**Background:** Necrotizing enterocolitis (NEC) is a devastating gastrointestinal disease affecting 7–11% of very low birth weight (VLBW) preterm infants, with mortality up to 30%. Neonatal intensive care units (NICUs) have developed various NEC prophylaxis programs combining probiotics, antibiotics, and feeding protocols, resulting in substantially decreased NEC rates. However, the mechanisms by which different regimens influence the gut microbiome and metabolome had not been systematically compared. This study aimed to deeply analyze the effects of different NEC prevention strategies on the preterm infant gut microbiome across multiple microbial kingdoms and functional levels.
**Methods:** The authors conducted a prospective, triple-center cohort pilot study of 55 VLBW preterm infants (<1500 g) across three Austrian NICUs (Graz, Klagenfurt, Leoben) between October 2015 and March 2017. The centers used different NEC prophylaxis regimens: Graz (G) used Lactobacillus rhamnosus LCR 35 probiotic, enteral gentamicin, oral nystatin, and favored human milk (HM); Klagenfurt (K) used Bifidobacterium longum subsp. infantis NCDO 2203 and Lactobacillus acidophilus NCDO 1748 probiotics, intravenous fluconazole, and mainly formula milk (FM); Leoben (L) used no probiotics, enteral gentamicin, oral nystatin, and favored HM. Fecal samples were collected every other day from meconium through two weeks of age (7 time points). A total of 383 samples were analyzed using 16S rRNA gene amplicon sequencing (bacterial, archaeal, fungal), shotgun metagenomic sequencing on a subset at three time points, and NMR-based metabolomics on 111 samples. Metagenome-assembled genomes (MAGs) were generated, and functional profiling, antibiotic resistance gene analysis, and virulence factor analysis were performed.
**Key Results:** The three regimens drove distinct center-specific microbiome profiles across all microbial domains. At time point 7 (days 13–21), Klagenfurt samples were dominated by Bifidobacterium at >82% relative abundance, while Graz and Leoben were dominated by Enterococcus (~77%). PERMANOVA confirmed that the combined variables of Bifidobacterium administration, feeding protocol, and gentamicin administration explained similar variance (R²=0.6763 at tp7, p=0.001) as hospital grouping. B. longum subsp. infantis NCDO 2203 (the administered probiotic in K) accounted for 95% of all Bifidobacterium reads and showed evidence of active replication (iRep values >1), while co-administered L. acidophilus NCDO 1748 reached only 0.21% relative abundance versus 75.69% for B. infantis. In Graz, L. rhamnosus LCR 35 reached 21% relative abundance (seven-fold increase over natural levels). Archaeal signatures were detected in all infants, with Methanobrevibacter present across all time points. Fungal and phage profiles also showed center-specific patterns. Functionally, Klagenfurt samples showed significantly reduced genes for osmotic stress (K:G q<0.001, K:L q<0.001), acid stress (K:G q=0.033, K:L q=0.029), and respiration (K:G q=0.007, K:L q=0.014), but increased polysaccharide metabolism genes (K:G q<0.001, K:L q<0.001). HMO gene clusters were seven-fold higher in K (307 hits) versus G (45 hits) and L (0 hits). Metabolomics confirmed significantly lower levels of most fucosylated HMOs in K samples at tp7. Antibiotic resistance gene hits were substantially lower in K (35 hits) versus G (93 hits) and L (123 hits). Virulence factors were also reduced in K (64 hits) versus G (193 hits) and L (173 hits). NEC incidence rates were 2.2% in K, 2.7% in G, and 4.6% in L, all well below the global average.
**Clinical Implications:** This study demonstrates that different NEC prophylaxis regimens profoundly shape the preterm infant gut microbiome across all microbial kingdoms. The regimen combining B. infantis NCDO 2203 with formula milk (Klagenfurt) achieved the highest Bifidobacterium dominance and lowest antibiotic resistance gene and virulence factor carriage, but the potential benefit of HMO metabolism was limited by formula feeding (which lacks HMOs). The authors suggest that combining B. infantis NCDO 2203 with human milk feeding may represent an optimal synergistic approach for NEC prevention. The study also highlights that enteral gentamicin administration (used in Graz and Leoben) did not show consistent negative correlations with specific taxa or induce detectable gentamicin resistance at tp7. Notably, one infant who later developed NEC (K16) showed early E. coli bloom with an active MAG at tp7, suggesting potential for microbiome-based NEC monitoring. The study is limited by its small cohort size, short survey period (first weeks of life), and inability to disentangle individual factors due to the multi-component nature of each center's regimen.