The findings of the current study confirm that fecal calprotectin levels are elevated in premature infants, compared with older age groups, with a wide range of inter- and intra-individual variation during the first few weeks of life. They further demonstrate that the most significant factors that affect calprotectin excretion are ante and per natal antibiotic treatment, volume of enteral feeding (ml/Kg/day), the occurrence of unplanned interruptions of enteral feeding, and the gastrointestinal bacterial colonization.
Consistent with previous studies 
, the fecal calprotectin levels observed in healthy preterm infants in the current study were high and clearly exceed those reported in healthy adults and children. Likewise, we observed a wide range of inter- and intra-individual variation. Such range of variation is unlikely to result from a poor stability of calprotectin in stools, since calprotectin resists proteolysis, and is stable at room temperature for up to a week 
. Although stool samples were collected from babies' diapers in the current study (as in all studies carried out in infants), the sampling technique cannot account for the variation either, since the changes in calprotectin concentrations found in the current study largely exceed what could be accounted for by water absorption 
. Besides, the changes observed cannot arise from the ingestion of calprotectin present in human milk, since, in agreement with earlier findings 
, none of the human milk samples we analyzed (n
10, data not shown) contained any detectable calprotectin (threshold: 15 mg/L). Thus, the wide range of f-calprotectin values may reflect true inter- and intra-individual variability in calprotectin fecal excretion in that patient population.
The high calprotectin levels observed in neonates may reflect the increased transepithelial migration of neutrophil granulocytes and/or macrophages into the intestinal lumen of preterm infants. As Berstad et al reported a significant correlation between calprotectin levels in gut lavage fluid and intestinal permeability 
, the increased migration of neutrophil granulocytes and/or macrophages into the gut lumen might be related to the higher intestinal permeability associated with intestinal mucosal immaturity. However, f-calprotectin levels are similar in preterm and full term infants, although intestinal permeability is higher in preterm than in term infants 
In univariate analysis, levels of calprotectin in the initial samples did not correlate with gestational age, birth weight or the mode of delivery, supporting earlier reports 
. Although other authors found a correlation of calprotectin with gestational age, the latter correlation was found with early determination in meconium, whereas calprotectin levels subsequently decreased during the first week postnatal 
. Likewise, we found no significant correlation between f-calprotectin and type of feeding (formulas as exclusive or predominant source of feeding vs.
human milk) in univariate analysis, consistent with other studies 
By contrast, gut microbiota appeared to influence calprotectin excretion in the present cohort, as suggested by Josefsson et al 
. In the current study gut microbiota was analyzed by culture, which allowed the isolation and identification of the main bacterial genera, even in a sub-dominant status. A limitation of this approach could be the inability to detect the uncultivable part of the microbiota. However, by contrast with adult gut microbiota, the latter accounts for a very small fraction of the overall bacterial population in preterm neonates 
. We found correlations between f-calprotectin and both microbiota per se
, and other factors known to influence gut bacterial establishment as well. The current report thus is first to provide evidence for an effect of intestinal bacterial colonization on fecal calprotectin excretion Three lines of evidence point to such effect. First in both univariate and multivariate analyses, f-calprotectin correlated positively with intestinal colonization by Staphylococcus
. Secondly, in univariate analysis, fecal calprotectin levels correlated positively with postnatal and postconceptional ages, factors known to influence the gut microbiota composition. However, this correlation did not remain significant in multivariate analysis in accordance with Campeotto et al 
, as opposed to the study by Josefsson et al 
. Thirdly, the use of antibiotics impacted f-calprotectin levels. Indeed, throughout the study f-calprotectin correlated negatively with ante and per natal antibiotics in univariate and multivariate analyses. To our knowledge, the current study is first to demonstrate such impact. Indeed, earlier studies have shown changes in the gut microbiota establishment in infants born from mothers who had received antibiotic per partum 
. Likewise, a negative correlation was found with neonatal antibiotic courses in univariate analysis, but in contrast to other studies 
this correlation did not remain significant in multivariate analysis. However, in the latter study a correlation was only observed in infants treated with cefotaxim and meropenem, two broad-spectrum antibiotics.
To summarize, factors known to delay gut bacterial colonization (both ante- and post-natal antibiotic treatments) correlated negatively with fecal calprotectin levels, whereas factors known to favor gut bacterial colonization (gestational age and postconceptional age) correlated positively with fecal calprotectin levels. This is in accordance with the study of Mohan et al, who described a decrease in f-calprotectin levels in infants supplemented with a probiotic strain 
. In the latter study, probiotic supplementation increased bifidobacteria levels, and decreased the levels of clostridia, a genus positively correlated in with f-calprotectin in the current study. By contrast, a recent study did not find any correlation between gut microbiota colonization and f-calprotectin: however, the culture techniques used in that report did not allow detection of clostridia 
Otherwise, we observed a highly significant, positive correlation between the volume of enteral feeding (mL/Kg/day) and f-calprotectin excretion in multivariate analysis, as previously reported 
Taken together, the current results suggest that exposure to two kinds of luminal ‘antigens’−i.e, commensal intestinal bacteria, and dietary antigens− might induce a state of “physiological” subclinical intestinal inflammation in preterm infants as well as in full-term infants. It is tempting to speculate that intestinal bacteria, and specific, individual components of the commensal microbiota might have variable abilities to stimulate transepithelial granulocyte migration and/or to induce calprotectin release from leucocytes and macrophages, as shown in gnotobiotic piglets colonized with various strains of Escherichia coli 
. Thus, the stress induced by birth per se
and by the adaptation to the extra-uterine life, particularly concerning gut bacterial colonization, rather than the degree of mucosal maturation or term of birth, may account for the higher fecal calprotectin levels found in infants, compared to older age groups. This high intestinal expression of calprotectin, known to display antimicrobial properties, might participate in the mechanisms of defense in neonates, whose intestinal immune system is not mature.
Despite these “physiological” high levels of f-calprotectin, several studies strongly suggest that a rise in f-calprotectin above this high, baseline levels may be a candidate, non-invasive marker of gastrointestinal diseases, in particular NEC 
. These studies reported a significant rise in f-calprotectin levels in infants suffering from gastrointestinal disease, particularly from NEC. Several thresholds for suspicion of NEC have been proposed, i.e. 200mg/L in Caroll's study 
, 2000 µg/g in the study by Josefsson 
, and 636 µg/g for the study by Campeotto et al 
. The threshold in Caroll's study appears too low. Indeed, the latter study, performed in a very small group of infants (6 infants with NEC, and 6 healthy controls), reported low f-calprotectin values in healthy preterm infants as compared with the literature. In our study, we did not observe any case of NEC, and median f-calprotectin excretion was 138 µg/g [interquartile, 58–271 µg/g]. Using the three thresholds reported among the infants in our cohort 34 infants should have been suspected of having NEC using Caroll's threshold, 0 with Josefsson's cut-off, and 3 with Campeotto's threshold. The rise in calprotectin levels we observed in cases of intolerance to enteral feeding suggests the monitoring of f-calprotectin might be useful as a warning signal for gastrointestinal disease and/or poor tolerance of enteral nutrition. However, the lack of specificity we observed might be due to the “physiological inflammation” linked to the bacterial colonization. Many more studies with much larger cohorts are warranted to confirm whether such thresholds or cut-off values could be recommended for routine use in clinical practice.
Finally, it may seem paradoxical to observe that a rise in calprotectin is observed in parallel with increased volume of enteral feeding, but a rise is also observed when there is a need to interrupt enteral feeding, i.e., in cases of poor digestive tolerance. To further address this issue we divided samples according to the terciles of enteral feeding volume received, and, within each tercile, we separated samples depending on the need to interrupt enteral feeding or not. As shown in , on one hand, calprotectin tends to be higher for the upper tercile of enteral feeding volume: this is consistent with the fact that calprotectin increases with enteral volume. On the other hand, for any given tercile of enteral volume administered, calprotectin was higher when feeding intolerance occurred. Should we have very large numbers of feeding interruptions within each tercile, we might be able to define ‘safe’ levels of calprotectin for a given volume of feeding. Interestingly, all instances when feeding had to be interrupted were associated with a calprotectin level >205 µg/g. This is consistent with the good sensitivity of calprotectin, yet this cutoff level has a poor specificity as well, as discussed with the ROC curve (see above).
Fecal calprotectin as a function of the tercile of enteral feeding volume and tolerance to enteral feeding.
In conclusion the current study demonstrates for the first time that calprotectin excretion can be linked to the gut bacterial establishment. We speculate that the enhanced expression of this protein possessing many potential functions including antimicrobial properties may participate in the innate immune system, and thus be of benefit for the developing gut in both full term and premature infants. Clinical situations may therefore occur when a pathological rise in calprotectin might be offset by a physiological increase.