Study population
Fecal samples were collected between February 2013 and October 2017 from women enrolled in the ENDIA study, a prospective, pregnancy-birth cohort study that follows 1500 Australian children who have a first-degree relative with T1D [23]. Thirty-five women (36 pregnancies) with T1D and 31 women (34 pregnancies) without T1D had each provided from one to three fecal samples across pregnancy (total 134 samples) for analysis by shotgun WMS (Fig. 1). Table 1 summarizes and compares characteristics of the T1D and non-T1D pregnancies.
Table 1
Summary of characteristics of non-T1D and T1D pregnancies
General | Non-T1D | T1D | P-value ** |
Overall number of samples: n (%) | 66 (49.3) | 68 (50.7) | |
Trimester 1 | 11 (16.7) | 12 (17.6) | |
Trimester 2 | 21 (31.8) | 22 (32.4) | |
Trimester 3 | 34 (51.5) | 34 (50.0) | |
All three trimesters (% pregnancies) | 12 (35.3) | 12 (33.3) | |
All three trimesters (% samples) | 12 (18.2) | 12 (18.2) | |
Gestational age in days at fecal sample: mean (SD) |
Trimester 1 | 75.4 (16.5) | 75.3 (16.5) | |
Trimester 2 | 150.8 (26.6) | 148.4 (26.6) | |
Trimester 3 | 247.6 (14.6) | 234.7 (14.6) | 0.001 |
Maternal |
Overall number of pregnancies | 34 | 36 | |
Age in years at conception: mean (SD) | 33 (4.1) | 32.3 (4.0) | |
Paternal missing n (%) | 1 (2.9) | 2 (5.6) | |
Assisted conception: n (%) | 3 (8.8) | 4 (11.1) | |
Twin pregnancy: n (%) | 0 (0.0) | 0 (0.0) | |
Nulliparous: n (%) | 14 (41.2) | 18 (50.0) | |
Pre-eclampsia: n (%) | 0 (0.0) | 4 (11.1) | 0.018 |
Group B Streptococcus positive: n (%) | 7 (20.6) | 1 (2.8) | |
Genito-urinary infections: n (%) | 2 (5.9) | 5 (13.9) | |
Pre-pregnancy BMI: mean (SD) | 24.7 (4.8) | 25.8 (4.7) | |
Underweight (< 18.5): n (%) | 0 (0.0) | 0 (0.0) | |
Normal weight (18.5–24.9): n (%) | 21 (61.8) | 18 (50.0) | |
Overweight weight (25-29.9): n (%) | 6 (17.6) | 10 (27.8) | |
Obese (> 30): n (%) | 7 (20.6) | 8 (22.2) | |
Gestational weight gain (kg): Mean (SD) | 13 (5.0) | 11.5 (4.9) | |
Gestational weight gain (kg): n (%) | 2 (5.9) | 2 (5.6) | |
Paternal | | | |
Age in years at conception: mean (SD) | 34.8 (5.2) | 32.7 (4.8) | 0.093 |
Pre-pregnancy BMI: mean (SD) | 28.6 (4.3) | 27.6 (4.3) | |
Underweight (BMI < 18.5): n (%) | 0 (0.0) | 1 (2.8) | |
Normal weight (BMI18.5-24.9): n (%) | 4 (11.8) | 6 (16.7) | |
Overweight (BMI 25-29.9): n (%) | 10 (29.4) | 8 (22.2) | |
Obese (> 30): n (%) | 6 (17.6) | 9 (25.0) | |
Missing: n (%) | 14 (41.2) | 12 (33.3) | |
Maternal demographics | | | |
Born in Australia: n (%) | | | |
Yes | 30 (88.2) | 24 (66.7) | |
Unknown | 0 (0.0) | 1 (2.8) | |
Education beyond High School: n (%) | | | |
Yes | 30 (88.2) | 29 (80.6) | |
Unknown | 0 (0.0) | 0 (0.0) | |
Lives in a metro area: n (%) | 31 (91.2) | 35 (97.2) | |
Socio-Economic Indexes for Areas (SEIFA) Index of Relative Socio-Economic Disadvantage (IRSD) |
Quintile 1 n (%) | 2 (5.9) | 1 (2.8) | |
Quintile 2 n (%) | 2 (5.9) | 3 (8.3) | |
Quintile 3 n (%) | 13 (38.2) | 8 (22.2) | |
Quintile 4 n (%) | 4 (11.8) | 9 (25.0) | |
Quintile 5 n (%) | 13 (38.2) | 15 (41.7) | |
Smoking during pregnancy: n (%) | 3 (8.8) | 0 (0.0) | |
Household smoking during pregnancy: n (%) | 5 (14.7) | 6 (16.7) | |
Adults in house during pregnancy: n (%) | | | |
One | 0 (0.0) | 2 (5.6) | |
Two | 31 (91.2) | 29 (80.6) | |
More than two | 3 (8.8) | 5 (13.9) | |
Children in house during pregnancy: n (%) | | | |
None | 14 (41.2) | 18 (50.0) | |
One | 8 (23.5) | 10 (27.8) | |
Two | 5 (14.7) | 7 (19.4) | |
More than two | 7 (20.6) | 1 (2.8) | |
Furred pet ownership during pregnancy: n (%) | 24 (70.6) | 20 (55.6) | |
Diet and physical activity in pregnancy | | | |
Diet: mean (SD) | | | |
Energy/day (kJ) | 6617.3 (2277.5) | 6445.6 (2185.9) | |
Fat (g) | 68.8 (27.0) | 71.7 (26.8) | |
Protein (g) | 77.1 (29.9) | 80.3 (29.6) | |
Carbohydrate (g) | 163.8 (57.9) | 142.4 (52.6) | |
Fiber (g) | 18.4 (6.1) | 17.9 (5.9) | |
Diet: Missing: n (%) | 0 (0.0) | 2 (5.6) | |
Alcohol consumed: n (%) | | | |
Yes | 6 (17.6) | 7 (19.4) | |
Unknown | 0 (0.0) | 2 (5.6) | |
Metabolic equivalent of task (MET) (h/wk): mean (SD) | 254.5 (100.9) | 267.9 (102.2) | |
Biological data | | | |
HbA1c (%) | | | |
Trimester 1: median (IQR) | -- | 6.8 (1.6) | |
Trimester 2: median (IQR) | -- | 6.1 (1.3) | |
Trimester 3: median (IQR) | -- | 6.1 (0.8) | |
Trimester 1: missing | -- | 1 (8.3) | |
Trimester 2: missing | -- | 3 (13.6) | |
Trimester 3: missing | -- | 14 (41.2) | |
1,5-anhydroglucitol (AG) (ug/mL) | | | |
Trimester 1: median (IQR) | 14.1 (13.1) | 3.4 (1.5) | |
Trimester 2: median (IQR) | 11.5 (4.9) | 2.5 (2.3) | |
Trimester 3: median (IQR) | 8.1 (6.3) | 2.4 (1.3) | |
Trimester 1: mean (SD) | 14.1 (5.9) | 3.4 (2.9) | < 0.001 |
Trimester 2: mean (SD) | 11.2 (5.1) | 2.3 (2.2) | < 0.001 |
Trimester 3: mean (SD) | 8.7 (3.8) | 2.4 (2.0) | < 0.001 |
Trimester 2: missing n (%) | 1 (4.8) | 1 (4.5) | |
Trimester 3: missing n (%) | 2 (5.9) | 8 (23.5) | |
Serum vitamin D (nmol/L): mean (SD) | | | |
Trimester 1 | 83 (26.9) | 76.7 (25.8) | |
Trimester 2 | 96.7 (27.0) | 85.5 (24.5) | |
Trimester 3 | 92.9 (31.6) | 96.4 (29.9) | |
Trimester 1: missing n (%) | 0 (0.0) | 1 (8.3) | |
Trimester 3: missing n (%) | 2 (5.9) | 3 (8.8) | |
Vitamin B6 (nmol/L): mean (SD) | | | |
Trimester 3 | 76 (75.8) | 70 (102.4) | |
Vitamin B12 (nmol/L): mean (SD) | | | |
Trimester 3 | 84 (116) | 154 (138) | |
Maternal HLA: n (%) | | | |
DR34 | 3 (8.8) | 14 (38.9) | |
DR3 or DR4 | 20 (58.8) | 19 (52.8) | |
DRXX | 11 (32.4) | 3 (8.3) | 0.002 |
Known supplements in pregnancy | | | |
Antibiotics: n (%) | 9 (26.5) | 10 (27.8) | |
Anticoagulants: n (%) | 3 (8.8) | 6 (16.7) | |
Antihypertensive agents: n (%) | 0 (0.0) | 4 (11.1) | |
Known other supplements pre-pregnancy and pregnancy |
Biotin: n (%) | 12 (35.3) 29 (85.3) | 9 (25.0) 30 (83.3) | |
Calcium: n (%) | 13 (38.2) 30 (88.2) | 10 (27.8) 33 (91.7) | |
Iron | 14 (41.2) 32 (94.1) | 9 (25.0) 35 (97.2) | |
Magnesium: n (%) | 14 (41.2) 31 (91.2) | 9 (25.0) 32 (88.9) | |
Selenium: n (%) | 12 (35.3) 29 (85.3 | 9 (25.0) 30 (83.3) | |
Vitamin B1: n (%) | 14 (41.2) 32 (94.1) | 9 (25.0) 32 (88.9) | |
Vitamin B2: n (%) | 14 (41.2) 31 (91.2) | 9 (25.0) 32 (88.9) | |
Vitamin B3: n (%) | 14 (41.2) 31 (91.2) | 9 (25.0) 32 (88.9) | |
Vitamin B5: n (%) | 9 (26.5) 18 (52.9) | 6 (16.7) 16 (44.4) | |
Vitamin B6: n (%) | 14 (41.2) 32 (94.1) | 9 (25.0) 32 (88.9) | |
Vitamin B9 (folate): n (%) | 14 (41.2) 32 (94.1) | 15 (41.7) 36 (100.0) | |
Vitamin B12: n (%) | 14 (41.2) 32 (94.1) | 9 (25.0) 30 (83.3) | |
Vitamin D: n (%) | 14 (41.2) 33 (97.1) | 10 (27.8) 33 (91.7) | |
Vitamin E: n (%) | 13 (38.2) 28 (82.4) | 8 (22.2) 27 (75.0) | |
OTHER | | | |
Vaccine: n (%) | | | |
Yes (Flu only) | 1 (2.9) | 3 (8.3) | |
Yes (Pertussis only) | 3 (8.8) | 3 (8.3) | |
Yes (Flu and Pertussis) | 11 (32.4) | 10 (27.8) | |
Mode of delivery: n (%) | | | |
Vaginal | 25 (73.5) | 14 (38.9) | |
Caesarean (with labour) | 1 (2.9) | 4 (11.1) | |
Caesarean (without labour) | 8 (23.5) | 18 (50.0) | 0.011 |
Log transformation was used for Age at conception: Paternal |
Square root transformation was used for all diet variables except carbohydrate and fibre |
Square root transformation was used for 1,5-AG in all trimesters and for vitamin D in trimester 2. |
Hb1A-c, 1,5-AG and vitamin D are based on samples, not pregnancies. |
** Blank cells indicate P value non-significant. P values for HLA are determined against DRXX as baseline. |
NM = not measured |
Whole metagenomic sequencing
The WMS dataset, 47,766,763 ± 10,956,057 (mean ± SD) paired-end reads per sample, was obtained using an Illumina NovaSeq 6000. Raw reads (SRA accession: PRJNA604850) were pre-processed using KneadData bioBakery tool [24] to eliminate human DNA sequences and filter sequences with poor quality which on average removed 6% of the reads. After quality control and read filter steps, 44,940,628 ± 10,572,188 (mean ± SD) paired-end reads per sample were obtained (Excel file E0).
Taxonomic diversity and composition of the gut microbiome in women with and without T1D during pregnancy
Sequences were analysed with MetaPhLan2 implemented within the HUMAnN2 pipeline. Overall, 340 bacterial species were identified, with an average of 93 ± 13 (mean ± SD) species per sample. The top 25 most abundant species accounted for more than 50% of the gut microbiome composition of each subject in any given trimester (Figure S1).
Alpha diversity (observed richness or number of species) per sample was calculated and generalized estimating equations (GEE) were applied to test for differences between women without and with T1D, and between trimesters, and to determine if there was an interaction between T1D status and trimester. No differences were found in richness due to T1D status or time, or interactions (Figure S2, Excel file E1).
For analysis of beta diversity, Bray-Curtis coefficients were calculated between sample pairs, ordinated and plotted by principal coordinate analysis (PCoA) for each taxonomic level (Figs. 2, S3 and S4). To test for differences in beta diversity, a repeated-measure aware permutational analysis of variance (RMA-PERMANOVA) of the Bray-Curtis coefficients was performed on proportional log transformed data. This revealed a significant interaction between T1D status and time at all taxonomic levels. Therefore, differences between women with and without T1D were assessed within trimesters. No significant differences were detected in trimesters 1 and 2. However, differences were significant at the strain (P = 0.002), species (P = 0.001), genus (P = 0.070) and family (P = 0.034) levels in trimester 3 (Excel file E2).
To rule out the possibility that these results were influenced by the difference in sample size between trimesters 1 and 3, we performed a sensitivity analysis by subsampling trimester 3 to the size of trimester 1 (n = 23), using samples of trimester 3 from the same women in trimester 1, and repeated the beta diversity analysis. Similar to the complete trimester 3 dataset, differences were significant at the strain (P = 0.003), species (P = 0.003), genus (P = 0.043) and family (P = 0.047), but also phylum (P = 0.09), taxonomic levels (Excel file E2).
Differences in beta diversity reflect differences in taxonomic composition. To identify differences in specific taxa between women with and without T1D in pregnancy, differential abundance was analysed in limma. Only taxa for which the prevalence (i.e. proportion of samples with those taxa) was above 50% in at least one group and with a log2 fold-change (logFC) greater than 0.5 or less than − 0.5 were considered. Across all trimesters, the species Bacteroides caccae (FDR 0.03) and its unique strain (unclassified) in the dataset (FDR 0.03), as well as the order Enterobacteriales (FDR 0.07) were increased in women with T1D (Fig. 3; Excel file E3). On the other hand, species Bacteroidales bacterium ph8 (FDR 0.034) and its strain (GCF000311925) (FDR 0.03), the genus (FDR 0.08) and family (FDR 0.08) to which Bacteroidales bacterium ph8 belongs, and the order Bifidobacteriales (FDR 0.07), were decreased in women with T1D (Fig. 3; Excel file E3).
Differences between women with and without T1D were also assessed within trimesters. In trimesters 1 and 2, taxa were not significantly different. However, several differences were found in trimester 3, in which the unique strain (unclassified) of Bacteroides caccae (FDR 0.004), the species Bacteroides caccae (FDR 0.004), the species Bacteroides vulgatus (FDR 0.04) and its unique strain (unclassified) (FDR 0.04) and Bacteroides uniformis (FDR 0.04) were increased in women with T1D, while the species Bacteroidales bacterium ph8 (FDR 0.01) and its strain (GCF000311925; FDR 0.005), and the genus (FDR 0.08) and family (FDR 0.08) of Bacteroidales bacterium ph8 and the order Bifidobacteriales (FDR 0.07), were decreased (Fig. 3; Excel file E3). A significant Spearman correlation (R2 > 0.4) was found between B. caccae and B. vulgatus (R2 = + 0.43; adj.P = 0.013).
A sensitivity analysis of differential abundance was also applied to the subset of trimester 3 samples referred to above: 13 species, 13 strains, 2 genera, 3 families, 2 orders and 3 phyla were detected as differentially abundant. From these, Bacteroides caccae and Bacteroides uniformis, an unclassified strain of Bacteroides caccae and the order Bifidobacteriales were also detected in the larger dataset of trimester 3 samples. Differential abundance results are summarised in Excel file E3.
In order to identify the bacterial species that were most abundant within the Enterobacteriales and Bifidobacteriales orders we plotted the average relative abundance in women with and without T1D (Figure S5A). Escherichia coli was the most abundant species within Enterobacteriales and, together with an unclassified species of the genus Escherichia, accounted for almost the complete abundance of this order. In addition, a significant Spearman correlation was found between between E. coli and Coprococcus sp. ART55_1 (R2= -0.6, adj.P = 0.09). Bifidobacterium adolescentis and Bifidobacterium longum were the most abundant species within Bifidobacteriales (Figure S5A). A lmer test applied to test differences in the abundance of these four species between women with and without T1D revealed that the abundance of E. coli in trimester 3 and of B. adolescentis in trimester 1 were significantly different between women with and without T1D (P = 0.01 for both; Figure S6).
Effect of gestation time and other factors on the gut microbiome during pregnancy
No significant differences in alpha diversity were detected in women with or without T1D according to time, analysed either by days of gestation (P-value 0.5) or by trimester (P-values > 0.6), i.e. as continuous or categorical variables, respectively (Excel file E1). Due to the significant interaction between T1D status and time, differences in beta diversity across time (days or trimesters) were assessed separately in women with and without T1D (Excel file E2). Differences were detected only at the strain (P-value 0.03) and species (P-value 0.06) levels in women without T1D with time as continuous variable (Excel file E2). However, in women with T1D, differences in beta diversity across days of gestation and between trimesters were significant at all taxonomic levels except order and phylum (Excel file E2). These observations suggested that the microbial community structure across pregnancy is less stable in women with T1D. Therefore, we sought to identify differentially abundant taxa across trimesters separately within each group.
Throughout pregnancy, in women with T1D, the abundance of an unclassified species of the family Peptostreptococcaceae (FDR 0.02), the species Odoribacter splanchnicus (FDR 0.098) the genus Prevotella (FDR 0.066) dominated by the species Prevotella copri, (Figure S5B) and the phylum Verrucomicrobia (FDR 0.043) decreased, while an unclassified strain of species Streptococcus thermophilus (FDR 0.099) and the species Streptococcus thermophilus (FDR 0.04) and family Porphyromonadaceae (FDR 0.092) increased (Excel file E4; Figure S7). In women without T1D, an Anaerostipes hadrus GCF000332875 strain (FDR 0.038) and species Anaerostipes hadrus (FDR 0.059), an unclassified strain of Haemophilus parainfluenzae (FDR 0.001) and species Haemophilus parainfluenzae (FDR 0.003), genus Haemophilus (FDR 0.004), family Pasteurellaceae (FDR 0.002), strain (GCF000218445 [FDR 0.04]) and species of Lachnospiraceae bacterium 1157FAA (FDR 0.055) and an unclassified species of Veillonella genus (FDR 0.083) decreased during pregnancy (Excel file E4; Figure S7). Furthermore, in women without T1D strains Ruminococcus sp 5139BFAA GCF000159975 and Lachnospiraceae bacterium 3157FAACT1 GCF000218405 (FDR 0.063 and 0.075, respectively) and their corresponding species (FDR 0.065 and 0.075), unclassified strains of Streptococcus thermophilus (FDR 0.06) and Bifidobacterium_animalis (FDR = 0.06) and their corresponding species (FDR 0.06 for both) increased throughout pregnancy (Excel file E4; Figure S7).
As expected, women with and without T1D differed in serum 1,5-anhydroglucitrol (1,5-AG), a marker of short-term glycemic control [25] (Table 1), but in women with T1D serum 1,5-AG was related to beta diversity only at the phylum level (Excel file E2). Mode of delivery had an effect on the beta diversity only at the family level (Excel file E2). No significant associations were found between beta diversity and age at conception, body mass index (BMI), parity, carbohydrate or fibre intake (Excel file E2). However, a difference was observed in the microbiome composition at the strain and species levels according to the human leukocyte antigen (HLA) class II type (Excel file E2). The model used to test for differences in beta diversity between women with and without T1D was adjusted for HLA type. HLA type accounted for 3.2% of the variation [R2] in beta diversity in trimester 3 (Excel file E2). After controlling for this effect, T1D status explained 2.9% of the variation and the difference in beta diversity between women with and without T1D women was statistically significant (P = 0.004) (Excel file E2). Finally, even though for the differential abundance analysis an adjustment for HLA type was included in the model, an additional analysis was performed to detect differences in the abundance of specific taxa due to HLA type and to verify that the taxa that were detected as differentially abundant due to T1D status were not affected by HLA type. Differences due to HLA type were detected only between HLADR34 and HLADR3X and DR4X for the abundance of strain Eubacterium ramulus GCF000469345 and species Eubacterium ramulus in trimester 1 and an unclassified strain of species Eubacterium rectale and species Eubacterium rectale in trimester 3, which were decreased in women with HLA DR34 Excel file E3. None of the taxa identified as differentially abundant due to T1D status were significantly affected by HLA type.
Validation of differentially abundant species by qPCR
To validate the findings from metagenomic sequencing, we analysed the relative abundance of two of the top-ranked differentially abundant bacteria, Bacteroides caccae and Bacteroides vulgatus, in the same cohort of T1D and non-T1D mothers in trimester 3. Relative abundances obtained by metagenomic sequencing and qPCR were strongly correlated (Spearman R = + 0.91 and + 0.74 for B. caccae and B. vulgatus, respectively). By fitting linear models in lmer with conception age, BMI, parity and HLA type introduced as fixed effects, and “woman ID” and processing batches as random effects, qPCR confirmed the increase in relative abundance of B. caccae (P = 0.00005) and B. vulgatus (P = 0.04) in women with T1D (Figure S8).
Functional annotation of gut microbiome taxa
Sequences processed with HUMAnN2 were annotated, complete metabolic pathways quantified, gene abundances calculated and regrouped into KO (Kegg Orthology) and MetaCyc reaction functional categories. A total of 451 complete pathways, 5,628 KO and 3,204 MetaCyc reaction categories were obtained. No significant interaction in richness was detected between factors T1D status and time. In the model in which time was considered as a continuous variable richness was significantly higher in women with T1D for all three functional categories (Figure S9, Excel file E1). For beta diversity, the interaction between T1D status and time was significant. Therefore, differences between groups were assessed within each trimester, but were significant for the three functional categories only in trimester 3 (Fig. 4; Excel file E2).
Women with and without T1D displayed significant differences in the abundance of a number of features identified in pathways, KO and MetaCyc categories; these are comprehensively listed in Supplementary Excel files E5-10. Selected functions, namely LPS production, vitamin K2 synthesis, vitamin B6 synthesis, vitamin B12 synthesis, short chain fatty acid (SCFA) synthesis and mucin degradation, and the principal bacterial species contributing to these functions, are summarized in Table 2. Examples of bacteria contributing to a functional feature are shown in Figures S10 and S11. Of interest, a pathway (PWY1269: CMP-3-deoxy-D-manno-octulosonate biosynthesis I), 17 KO gene categories and two MetaCyc reactions (DARAB5PISOM-RXN and UDPGLCNACEPIM-RXN) involved in the synthesis of bacterial lipopolysaccharides (LPSs) were enriched in women with T1D (Excel file E5-E7; Table 2; Fig. 5, Figure S10A). Seven pathways and 6 KO categories involved in vitamin K2 synthesis were also increased in women with T1D. (Excel files E5-E6; Table 2; Fig. 5; Figure S10B. In addition, two KO categories increased in women with T1D in trimester 3 were involved in antibiotic tolerance (K03771) and biofilm formation and (K18831) (Excel file E6).
Table 2
Pathways and enzymes differentially abundant in T1D women.
Function (Change ref. T1D) | Function IDs | Principal bacterial contributors * (bacterial clusters are in bold) |
LPS production (↑) | PWY-1269 K00748 K00912 K00979 K01447 K01627 K01791 K02517 K02536 K02852 K03270 K03771 K05807 K06041 K06142 K07091 K11720 K11934 DARAB5PISOM-RXN UDPGLCNACEPIM-RXN | Escherichia coli, Akkermansia muciniphila, Bacteroides caccae, Alistipes finegoldii, Bacteroides dorei, Odoribacter splanchnicus, Alistipes shahii, Bacteroides fragilis, Bacteroides vulgatus, Bacteroides faecis, Bacteroides finegoldii, Bacteroides ovatus, Bacteroides sp 2 1 22, Bacteroides sp 4 3 47FAA, Bacteroides thetaiotaomicron, Bacteroides xylanisolvens, Alistipes onderdonkii, Bacteroides stercoris, Bacteroides uniformis, Comamonas testosteroni, Enterobacter cloacae, Escherichia sp TW09276, Achromobacter xylosoxidans, Aggregatibacter aphrophilus, Azospira oryzae, Bacteroides cellulosilyticus, Bacteroides sp 1 1 6, Bacteroides sp 3 2 5, Campylobacter concisus, Campylobacter curvus, Campylobacter hominis, Chryseobacterium taeanense, Citrobacter freundii, Citrobacter koseri, Cronobacter sakazakii, Delftia acidovorans, Desulfovibrio desulfuricans, Escherichia fergusonii, Eubacterium siraeum, Haemophilus influenzae, Neisseria flavescens, Neisseria meningitidis, Neisseria subflava, Parabacteroides goldsteinii, Parabacteroides merdae, Porphyromonas asaccharolytica, Prevotella denticola, Prevotella melaninogenica, Pseudomonas nitroreducens, Pseudomonas putida, Salmonella enterica, Serratia liquefaciens, Shigella flexneri, Shigella sonnei |
Vitamin K2 synthesis (↑) | PWY-5838 PWY-5845 PWY-5850 PWY-5860 PWY-5861 PWY-5862 PWY-5896 K00330 K00334 K00338 K00340 K00343 K02523 K02548 | Escherichia coli, Akkermansia muciniphila, Bacteroides sp 4 3 47FAA, Bacteroides dorei, Bacteroides fragilis, Alistipes finegoldii, Bacteroides vulgatus, Bacteroides salyersiae, Bacteroides sp 1 1 6, Bacteroides thetaiotaomicron, Odoribacter splanchnicus, Alistipes onderdonkii, Alistipes shahii, Bacteroides caccae, Bacteroides ovatus, Bifidobacterium longum, Enterobacter cloacae, Klebsiella pneumoniae, Porphyromonas asaccharolytica |
Vitamin B6 synthesis (↓) | K06215 | Eubacterium rectale, Desulfovibrio piger |
Vitamin B12 synthesis (↓) | COBALSYN-PWY K02189 K03394 K05934 K05936 K06042 RIBAZOLEPHOSPHAT-RXN RXN-8770 2.7.1.156-RXN COBINAMIDEKIN-RXN COBINPGUANYLYLTRANS-RXN RXN-14063 | Eubacterium rectale, Faecalibacterium prausnitzii, Roseburia intestinalis, Desulfovibrio piger, Ruminococcus torques, Ruminococcus obeum, Citrobacter freundii, Citrobacter koseri, Collinsella aerofaciens, Coprococcus catus, Fusobacterium nucleatum, Fusobacterium periodonticum, Klebsiella sp MS 92 3, Lactobacillus reuteri, Megamonas funiformis, Megamonas hypermegale, Megamonas rupellensis, Methanosphaera stadtmanae, Morganella morganii, Roseburia hominis, Salmonella enterica, Streptococcus australis, Streptococcus parasanguinis, Streptococcus sanguinis, Veillonella dispar, Veillonella parvula, Veillonella sp oral taxon 158 |
SCFA production (↓) | GLUCUROCAT-PWY P42-PWY PWY-5177 PWY-6507 PWY-7242 K00016 K00074 K00248 K00626 K01571 K01625 K01715 K03856 K04070 K03785 K15634 METHYLACETOACETYLCOATHIOL-RXN RXN-12561 RXN0-2044 RXN-11245 RXN-16133 RXN-12705 RXN-14275 RXN-12750 RXN-12490 RXN-11662 RXN-12570 ACETYL-COA-ACETYLTRANSFER-RXN OHACYL-COA-DEHYDROG-RXN | Faecalibacterium prausnitzii, Ruminococcus torques, Eubacterium rectale, Roseburia intestinalis, Anaerostipes hadrus, Lachnospiraceae bacterium 5 1 63FAA, Roseburia inulinivorans, Roseburia hominis, Ruminococcus champanellensis, Bacteroides sp 3 1 19, Bifidobacterium adolescentis, Bifidobacterium animalis, Eggerthella lenta, Eubacterium eligens, Eubacterium ventriosum, Ruminococcus bromii, Ruminococcus obeum, Treponema succinifaciens |
Mucin degradation (↓) | K01207 | Eubacterium rectale, Bifidobacterium adolescentis, Eubacterium siraeum, Ruminococcus bromii, Adlercreutzia equolifaciens, Roseburia intestinalis, Bifidobacterium bifidum, Streptococcus parasanguinis, Megamonas hypermegale |
* on average more abundant in the group with increased abundance of a given function. |
The enzyme pyridoxal 5'-phosphate synthase (K06215) involved in the deoxyxylulose 5-phosphate (DXP)-independent pathway for vitamin B6 synthesis, one pathway, five KO categories and six metaCyc reactions related to vitamin B12 (cobalamin) synthesis and five pathways, 11 KO categories and 13 MetaCyc reactions involved in SCFA synthesis, including pyruvate and acetyl-CoA production and butyrate synthesis from acetate or lactate, were decreased in women with T1D (Excel file E5-E7; Table 2; Fig. 5; Figure S11). The abundance of beta-N-acetylhexosaminidase (K01207) involved in the degradation of mucin, was also significantly reduced in women with T1D, but again only in trimester 2 (Table 2; Excel file E6; Fig. 5; Figure S11).
Identification of bacterial clusters based on differentially abundant functional features
Differentially abundant functional features derived from HUMAnN2 were contributed not by a single species but rather a combination of species. Therefore, relative abundances of the principal contributing species in each of the six selected functions could be grouped into clusters (Table 2). For functions with three or more features, only principal contributors to at least three features were considered. For each cluster, a linear model was fitted with lmer using the same design as for the differential abundance analysis. This confirmed that women with T1D had an increased abundance of bacterial clusters contributing to production of LPS and synthesis of vitamin K2 and a decreased abundance of bacterial clusters contributing to synthesis of vitamins B6 and B12, production of SCFA and degradation of mucin (Figure S12).
Markers of gut pathology
Because the composition and function of the gut microbiome of women with T1D was suggestive of a pro-inflammatory state we sought evidence for gut inflammation in women with T1D. Fecal calprotectin, released from neutrophils and monocytes, is a marker of intestinal inflammation that may result in increased epithelial permeability [26]. Serum intestinal fatty acid-binding protein (I-FABP) is a marker of intestinal epithelial damage [27]. Fecal calprotectin and serum I-FABP were measured in trimester 3 in 61 women (32 with T1D) and 55 women (27 with T1D), respectively. Fecal calprotectin was increased in women with T1D compared to those without T1D (112 ± 148 vs. 36 ± 28 [mean ± SD] mg/kg: P-value 0.04; Mann-Whitney test). Serum I-FABP was also increased in women with T1D compared to women without T1D (587 ± 235 vs 314 ± 185 [mean ± SD] pg/ml: P-value 0.0003; Mann-Whitney test) (Figure S13). However, these markers did not significantly correlate (Spearman R > 0.4) with any of the individual taxa that were differentially abundant between T1D and non-T1D women.