Study population
We enrolled 34 preterm and term neonates who were hospitalized in the NICU or nursery room (T, n = 12; P, n = 22) and collected 60 stool samples at 7 and 28 days of age. After DNA extraction and amplification, we investigated 38 valid samples from 34 infants using the ONT. Illumina analyzed 25 overlapping samples from 17 infants to test the effectiveness of ONT.
The T and P groups were different in gestational age (38.8 ± 0.9 vs 32.3 ± 2.8, p < 0.05) and birth weight (3190 ± 229.4 vs 1688 ± 600.2, p < 0.05). However, there were no differences in sex, delivery mode, the start of feeding day, or current breastfeeding. T; the control group was not exposed to antibiotics. However, according to the NICU guidelines of Hanyang University Hospital, most preterm infants are exposed to antibiotics. Preterm infants were divided into two subgroups (VP: very preterm infants, n = 11; LP: moderate-to-late preterm infants, n = 11). Except for GA and birth weight, there were no significant differences in the use of antibiotics, duration of antibiotics, delivery mode, and current breastfeeding between VP and LP groups (Supplement Table 1). Baseline characteristics and neonatal outcomes are detailed in Table 1.
Table 1
Relevant clinical characteristics of infants whose gut microbiome was analyzed1,2,3
Characteristics
|
Term
(n = 12)
|
Preterm
(n = 22)
|
p
|
GA, weeks
|
38.8 ± 0.9
|
32.3 ± 2.8
|
< 0.001
|
Birth weight, g
|
3175.8 ± 229.4
|
1688.6 ± 600.2
|
< 0.001
|
Male, n
|
9 (75.0)
|
8 (72.2)
|
0.071
|
Apgar score
1-min
|
8.3 ± 1.0
|
4.2 ± 2.1
|
< 0.001
|
5-min
|
9.5 ± 0.6
|
7.0 ± 1.5
|
< 0.001
|
C-section, n
|
9 (75.0)
|
22 (100)
|
0.037
|
Antibiotics exposure, n
|
N/A
|
20 (90.9)
|
|
Breastmilk feeding, n
|
9 (75.0)
|
10 (45.5)
|
0.152
|
Hospitalization, days
|
4.9 ± 0.9
|
43.6 ± 24.9
|
< 0.001
|
1Abbreviation: GA, gestational age; C-section, cesarean section. |
2Term infants were born at a GA ≥ 37 weeks. Antibiotic exposure refers to administering antibiotics to neonates in the first 48 h of life. The delivery mode was either cesarean section or vaginal delivery. Breastmilk feeding was defined as breast milk consumption at the time of sample collection. |
3Data are presented as n (%) or mean (± SD), unless otherwise stated. |
Comparison of ONT vs Illumina sequencing performance in neonates
We sequenced 17 neonates (25 samples) from the study population with Illumina to compare ONT and Illumina sequencing data for discriminatory power, validating profiling and abundance of microbiome up to the genus level. This approach showed that the ONT sequencing depth was sufficient to capture the bacterial diversity in the neonates’ stool samples. Taxonomic assignments using ONT versus Illumina sequencing data were largely consistent with each other (log-transformed Pearson’s r = 0.906) (Fig. 2a), especially for Enterococcus, Escherichia, Klebsiella, Raoultella, Staphylococcus, and Streptococcus.
Two sequencing reads correlated well, showing that the Firmicutes phylum was the most dominant, followed by Proteobacteria in T1 and P1, Firmicutes was still dominant in T2, and Proteobacteria was increased in P2. However, the relative abundance of Actinobacteria (and the associated genus Bifidobacterium) was higher according to Illumina results, and it was also the same in T2 and P2 (T1 vs. T2; p = 0.073, P1 vs. P2; p = 0.099, respectively) (Fig. 2b, c). Although highly correlative, the total reads mapped for the ONT sequencing tended to exhibit greater read counts with a higher percentage of mapped reads compared to the results obtained using Illumina.
Taxonomic composition and abundance differences at phylum-level (T vs P)
Firmicutes phylum predominated both T1 and P1 profiles. The dominant phylum identified in T1 was Firmicutes (99.05%), followed by Actinobacteria (0.68%), Proteobacteria (0.22%), and Bacteroidetes (0.05%). In P1, Firmicutes was the most abundant (96.05%), followed by Proteobacteria (3.86%) and Actinobacteria (0.06%). No Bacteroidetes were present (Fig. 3a).
Proteobacteria and Firmicutes dominated the P2 microbiome.
Over time, Proteobacteria increased significantly in P2 (P1 vs. P2, p < 0.011), almost at the same level as Firmicutes. In contrast, Actinobacteria barely increased in abundance. There were statistically significant increases in the abundances of Firmicutes, Proteobacteria, and Actinobacteria (p < 0.001, p < 0.001, and p < 0.001, respectively) at 1 month (Fig. 3a).
Same phylum but different features at genus-level
Genus-level profiles showed different bacterial compositions in T1 vs. P1. Within the same Firmicutes phylum, Staphylococcus (40.33%) was the most abundant in T1, followed by Enterococcus (39.76%) and Lactobacillus (0.41%), whereas in P1, Enterococcus (65.58%) was dominant, followed by Staphylococcus (27.01%) and Lactobacillus (0.09%) (Fig. 3b).
Prematurity influenced the abundance of Klebsiella, Streptococcus, Lactobacillus, and Bifidobacterium genus not at 7 days but at 1 month. There were no significant differences between T and P at the genus level after seven days of life. However, at 1 month, we observed a noticeable increase in the relative abundance of Klebsiella, a potentially pathogenic bacterial member, (p = 0.002) in P2, and Streptococcus and beneficial Bifidobacterium and Lactobacillus in T2 (p = 0.001, p = 0.026, p < 0.001, respectively) (Fig. 3c).
Genus-level profiles were different in T1 and T2. Firmicutes remained the most abundant phylum from T1 to T2; however, the dominant genera changed (Staphylococcus (40.33%), Enterococcus (39.76%), and Lactobacillus (0.41%) to Streptococcus (37.55%), Lactobacillus (22.07%), and Enterococcus (13.50%)) (Fig. 3b).
Profiling up to species-level through ONT application
There were differences at the species-level, even in the same genus (T1 vs. P1). The T1 showed higher abundance of Enterococcus faecalis (p = 0.081) and Staphylococcus epidermidis (p = 0.004), whereas in P1, Enterococcus faecium (p < 0.001) and Staphylococcus hemolyticus (p = 0.034) were more abundant (Fig. 3e).
Beneficial bacteria increased in T2, whereas pathogenic bacteria in P2. Bifidobacterium longum, Lactobacillus gasseri, and Streptococcus salivarius significantly increased in T2 (p = 0.017, p = 0.001, p < 0.001, respectively), whereas in P2, Klebsiella pneumoniae was increased, but there was no statistical difference (p = 0.096) (Fig. 3e).
Gut microbiome diversity
α-diversity. Bacterial α-diversity was measured using Pielou's evenness index (microbial evenness) and Faith’s phylogenetic diversity index (species richness). There was a statistically significant increase in both evenness and diversity in T (p < 0.002 and p < 0.002, respectively), but not in P (p = 0.156 and p = 0.156, respectively) (Fig. 4a).
β-diversity. Subsequently, we evaluated the similarity of the gut microbiota (β-diversity) for both bacterial composition and phylogenetic relationships of the components. There was a significant difference in the β-diversity between T1 and T2 (p = 0.005), but no differences were observed between P1 and P2 (p = 0.050). Samples from T1 and P1 were dispersed and intermingled, suggesting that the bacterial composition of T1 was largely similar to that of P1 (p = 0.327). However, there was a statistically significant difference between T2 and P2 (p = 0.001) (Fig. 4b).
Differences in bacterial abundance among preterm infants by GA (VP vs. LP)
There were significant differences between VP and LP. To investigate the impact of the degree of prematurity on the gut microbiome of premature infants, we divided preterm infants into two groups (VP and LP). VP infants had the greatest abundance of Klebsiella pneumonia (Klebsiella genus/Proteobacteria phylum) (p = 0.014), whereas LP, later GA premature infants, had the greatest abundance of Enterococcus faecium (Enterococcus genus/Firmicutes phylum) (p = 0.008) (Fig. 5a, b, c). Across all time points, in T, Lactobacillus gasseri (Lactobacillus genus/Firmicutes phylum) (p = 0.023) and Streptococcus salivarius (Streptococcus genus/Firmicutes phylum) (p = 0.013) were significantly increased.