Clinical features of WMI and nWMI groups

From April 2022 to December 2022, the Affiliated Hospital of Guangdong Medical University enrolled a total of 83 subjects who met the inclusion criteria. The study finally included 71 cases after excluding 4 deaths, 6 patients who failed to complete MRI + DTI testing, 1 case lost to treatment, and 1 case of inherited metabolic disease. Our sample finally included 23 cases of WMI and 48 cases of nWMI. The incidence of WMI in preterm infants was 32.3% (23/71). The WMI group included 12 cases (16.9%) of mild WMI, 7 cases (9.8%) of moderate WMI, and 4 cases (5.6%) of severe WMI. The incidence of moderate and severe WMI was 15%. The GA of the WMI group was 30.0 ± 1.8 weeks, while that of the nWMI group was 29.9 ± 1.7 weeks; the difference was not statistically significant. No significant differences in birth weight, sex, prenatal hormone levels, prenatal antibiotic use, APGAR score and postnatal antibiotic use were found between the two groups (P > 0.05), as shown in Table 1.

Table 1 Demographic characteristics and clinical data for the participantsDifferences in ADC and FA values between WMI and nWMI groups

No significant difference was found in FA and ADC values of ROIs in different areas of the left and right cerebral hemispheres. Therefore, these values were averaged before comparison of the two groups. Significant differences were identified in ADC values in 3 ROIs including the occipital white matter, paraventricular white matter, and splenium of corpus callosum between the WMI and nWMI groups (Table 1s, Fig. 1A). Comparison of the FA values of ROIs in different areas of the two groups revealed significant differences in frontal white matter, paraventricular white matter, and splenium of corpus callosum (Table 2s, Fig. 1B). These results indicated that brain myelination was delayed in the WMI group, especially in the paraventricular white matter and the splenium of corpus callosum.

Fig. 1

Comparison of DTI values in ROIs between WMI and nWMI group. A Comparison of ADC values in ROIs between WMI and nWMI group. B Comparison of FA values in ROIs between WMI and nWMI group. *P < 0.05, **P < 0.01

Comparison of gut microbiota richness and diversity

The microbial community richness and community evenness were evaluated via alpha diversity analysis. The alpha diversity of preterm infants gradually decreased with age and stabilized at 4 weeks after birth (Fig. 2). As shown in Fig. 2A-C, the Ace, Chao, and Shannon indices of the WMI1 group were the highest in the WMI group, with statistically significant differences. In addition, the Shannon index of the WMI1 group was significantly higher than that of nWMI1 (P = 0.0035), while the Simpson index was significantly lower than that of the nWMI1 group (P = 0.002) (Fig. 2A-D). Significant differences were found in the diversity indices including Shannon and Simpson indices between WMI and nWMI groups, indicating that the diversity of WMI was higher than in children with nWMI (Fig. 1S).

PCoA based on Bray-Curtis distance showed that the structures of WMI1 and nWMI1 group were significantly separated at the genus level (R2 = 0.084, P = 0.02; Fig. 2E). Both the WMI and nWMI groups showed significant structural differences in meconium sample flora on days 14 and 28 (Fig. 2F-H). PCA analysis further revealed significant differences in overall microbial diversity between patients with WMI and nWMI controls (R2 = 0.027, P = 0.022; Fig. 1S). WMI samples were more dispersed compared with clustered nWMI samples, indicating the diversity in the composition of the bacterial community.

Fig. 2

Comparison of alpha diversity and beta diversity index. (AD) Comparison of alpha diversity index, Ace (A), Chao (B), Shannon (C) and Simpson (D) index of WMI1, WMI14, WMI28, nWMI1, nWMI14, and nWMI28 at the genus level. E The gut microbiota composition was significantly different at the genus level between WMI1 and nWMI1(PCoA). F There were significant differences in the composition of gut microbiota among six groups (PCoA). G There were significant differences in the composition of gut microbiota at the genus level among nWMI1, nWMI14 and nWMI28 groups (PCoA). H There were significant differences in the composition of gut microbiota at the genus level among WMI1, WMI14 and WMI28 groups (PCoA). Differences between groups were compared using the Wilcoxon rank sum test. *P < 0.05, **P < 0.01, ***P < 0.001

Comparison of gut microbiota structure in feces

The relative abundances of OTUs in the top 10 phyla are shown (Fig. 3A). Firmicutes and Proteobacteria were the main phyla in the WMI and nWMI groups, followed by Actinobacteria and Bacteroidetes (Fig. 3A). Compared with WMI1, the relative abundance of Bacteroides, Fusobacteriota, and Verrucomicrobia was significantly lower in WMI14 and WMI28 (Kruskal-Wallis rank sum test, Fig. 2S). The results showed that the levels of Bacteroidetes, Actinobacteria and Verrucomicrobia in the WMI1 group were significantly higher than in the nWMI1 group (Wilcoxon rank test, Fig. 2S). The relative abundance of Actinobacteria and Bacteroidetes in the WMI group was higher than in the nWMI group, and the difference was significant (Wilcoxon rank sum test, P values, 0.0219 and 0.0356, respectively; Fig. 3B). The relative abundance of Cyanobacteria was lower than in the nWMI group.

At the class level, compared with WMI14 and WMI28, the relative abundance of Clostridia, Bacteroidia and Alphaproteobacteria in WMI1 was significantly higher than in WMI14 and WMI28 (Fig. 2S). The relative abundance of Clostridia, Bacteroides, and Actinobacteria in WMI1 was higher than in nWMI1 group based on Wilcoxon rank sum test, and the difference was statistically significant (Figs. 2S and 3C). Compared with the nWMI group, the relative abundance of Actinobacteria and Bacteroides was significantly increased in the WMI group (Fig. 3D).

According to the Wilcoxon rank sum test, compared with nWMI1, the relative abundance of Klebsiella and Parabacteroides in the WMI1 group increased significantly, and the difference was statistically significant (Figs. 2S and 3E). Compared with nWMI14, the relative abundance of Staphylococcus in the WMI14 group was significantly increased, and the difference was statistically significant (Figs. 2S and 3E). At the genus level, the abundance of Staphylococcus and Acinetobacter in WMI14 was significantly higher than in WMI1 and WMI28 (Figs. 2S and 3E). The relative abundances of Staphylococcus, Bifidobacterium, Acinetobacter and Lactobacillus in the WMI group were higher than in the nWMI group, and the difference was statistically significant (Fig. 3F).

Compared with nWMI1, the relative abundance of Klebsiella pneumoniae and Parabacteroides distasonis_ATCC_8503 in the WMI1 group increased significantly, and the difference was statistically significant (Figs. 2S and 3G). Compared with nWMI14, the relative abundance of Staphylococcus caprae in the WMI1 group increased significantly, and the difference was statistically significant (Figs. 2S and 3G). At the species level, 153 species were significantly different between the WMI and nWMI groups; 149 species were enriched in the WMI group, including Staphylococcus caprae, Acinetobacter johnsonii, and Parabacteroides distasonis_ATCC_8503, while 4 species were decreased, including Bifidobacterium longum (Fig. 3H).

Fig. 3

16S rRNA gene sequencing analysis. A The histogram shows six groups of dominant species at the phylum level. B Histogram showing differences in WMI and nWMI at the phylum level. C The histogram shows six groups of dominant species at the class level. D Histogram showing the difference between WMI and nWMI at the class level. E The histogram shows six groups of the top 15 dominant species at the genus level. F Histogram showing the difference between WMI and nWMI at the genus level. G Histogram shows six groups of the top 15 dominant species at the species level. H Histogram showing the difference at the species level between WMI and nWMI. *P < 0.05, **P < 0.01, ***P < 0.001

LEfSe analysis

According to the LDA score, the abundance of c_Bacteroidia, p_Bacteroidota, c_Clostridia, o_Bacteroidales, g_Klebsiella and s_Klebsiella pneumoniae in WMI1 was higher than in nWMI1 (LDA > 4, Fig. 4A). After 2 weeks of birth, f_Staphylococcus, o_Staphylococcales, g_Staphylococcus, and s_Staphylococcus caprae were significantly enriched in WMI14 patient samples compared with nWMI14 based on LEfSe analysis (LDA > 4, Fig. 4B). According to the LDA score, the dominated bacteria in the WMI group were g_Staphylococcus, s_Staphylococcus_caprae, f_Staphylococcaceae, o_Staphylococcales, p_Bacteroidota, p_Actinobacteriota and g_Acinetobacter (Fig. 4C).

Fig. 4

LefSe analysis. A Histogram of LDA values in WMI1 and nWMI1 groups. B Histogram of LDA values in WMI14 and nWMI14 groups. C Histogram of LDA values in WMI and nWMI groups

Analysis of differential metabolites in WMI and nWMI groups

Significantly different metabolites were selected based on the variable importance in the projection (VIP) obtained from the Orthogonal Partial Least Squares Discriminant Analysis (OPLS-DA) model and the P value of the t test. Metabolites with VIP > 1 and P < 0.05 were considered significantly different metabolites. A total of 139 potential metabolic markers were identified between WMI and nWMI groups; 32 metabolites were significantly upregulated in the WMI group, while 107 metabolites were significantly downregulated (VIP > 1 and P < 0.05; Fig. 5A). A total of 184 significantly different metabolites were identified compared with the nWMI1 group, with the levels of 81 increased and 103 decreased in the WMI1 group (Fig. 5B). A total of 43 significantly different metabolites were identified, including 14 increased and 29 decreased compounds in the WMI14 group compared with the nWMI14 group (Fig. 5C). A total of 254 significantly different metabolites were identified in the WMI28 group compared with the nWMI28 group, with the levels of 44 increased and 210 decreased (Fig. 5D). The significantly upregulated metabolites included mainly lipids and lipid-like molecules, and organic compounds that contain oxygen in the WMI group. The significantly downregulated metabolites included mainly organic acids and derivatives. Clear differences were found in metabolites between the WMI and nWMI groups (Fig. 5E-H).

Fig. 5

Significant difference analysis of metabolites. A In positive and negative ion modes, the volcano plot shows the overall distribution of significant difference metabolites in WMI and nWMI group. B Volcano plot showing different metabolites in the WMI1 and nWMI1 groups. C Volcano plot showing different metabolites in the WMI14 and nWMI14 groups. D Volcano plot showing significantly different metabolites in the WMI28 and nWMI28 groups. E PLS-DA plot of metabolite differences between WMI and nWMI group. F PLS-DA plot of metabolite differences between WMI1 and nWMI1 group. G PLS-DA plot of metabolite differences between WMI14 and nWMI14 group. H PLS-DA plot of metabolite differences between WMI28 and nWMI28 group

Functional indicators of the faecal metabolome

Compound classification (lipids) of differential metabolites annotated by KEGG, FA01 Fatty Acids and Conjugates, ST04 Bile acids and derivatives, and ST02 steroids in WMI and nWMI groups were the most (Fig. 6A). Among the functional pathways annotated by KEGG, Amino acid metabolism, Biosynthesis of other secondary metabolites and Lipid metabolism pathway were the most significant (Fig. 6B). KEGG enrichment analysis revealed statistically significant differences in 17 metabolic pathways between WMI and nWMI groups (adjusted P value < 0.05). Compared with nWMI group, taurine and hypotaurine metabolism, arginine biosynthesis, phenylalanine, tyrosine and tryptophan biosynthesis, cyanoamino acid metabolism, and primary metabolic pathways such as primary bile acid biosynthesis were downregulated in the WMI group (Fig. 6C, D). These results indicated that metabolomics and related pathways were significantly altered in WMI compared with nWMI.

Fig. 6

Changes in KEGG metabolic pathways and functions. A Compound classification (lipids) of differential metabolites in WMI and nWMI groups. B Major functional pathways of differential metabolites in WMI and nWMI groups. C KEGG metabolic pathway enrichment map (histogram) of differential metabolites in WMI and nWMI groups. D KEGG metabolic pathway enrichment map (bubble plot) of differential metabolites in WMI and nWMI groups

Integration of 16S rRNA genes and metabolomes

As shown in Fig. 7A, a general positive correlation was found between the WMI-characteristic Bacteroidetes and metabolites such as didesethyl flurazepam, cinobufagin, N-acetylneuraminic acid and adenosine 3’-monophosphate, and a negative correlation with metabolites such as cholic acid and allocholic acid (correlation coefficients were − 0.6135 and − 0.6253; P < 0.0001). Actinobacteria was positively correlated with cyclocalamin and crocin 4 (Fig. 7A).

As shown in Fig. 7B, the WMI group showed a negative correlation between Staphylococcus and cholic acid, allocholic acid, and 1,3-butadiene levels. The microbiota in the WMI group including Acinetobacter showed a positive correlation with the levels of cinobufagin, didesethyl flurazepam, N-acetylneuraminic acid, and adenosine 3’-monophosphate, and other metabolites; however, they were negatively correlated with cholic acid and allocholic acid (Fig. 7B).

The LEfSe analysis revealed that the WMI group was enriched in Staphylococcus species such as S. caprae, members of the phylum Bacteroidota, and Acinetobacter species. In the WMI group, metabolites such as didesethylflurazepam, cinobufagin, N-acetylneuraminic acid, and adenosine 3’-monophosphate were significantly upregulated, while cholic acid, allocholic acid, and 1,3-butadiene were significantly downregulated. Notably, Staphylococcus may affect WMI by downregulating metabolites such as cholic acid, allocholic acid, and 1,3-butadiene. However, members of phylum Bacteroidota and Acinetobacter species affect WMI by upregulating the levels of didesethylflurazepam, cinobufagin, N-acetylneuraminic acid, and adenosine 3’-monophosphate, while downregulating metabolites such as cholic acid and allocholic acid. The results suggested that patients with WMI carry a significantly dysregulated gut microbiota, which may lead to marked alterations in metabolomics.

Fig. 7

Spearman correlation analysis of gut microbiota abundance and metabolites. A Correlation analysis between gut microbiota and metabolites at the phylum level. B Correlation analysis between gut microbiota and metabolites at the genus level. *P < 0.05, **P < 0.01, ***P < 0.001

Correlation between DTI values and gut microbiota

The results showed that Escherichia-Shigella was positively correlated with the ADC value of splenium of corpus callosum (Fig. 8A), while Blautia species were positively correlated with the ADC value of frontal white matter (Fig. 8A). Escherichia-Shigella was negatively correlated with the FA value of periventricular white matter (Fig. 8B), while Klebsiella was negatively correlated with the FA value of occipital white matter (Fig. 8B). Therefore, Escherichia-Shigella may be related to the splenum of corpus callosum and periventricular white matter; Blautia may be related to brain damage associated with frontal white matter; and_Klebsiella is related to brain damage involving occipital white matter.

The heatmap showed that Bifidobacterium longum was negatively correlated with occipital white matter ADC value (Fig. 8C). K._pneumoniae was negatively correlated with the FA value of occipital white matter (Fig. 8D). Therefore, K._pneumoniae is related to brain damage of occipital white matter. B. longum may have a protective effect on occipital white matter, suggesting a novel therapeutic role of probiotics in WMI.

Fig. 8

Correlation analysis between ADC and FA values and differential gut microbiota. A At the genus level, the heatmap of the correlation between ADC value and gut microbiota. B At the genus level, the heatmap of the correlation between the FA value and gut microbiota. C At the species level, the heatmap of the correlation between the ADC value and gut microbiota. D At the species level, the heatmap of the correlation between the FA value and gut microbiota. *P < 0.05, **P < 0.01

Correlation between DTI values and differential metabolites

Based on Spearman correlation analysis, we found that cinobufagin and fumagillin were positively but weakly correlated with the ADC of genu of corpus callosum (Fig. 9A). Cyclocalamin was positively correlated with the ADC values of parietal white matter, anterior limb of internal capsule, and posterior limbs of internal capsule (Fig. 9A). However, N-docosahexaenoyl cysteine was negatively correlated with the ADC of anterior limb of internal capsule and ADC of posterior limbs of internal capsule (Fig. 9A).

Cyclocalamin and parietal white matter FA, anterior limb of internal capsule, and posterior limbs of internal capsule showed a negative correlation (Fig. 9B). Fumagillin, cinobufagin, and crocin 4 were negatively correlated with parietal white matter FA. N-Docosahexaenoyl cysteine was positively correlated with the FA values of parietal white matter, periventricular white matter, anterior limb of internal capsule, and posterior limbs of internal capsule.

Cinobufagin, fumagillin, cyclocalamin, isoaustin, crocin 4 and other metabolites were positively correlated with ADC values in different regions of white matter in the brain, and negatively correlated with FA values, which suggest brain damage. By contrast, N-docosahexaenoyl cysteine was negatively correlated with ADC values of different regions of white matter in the brain but positively correlated with FA, suggesting protective effects. Based on the correlation between the abundance of gut microbiota and metabolites, the characteristic Acinetobacter species and Bacteroidetes of WMI group were positively correlated with metabolites such as cinobufagin. Therefore, gut microbiota such as Acinetobacter and Bacteroidetes may affect white matter structure by upregulating the levels of metabolites such as cinobufagin.

Fig. 9

Correlation between intestinal metabolites and DTI values. A Correlation between intestinal metabolites and white matter ADC values. B Correlation of gut metabolites with white matter FA values. *P < 0.05, **P < 0.01