Main

Iron is a nutritional compound that is important for growth of preterm infants, development of the nervous system and hematopoiesis, and is also essential for enteric microbiota. Human milk is low in iron; however, iron absorption in the gut of neonates and preterm infants is enhanced by lactoferrin, which is the main whey protein in human milk. Furthermore, binding to lactoferrin protects nutritional iron from being hijacked by bacteria in the relatively iron-poor environment of the human intestine (1). Recent research indicates that gut bacteria dysbiosis precedes necrotizing enterocolitis (NEC) (2) and that oral iron supplementation may adversely affect the gut microbiome by selectively favoring the growth of pathogenic strains (3).

Another unique feature of iron metabolism in humans is the lack of a natural route for excreting excess iron. Iron overload has been described not only in adults, but also in infants (4, 5). Preterm infants, who are frequently supplemented with iron for long time periods, might be at a particular risk for iron overload, especially if they are carriers of variants with genetically determined high iron uptake.

In all populations, genetic variants influencing iron uptake are common. Large-scale studies involving more than 20,000 adults demonstrated a dose-dependent effect of the A allele of hemochromatosis gene (HFE) rs1800562, the G allele of HFE rs1799945, and the G allele of the transmembrane protease serin 6 gene (TMPRSS6) rs855791 on serum iron levels (6).

As genetic variants are randomly allocated at conception, iron uptake enhancement by the above-mentioned genetic variants is independent from postnatal nutrition, lactoferrin, iron supplementation, and other cofactors. We studied a large group of preterm infants with a birth weight below 1,500 g to describe short- and long-term effects of genetically altered iron uptake with special attention to the development of NEC.

Methods

Patients were enrolled in the German Neonatal Network (GNN) between 2009 and 2014 and in the predecessor study of the GNN between 2003 and 2008. Infants were eligible for enrollment if their birth weight was <1,500 g and their gestational age was <37 weeks+0 days. Parents gave written informed consent. The study parts were approved by the local committee on research in human subjects of the University of Lübeck and the local ethical committees at the other study centers. Infants were selected for analysis if genotyping of all three polymorphisms was successful and information concerning NEC surgery was available.

Clinical data were collected at the study sites until discharge or death of the infant and were transferred to the study center (University of Lübeck). Maternal descent was categorized as “German”, “other European”, “Turkey, Middle East, and Northern Africa”, “Asia”, “Sub-Sahara Africa,” and “Other or unknown”. Small for gestational age status was defined as a birth weight below the 10th percentile according to Voigt et al. (7). Treatment with iron was defined as any treatment with iron during the stay in the hospital. Erythropoietin treatment was defined as any prophylactic treatment with erythropoietin to avoid anemia of prematurity. Oral supplementation with Lactobacillus acidophilus/Bifidobacterium infantis probiotics was recorded as an optional variable until 2012 (written down as “additional medication”) and in all infants thereafter (checkbox on case record forms). We did not include nutritional data (e.g., breastmilk vs. formula) in our analysis as nutrition was not recorded before 2013. Transfusion of blood was defined as any blood transfusion during the stay in the hospital. Intraventricular hemorrhage (IVH) was defined as any IVH according to Papile et al. (8). Sepsis was defined as clinical sepsis with positive blood culture. NEC requiring surgery was defined as clinical NEC with need for laparotomy with or without resection of the necrotic gut and macroscopic diagnosis of NEC made by the attending surgeon. Focal intestinal perforation (FIP) requiring surgery was defined as occurrence of spontaneous intestinal perforation with the need for laparotomy and macroscopic confirmation of isolated FIPs (without inflammatory component) rather than NEC made by the attending surgeon. Death was defined as death during the stay in the hospital.

Follow-up at 5 years was carried out only in infants who were enrolled in the GNN study. All infants were tested by a dedicated team from the study center (one physician, two study nurses, and one medical student) at participating sites. The follow-up team was not aware of clinical or genetic data of the participating children. Body length was determined with Harpenden Portable Stadiometer (Holtain, Crosswell, UK) and body weight with a calibrated scale (M300020, ADE, Hamburg, Germany). Cognitive function was tested with the “Wechsler Preschool and Primary Scale of Intelligence Third Edition (WPPSI-III)”. Cerebral palsy was defined as gross motor function scale >1. Forced expiratory volume in one second (FEV1) (%) was assessed with “Easy on-PC” spirometry system (ndd, Zürich, Switzerland).

DNA samples from infants were obtained after birth during their stay in hospital using buccal swab or umbilical cord tissue and were transferred to the study center. DNA was extracted using a commercial DNA purification kit (Qiagen, Hilden, Germany). Genotypes were determined using the TaqMan 5′ nuclease assay (Applied Biosystems, Foster City, CA) and the 7900HT Real-Time PCR System. Context sequences and VIC/FAM labeling are provided in Supplementary Table S1 online.

Mendelian randomization uses genetic variation as instrumental variables for an intermediate phenotype. As genotypes are randomly transmitted to children, the intermediate phenotype is unaffected by classical confounding. For this reason, demonstration that a genetic variation known to influence the intermediate phenotype level also modifies the disease risk represents an indirect evidence of causal association between phenotype and disease.

In our study two polymorphisms of the hemochromatosis gene (HFE) and one polymorphism of the transmembrane protease serin 6 gene (TMPRSS6) were used as instrumental variables for the intermediate phenotype “high iron uptake”. All three polymorphisms are associated with increased serum iron levels in adults. In adults, each copy of the A allele of HFE rs1800562 increases the serum iron level by 0.37 SDs. The effect of the HFE rs1799945 G-allele and the TMPRSS6 rs855791 G-allele is less pronounced with an increase in serum iron levels by 0.19 SD (8). The effect of each polymorphism on NEC was tested separately as in (ref. 6). Outcome data are reported accordingly. In addition to that, we calculated genetically estimated iron uptake by adding the effects of all three polymorphisms for each patient. The patients were pooled according to their intestinal iron as “high intestinal iron” (no iron uptake enhancing alleles), “intermediate intestinal iron” (serum iron +0.19 to +0.38 SD, i.e., one A-allele rs1800562 or one or two G-allele rs1799945 or rs855791), and “low intestinal iron” (high iron uptake, serum iron >+0.5 SD).

We tested the hypothesis that a genetic background of high intestinal iron uptake is associated with a lower risk of NEC treated with surgery. This was carried out with a Mendelian randomization approach, which was already described (6). As three genetic variants were tested, the significance level was set to <0.05 with adjustments for three tests by Bonferroni–Holm. All further P values are descriptive. Outcome data stratified to the TMPRSS6/HFE genotype and estimated serum iron levels were compared with χ2-test, Fisher’s exact test, t-test, and Cochran–Armitage test for trend and multiple logistic regression analysis. All P values are two-sided. Statistical analyses were conducted with SPSS (version 22, IBM, Armonk, NY, USA) and R (Version 3.3.1, Vienna, Austria).

Results

We genotyped 12,342 preterm infants with a birth weight below 1,500 g. Genotyping was successful in 95.7% of the patients for rs855791, in 95% of the patients for rs1800562, and in 95.8% of the patients for rs1799945. Only patients with complete genotyping data for all three polymorphisms were analyzed (n=11,166). The frequency of NEC surgery was 2.5% (281/11,166). Mendelian randomization estimates for each polymorphism based on serum iron levels in adults (6) are given in Figure 1. The G-allele of rs855791 was significantly associated with a reduced risk for NEC surgery (nominal P=0.004 and adjusted P=0.011). Clinical data stratified to the rs855791 genotype are given in Table 1. NEC rate was high (3.1 %) in the subgroup of infants with the AA genotype, leading to low iron uptake and consecutive higher intestinal iron levels. Infants with the heterozygous AG genotype had an intermediate NEC rate of 2.6%. Infants with the GG genotype (with high iron uptake and consecutive lower intestinal iron levels) had the lowest risk for NEC surgery (2.0%, odds ratio (OR) for AA vs. GG 0.62, 95% confidence interval (CI) 0.44–0.87; P=0.006, Fisher’s exact test). The respective data for the HFE polymorphisms rs1799945 and rs1800562 are given as Supplementary Tables S2 And S3. Few infants carried the rs1800562 AA genotype (n=15). They had a significantly higher birth weight and a higher gestational age when compared with infants with the rs1800562 GG or AG genotype. No other significant differences were observed.

Figure 1
figure 1

Graphical representation of the Mendelian randomization approach. Serum iron levels were transferred from adult data (ref. 6).

Full size image
Table 1 Clinical data stratified to TMPRSS6-rs855791 genotype
Full size table

Frequencies of genotypes stratified to maternal descent are given in Supplementary Table S4. Although the iron uptake enhancing polymorphisms in the HFE gene (rs1800562 and rs1799945) were common in infants of German and European descent if compared with infants from Asia or Sub-Sahara Africa, the reverse was true for the TMPRSS6 rs855791 G allele.

Although the possible maximum of iron-uptake-enhancing alleles in our study is 6, none of the infants carried more than four alleles. To give an estimate of the combined effect of all three polymorphisms, we multiplied the number of iron-enhancing alleles for each polymorphism by the iron serum SD values from adult data (6) and then added them for each patient. Overall, 1,473 infants (13.2%) did not carry one of the iron-uptake-enhancing alleles and were classified as “low iron uptake”. Most infants (n=8,095, 72.5%) carried a single rs1800562 allele or one or two rs1799945 or rs855791 alleles. Their serum iron level was estimated to be “intermediate” (+0.28 SD if compared with non-carriers). In all, 1,598 infants (14.3%) had two alleles (including at least one rs1800562) or –three to four iron-uptake-enhancing alleles. Their serum iron level was estimated to be +0.62 SD above infants without iron-uptake-enhancing alleles and was classified as “high”. Genotypes and stratification are given in Supplementary Table S5; clinical data are given in Supplementary Table S6. The combined data of all three polymorphisms resulted in a more pronounced difference if compared with the rs855791 data. Rates for NEC surgery were reduced by ~50% in infants with high iron uptake if compared with infants without polymorphisms (3.2% vs. 1.8%, OR 0.54, 95% CI 0.34–0.87; P=0.01, Fisher’s exact test). This was also true if serum iron uptake variants were tested as a continuous variable in a multiple logistic regression analysis adjusted for gestational age, gender, multiple birth, maternal descent, small for gestational age status, treatment with probiotics, and participating study site (OR per additional estimated SD of serum iron: 0.44, 95% CI 0.22–0.88, P=0.02). A sensitivity analysis was performed, and parameters were stable. In addition to iron uptake, only study site, gestational age (OR 0.66/week, 95% CI 0.62–0.69, P<0.001), small for gestational age (OR 1.76, 95% CI 1.29–2.40, P<0.001), and treatment with L. acidophilus/B. infantis (Infloran) probiotics (OR 0.64, 95% CI 0.47–0.87, P=0.005) were significant predictors for NEC surgery. As treatment with probiotics is the only modifiable cofactor, we tested the effect of the rs855791 genotype stratified to probiotic treatment. In infants not receiving probiotics, risk for surgery-treated NEC was increased by 12% per A allele at rs855791 (OR 1.12, 95% CI: 1.08–1.70; nominal P=0.002); however this association was not observed in infants who received prophylactic probiotics (OR 1.08, 95% CI: 0.88–1.33; nominal P=0.592, Table 2).

Table 2 NEC surgery rate in infants stratified to treatment with probiotics and TMPRSS6 rs855791 genotype
Full size table

Data concerning the time of NEC surgery were available for 135 of 281 infants (48%) since we started recording this parameter in 2011. The median time of NEC surgery was day 18 of life (interquartile range: days 10–31). No significant differences were observed between infants with different genotypes.

Five-year outcome data were available for 946 infants. No significant differences were observed if infants with low iron uptake (rs855791 AA genotype) were compared with infants with high iron uptake (rs855791 GG genotype, Table 3).

Table 3 Outcome data at 5 years stratified to the TMPRSS6 rs855791 genotype
Full size table

Discussion

Here, we report that genetically determined low intestinal iron uptake (and consecutive higher intra-intestinal iron level) is associated with NEC in preterm infants. This finding is in line with a number of observational and experimental data supporting the hypothesis that high levels of intestinal iron might be a cofactor for the early development of gut dysbiosis and NEC.

Iron is essential for the growth of most bacterial species. A large-scale trial of oral iron supplementation was stopped because of excess morbidity and mortality due to infection in preschool children in a high malaria-transmission setting (9). Another recently published randomized controlled trial provides an explanation for this finding. It shows that oral iron supplementation adversely affects the gut microbiome by selectively favoring the growth of pathogens including Clostridium perfringens and Escherichia coli in Kenyan infants (3). These data indicate that intestinal iron levels might influence the development of local intestinal dysbiosis in the small intestine, preceding NEC (2).

The median time of NEC surgery in our study was day 18 of life, which is often before the point in time at which oral iron supplementation is introduced. Furthermore, randomized controlled trials of enteral iron supplementation have not reported adverse outcomes or increased NEC incidence in infants supplemented with iron (10, 11, 12).

Concerning the prophylactic effect of probiotics, our data suggest that certain probiotic species, such as the L. acidophilus/B. infantis preparation, which is commonly used in German neonatal intensive care units, might be able to reduce the iron supply for pathogenic strains in the small intestine. The combined use of both species might be of particular importance, as Lactobacillus species have little or no requirement for iron, but are rapidly overgrown if increasing amounts of iron are available for competing bacterial strains (13). Another iron-mediated protective effect of probiotics is suggested by recently published data from germ-free mice, indicating that intestinal cells favor iron storage after colonization with probiotics (14).

Formula nutrition of preterm infants is associated with an increased risk of NEC (15, 16). The low lactoferrin content and relatively higher iron content of formula nutrition might be a double risk, as both factors will increase intestinal iron. Treatment with oral lactoferrin reduces the risk of NEC and sepsis in preterm infants (17). The molecular mechanisms of these protective effects are unknown, as lactoferrin is a multifunctional protein that enhances iron uptake but has additional antibacterial, antiviral, and anti-inflammatory activities (18).

Both low and high serum iron levels are described as risk factors for long-term neurocognitive dysfunction in the literature (19). Long-term outcome data of our study are reassuring with regard to the current standard of nutritional iron supplementation in preterm infants. However, 5-year follow-up in our study is yet limited to a relatively small group of 946 infants.

The three polymorphisms, which were evaluated in our study, are associated with serum iron and ferritin levels in adults (6, 20). Although the polymorphisms are known determinants of iron status for years, the exact mechanism how these genetic variants enhance iron uptake is not clarified yet and pleiotrophic effects of these polymorphisms on other pathways cannot be excluded (21, 22, 23). A recent publication demonstrating that anemia is a risk factor for the development of NEC in preterm infants (24) is of particular importance with regard to pleiotrophic effects of the variants studied here, as the TMPRSS6-G-allele is associated with increased hemoglobin levels (25). However, effect sizes of single polymorphisms on complex traits like hemoglobin levels are rather small (25). A Mendelian randomization study targeting high and low hemoglobin levels and results of large-scale randomized trials of transfusion thresholds (NCT01393496) will probably be more informative with regard to the question whether anemia is a risk factor for NEC. No data concerning the interaction between specific alleles and infant iron or ferritin levels are published so far. Therefore, assumptions concerning serum iron and intestinal iron uptake in our study are entirely based on adult data, which is a major limitation of our study, at least with regard to short-term outcome data. Another limitation is the possible effect of the maternal genotype. Recently published data indicate that maternal HFE polymorphism influences umbilical cord blood lead levels. Therefore, iron transfer to the infants might also be influenced by the maternal genotype (26). Finally, the main determinant of iron uptake in our study was the TMPRSS6 rs855791 polymorphism. As spurious association is always a matter of concern, our data with regard to NEC should be confirmed by additional studies of similar size and set-up.

In summary, polymorphisms inducing low intestinal iron uptake in adults were associated with an increased rate of NEC in preterm infants. No increased risk was observed in infants receiving L. acidophilus/B. infantis probiotics. Long-term outcome data of infants with high iron uptake did not differ in infants with low or intermediate uptake, supporting the current practice of oral iron supplementation in preterm infants, which is not associated with NEC. Our data indicate that polymorphisms reducing intestinal iron uptake in adults might be an underestimated risk factor for very early intestinal dysbiosis and NEC (6).