Caco-2 cell assay (17) . Caco-2 cells are derived from a moderately differentiated human colon adenocarcinoma that differentiates into enterocyte-like cells without any inducers during in vitro culture. The differentiation occurs at late confluence when the cells exhibit typical brush-border microvilli and tight junctions. At this polarized stage, the cells express typical small-intestinal enzymes, e.g. sucrase-isoma-Itase, lactase, aminopeptidase, and alkaline phosphatase (18, 19) . We have previously described the use of these cells to study the pathogenesis and evaluate virulence of non-01 Vibrio cholerae (20) , and this model is now extensively used to study the pathogenesis of several other enteric bacterial species (31, 32) . In this study, we were not able to show that isolates from NEC case patients were more likely to be adherent than isolates from control infants. However, when selected strains were evaluated in a weanling rabbit ileal loop model, highly adherent strains (from case patients as well as from controls) produced typical NEC-like injury. Pathologic changes were not seen after infection with Gram-positive organisms. Although they do not explain why NEC occurred in some infants but not in others, these data do suggest that, in the case of E. coli, a high level of adherence may be necessary but not sufficient for production of NEClike injury. Results became more interesting when we attempted coinfection with Gram-positive and Gram-negative isolates from the same infant. Control Gram-negative iso-lates that were adherent in pure culture lost their ability to adhere in the Caco-2 model when coinfected with Gram-positive isolates from the homologous infant. With one exception, adherence of Gram-negative isolates from NEC case patients was not affected by coinfection with a homologous Gram-positive isolate; however, when these strains were coinfected with an enterococcal strain isolated from a control child, adherence was lost. These results were mirrored in the weanling rabbit model, with loss of adherence correlating with failure to produce typical NEC-like changes. The mechanisms responsible for these changes in adherence seen in association with coinfection remain to be determined. Among other possibilities, Gram-positive organisms could be saturating the adherence-receptor sites on the epithelial cells for the Gram-negative bacteria, or expression of adherence and virulence factors by the Gram-negative bacteria may be affected by the presence of Gram-positive organisms. Inasmuch as we were concentrating on the effects of adherence and colonization, we did not introduce other variables (substrate, pH, ischemia, temperature, etc.) into our experiments. These factors will need to be taken into consideration in subsequent studies. It may also be noted that our results dealing with protection are most compelling with respect to E. coli. Although we have examined other Gram-negative strains in our model systems, there is a need for further investigation with larger numbers of other Gram-negative strains before we can generalize the phenomenon of protection. NEC is undoubtedly the result of a complex interaction of factors, including gut maturity, prior ischemic injury, inflammatory mediators, and other unidentified factors (2, (33) (34) (35) (36) (37) . We cannot yet explain the pathophysiologic mechanisms underlying our observations. However, our data suggest that bacterial adherence and patterns of bacterial colonization of the neonatal gut play a pivotal role in the development of NEC. In this context, factors that change or modify the innate pattern of microbial colonization should be given serious consideration in the management of preterm infants. To begin with, infants under aseptic NICU care have very limited exposure to microorganisms. Use of broadspectrum antibiotics in the management of respiratory and other associated conditions in these group of infants could further damage the existing microflora. Endogenous antimicrobial peptides, such as defensins and cryptdins, have an increased antibacterial property on Gram-positive organisms compared with Gramnegative bacteria (38, 39) . Elimination of or decrease in the population of such Gram-positive organisms may have a profound effect in circumstances in which the infant's intestine is colonized with only two to three species of bacteria. Although gaining insight into the multiplicity of these factors will require substantial additional research, our observations now provide a starting point for such investigations and for judicious therapeutic interventions in this devastating disease. Epidemiology of necrotizing enterocolitis: a case control study Necrotizing enterocolitis Epidemiologic study of necrotizing enterocolitis among low-birth-weight infants Neonatal necrotizing enterocolitis: a nine-year experience. Epidemiology and uncommon observations Neonatal necrotizing enterocolitis: implications for an infectious disease Models of pathogenesis of necrotizing enterocolitis Analysis of enteric coagulasenegative staphylococci from neonates with necrotizing enterocolitis Delta-like toxin produced by coaghase-negativestaphylococci is associated with necrotizing enterocolitis Neonatal rotavirus-associated necrotizing enterocolitis: case-control study and prospective surveillance during an outbreak Association of coronavirus infection with neonatal necrotizing enterocolitis Gewolb IH 1992 Stool bacterial plasmid profiles in infants with necrotizing enterocolitis Mackowiak PA 1 9 8 2 -~e normal microbial flora Resistance of severe combined immunodeficient mice to infection with Ciyprosporidium patvum: the importance of intestinal microflora CIostridium dificile-associated diarrhea in infants: lack of association between the presence of toxin and the occurrence of illness Closrridium dificile: colonization and toxin production in a cohort of patients with hematologic malignancies Small-bowel colonization alone is a cause of diarrhea Enterocyte-like differentiation and polarization of the human colon carcinoma cell line Caco-2 in culture Characterization of the human colon carcinoma cell line (Caco-2) as a model system for intestinal epithelial permeability Epithelial polarity, villin expression, and enterocytic differentiation of cultured human colon carcinoma cells: a survey of twenty cell lines Development of an in virro model for study of non-01 Vibrio cholerae virulence using Caco-2 cells Elseviers D 1988 A bacterial pathogenicity determinant associated with necrotizing enterocolitis Necrotizing enterocolitis: intramural biochemistry in human neonates and a rabbit model Necrotizing enterocolitis: a practitioner's perspective Acute diarrhea in Baltimore children attending an outpatient clinic Human immune response to Campylobacrer jejuni proteins expressed in vivo Intraluminal pathogenesis of necrotizing enterocolitis Adhesive properties of enterobacteriaceae Host-specific fimbrial adhesins of non-invasive enterotoxigenic Escherichia coli strains Models for studying the role of bacterial attachment in virulence and pathogenesis Consequences of microbial attachment: directing host cell functions with adhesins Common themes in microbial pathogenesis A new in virro model of Enramoeba hisrolytica adhesion, using the human colon carcinoma cell line Caco-2: a scanning electron microscopic study Kosloske AM 1990 A unifying hypothesis for pathogenesis and prevention of necrotizing enterocolitis Colonic fermentation of carbohydrate in the premature infant: possible relevance to necrotizing enterocolitis Necrotizing enterocolitis of the neonate Intestinal ischemia and necrotizing enterocolitis Bacterial colonization of neonates admitted to an intensive care environment Defensins: antimicrobial and cytotoxic peptides of mammalian cells Cryptdins: antimicrobial defensins of the murine small intestine Acknowledgments. The authors thank Aaron Joseph and Wendy Martin for the excellent technical help.