Participant characteristics
A total of 194 AF sample were obtained at the second trimester. At the time of delivery, 55 AF samples from 51 pregnancies (including 47 single pregnancies and 4 twin pregnancies) of term, non-labored and elective cesarean section deliveries were collected, along with 55 AF counterparts from the second trimester, forming Cohort 1. Twenty-two women (including 20 single pregnancies and 2 twin pregnancies) experienced adverse pregnancy outcomes (1 woman with twin was stillborn at 22 gestational weeks,1 woman with twin was underwent late abortion at 27 gestational weeks; 2 women were spontaneous preterm birth, placental histopathology of 13 women were positive, and 5 women suffered both). However, the AF samples were lost at the time of delivery, and their 24 AF samples from the second trimester were included in cohort 2 (Fig. S1).
The demographic and clinical characteristics of the two cohorts are shown in Table 1. Gravidity, gestational age at amniocentesis, the ratio of twins and infant gender were comparable between the two cohorts, while maternal age, parity, gestational age at delivery, birth weight, and cesarean section rate were significantly different between cohort 1 and cohort 2, which were intrinsically related to pregnancy outcomes.
Table 1
Participant characteristics
Characteristic
|
Cohort 1(n = 51)
|
Cohort 2 (n = 22)
|
P value
|
Maternal age (mean ± SD, years)
|
37.6(2.7)
|
35.8(3.5)
|
0.02
|
Gravidity (mean ± SD)
|
2.4(1.1)
|
2.1(0.8)
|
> 0.05
|
Parity (mean ± SD)
|
1.6(0.5)
|
1.3(0.6)
|
0.02
|
GA at amniocentesis (mean ± SD, weeks)
|
21.7(0.6)
|
21.5(1.2)
|
> 0.05
|
GA at delivery (mean ± SD, weeks)
|
38.2(1.0)
|
35.2(6.1)
|
< 0.01
|
Twins, n (%)
|
4(7.2)
|
2(8.3)
|
> 0.99
|
Birth weight(g)
|
3231(448.1)
|
2778(960.3)
|
< 0.01
|
Infant gender, n (%)
|
|
|
0.33
|
Male
|
27(49.1)
|
14(63.6)
|
|
Female
|
28(50.9)
|
8(36.4)
|
|
Delivery, n (%)
|
|
|
< 0.01
|
Vaginal
|
0(0)
|
12(54.5)
|
|
Cesarean
|
51(100)
|
10(45.5)
|
|
Complications a
|
|
|
|
Stillbirth
|
-
|
1
|
|
Abortion
|
-
|
1
|
|
Preterm birth
|
-
|
2
|
|
Histological Chorioamnionitis
|
-
|
13
|
|
Preterm birth and
Histological Chorioamnionitis
|
-
|
5
|
|
a Complications indicate adverse pregnancy outcomes of the cohort 2. GA, gestational age. The "n" means the number of pregnant women.
No bacteria found by bacterial culture
All AF samples and sampling negative controls were performed aerobic and anaerobic culture, as well as genital mycoplasma cultivation. Among the 134 AF samples and 4 sampling negative control, only one second trimester AF sample (subject 45) yielded Paenibacillus lactis in aerobic condition. However, no bacteria found in the its delivery AF counterpart, and the subject 45 did not suffered any adverse pregnancy outcome. As Paenibacillus lactis is a common environmental contaminant in clinical culture, we incline to confirm positive results in the sequencing data.
Amniotic fluid has low bacterial biomass
To quantify the microbial biomass in AF samples, the qPCR assay was used to measure the copy numbers of the 16S rRNA gene. A standard curve over a range of 10–107 gene copies was generated by the linear regression analysis of an E coli plasmid (Fig. 1a, slope=-4.4936; R2 = 0.99). Compared with stool samples (med = 1.5×107 copies/mL, min = 1×107 copies/mL, max = 5.1×107 copies/mL) and vaginal swabs (med = 3.7×105 copies/mL, min = 1.2×105 copies/mL, max = 6.2×105 copies/mL), AF samples contained extremely low bacterial biomass (Fig. 1b, med = 67 copies/mL, min = 21 copies/mL, max = 1401 copies/mL for the second trimester AF in cohort 1, med = 207 copies/mL, min = 67 copies/mL, max = 4950 copies/mL for the delivery AF, med = 180 copies/mL, min = 64 copies/mL, max = 92400 copies/mL for the second trimester AF in cohort 2). Furthermore, the 16S rRNA gene copy number was assessed in negative control samples. There was no significant difference between the second trimester AF in cohort 1 and the sampling controls (med = 52 copies/mL, min = 30 copies/mL, max = 125 copies/mL; Mann-Whitney U test; U = 81; P = 0.4038), while the second trimester AF in cohort 1 contained higher numbers of 16S gene copies than the extraction controls (med = 25 copies/mL, min = 7 copies/mL, max = 80 copies/mL, U = 56; P < 0.01) and amplification controls (med = 17 copies/mL, min = 5 copies/mL, max = 50 copies/mL; U = 24; P < 0.0001), indicating that the potential contamination may occur during the sampling procedure. The delivery AF showed a significant increase in bacterial DNA copies in the comparison to the second trimester AF counterparts (U = 389; P < 0.001), which suggested the varying bacterial biomass as pregnancy progressed, but still with extremely low biomass. Interestingly, the numbers of 16S gene copies of the second trimester AF in cohort 1 were significantly lower than that of cohort 2 (U = 226; P < 0.0001). Since the copy number of 16S rRNA gene were both considerably low in negative control and AF samples, we attempted to confirm these results in the sequencing data.
The bacterial community structure of the second trimester AF in cohort 1 is indistinguishable from negative controls
The bacterial DNA of all the AF samples and control samples were assessed by 16S ribosomal RNA gene sequencing. As the technical control samples, the taxonomic composition and relative abundance of each artificial bacterial community was consistent with expectation (Figure S2), which guaranteed the reliability and stability of sequence data in this study.
At first, we compared the read counts of second trimester AF and delivery AF in cohort 1, as well as negative and positive control samples. The second trimester AF had a median read count of 10 (Fig. 2a, min = 0, max = 32, n = 55), which were much lower than that of stool (med = 68799, min = 53033, max = 97564, n = 4, P < 0.0001) and vaginal swab samples (med = 26914, min = 7658, max = 70343 n = 4, P < 0.0001). No difference was found between the second trimester AF and sampling controls (med = 7, min = 3, max = 10, n = 4, P = 0.29), extraction control (med = 2.5, min = 1, max = 37, n = 6, P = 0.28), or amplification control (med = 2, min = 0, max = 30, n = 6, P = 0.06). The read counts of the second trimester AF were slightly lower than those of the delivery AF counterparts (med = 10, min = 2, max = 55, n = 55, P = 0.04), and the latter was higher than that of amplification control (P < 0.01).
After annotation to bacterial operational taxonomic units (OTUs), we found similarly results, showing that the OTU numbers of the second trimester AF (Fig. 2b, med = 7, min = 0, max = 16) were much lower than those of stool (med = 1011, min = 952, max = 1107, n = 4, P < 0.0001) and vaginal swab samples (med = 56.5, min = 29, max = 68, n = 4, P < 0.0001). However, they did not differ significantly from any negative controls (P > 0.05 for all comparisons). The OTU numbers of the delivery AF samples (med = 8, min = 2, max = 23, n = 55) were higher than those of the second trimester AF (P < 0.01) and also significantly higher than those of sampling control (P = 0.01) and amplification control (P = 0.01). Furthermore, we analyzed within-sample diversity and found that the alpha diversity (Fig. 2c; Chao1, Shannon and Simpson indices) of the second trimester AF were comparable to that of sampling control (P = 0.25, P = 0.09 and P = 0.11, respectively), extraction control (P = 0.41, P = 0.40 and P = 0.41, respectively) and amplification control (P = 0.26, P = 0.19 and P = 0.21, respectively), and were significantly lower than that of stool samples (P < 0.0001 for all). Also, we found that the Chao1, Shannon and Simpson indices of the delivery AF samples were significantly higher than those of the second trimester AF counterparts (P = 0.02, P < 0.01 and P = 0.01, respectively).
To investigate the bacterial community composition, beta diversity of all samples was investigated by performing principal coordinate analysis (PCoA) based on unweighted Unifrac distance. As shown, the second trimester AF clustered with all negative controls (Fig. 2d, P = 0.12 for sampling control, P = 0.08 for extraction control and P = 0.06 for amplification control;), and were distinct from stool and vaginal swab samples (P = 0.01 and P < 0.01, respectively). The delivery AF samples were different from negative controls (P = 0.002 for sampling control, P = 0.027 for extraction control and P = 0.002 for amplification control), and also distinct from stool (P = 0.001) and vaginal swab (P = 0.025). Furthermore, we found the delivery AF samples appeared to cluster separately from the second trimester AF counterparts (P < 0.01), indicating varying bacterial signals from the second trimester to the time of delivery. In addition, we calculated the unweighted UniFrac distances to quantify the community similarity between sample types (Fig. 2e). Compared with the second trimester AF samples, the community distance between delivery AF samples and vaginal swab were significantly smaller (P = 0.001), suggesting an increased similarity to vaginal microbiota as the cervix shortens and the cervix mucus plug is lost near delivery.
Dynamic changes of predominant bacterial OTU signals across gestational timeline
Given the differences in microbial community structure between the second trimester AF samples and the delivery AF counterparts, we identified the specific OTU signals associated with bacterial genera based on the average abundance and distribution in each group, respectively. As shown, we highlighted the enrichment of 8 bacterial OTU genera specifically predominant in the second trimester AF, including Listeria, Enterococcus, Streptococcus, Bacteroides, Propionibacterium, Megasphaera, Prevotella, and Sphingomonas (Fig. 3a). Additionally, we identified 8 bacterial OTU genera specifically predominant in delivery AF, including Lactobacillus, Ureaplasm, Bacteroides, Streptococcus, Listeria, Enterococcus, Erwinia and Gardnerella (Fig. 3b).
The prevalence and abundance of the second trimester AF-specific OTU genera were significantly decreased in the corresponding delivery AF (Fig. 3a), and the second trimester AF-specific OTU genera showed a homogeneous distribution among samples, suggesting a systemic contamination. On the other hand, the delivery AF-specific OTU genera were rare in the corresponding second trimester AF, and some of them were common in vagina, prompting the hypothesis that microbiota might travel from vaginal to the amniotic cavity through ascending pathway as the cervix shortens and cervix mucus plug is lost near the delivery.
The specific bacteria OTUs of the second trimester AF in cohort 2 resemble vaginal microbiota
Microbial invasion of the amniotic cavity (MIAC) has been closely linked to obstetrical complication, including spontaneous preterm birth, late abortion, preterm premature rupture of membranes (PPROM), and histological and clinical chorioamnionitis[24]. In this study, we further retrospectively analyzed the second trimester AF from pregnancies who suffered preterm birth and/or intra-amniotic infection. Firstly, the read counts of the second trimester AF in cohort 2 were much more variable, with a median read count of 12 (Fig. 2a, min = 2, max = 63759, n = 24), which were higher than those of the second trimester AF in cohort 1 (P = 0.02) and much lower than those of stool (P < 0.001) and vaginal swab samples (P < 0.001). Similarly, the OTU numbers of the second trimester AF in cohort 2 (Fig. 2b, med = 9, min = 2, max = 117) were higher than those of the second trimester AF in cohort 1 (P < 0.01), sampling control (P = 0.03) and amplification control (P = 0.01), and also lower than those of stool (P < 0.0001) and vaginal swab samples (P < 0.01), indicating significant enrichment of microbial signals in amniotic cavity of symptomatic pregnancies at the second trimester.
Regarding within-sample diversity, the second trimester AF in cohort 2 had slightly higher alpha diversity than the second trimester AF in cohort 1 (Fig. 2c, P < 0.05 for Chao1 index, P = 0.09 for Shannon index, and P = 0.33 for Simpson index, respectively) and lower than that of stool (P < 0.0001, P < 0.001, P < 0.01, respectively) and vaginal swab (P < 0.01, P < 0.01 and P < 0.01, respectively), as expected.
Overall, the microbial community structure of the second trimester AF in cohort 2 was distinguishable from those of stool (Fig. 4a, P = 0.001) and negative controls (P = 0.01 for both sampling control and amplification control). When compared with the second trimester AF in cohort 1, the second trimester AF in cohort 2 showed a more variable and dissimilar distribution (P = 0.001). Notably, no significant difference was observed between the second trimester AF in cohort 2 and vaginal swab in the community structure (P = 0.06). These results were further consolidated by comparison of between-sample unweighted UniFrac distances. The UniFrac dissimilarity between the second trimester AF in cohort 2 and sampling control was significantly higher than that of the second trimester AF in cohort 1 and sampling control (Fig. 4b, P = 0.019), while the distance between the second trimester AF in cohort 2 and vaginal swab was significantly lower than that of the second trimester AF in cohort 1 and vaginal swab (P = 0.021), which stimulated the speculation that an ascending microbial colonization of the intrauterine cavity may have occurred at the second trimester.
To identify the bacterial OTUs unique to the second trimester AF samples in cohort 2, the sequence data were filtered by the following criterions: 1) absence in negative controls and the second trimester AF in cohort 1; 2) presence in at least two AF samples. Eventually, 9 bacterial OTUs at the genus level were identified, including Ureaplasma, Lactobacillus, Turicibacter, Bradyrhizobium, Streptococcus, Gardnerella, Ruminococcus, Anaeroplasma, Mucispirillum (Table 2). Ureaplasma was the most abundant and predominant genus, found in 13 AF samples from 12 women. Subject 94, who harbored the highest number of Ureaplasma reads (a total of 62,819), underwent spontaneous preterm birth at 34 gestational weeks with intact membrane and histologic chorioamnionitis, but cervixvaginal fluid and AF culture was negative. The pregnancy outcomes of women with these 9 bacterial OTU were described in Table 2.
Table 2
Clinical complications of pregnancies who harbor bacterial signals in cohort 2
Genus
|
Total reads
|
Prevalence
(n, %)
|
Clinical Complications
|
Ureaplasma
|
62,840
|
13, 54.2%
|
1. Subject 60 suffered histologic chorioamnionitis;
2. Subject 94 suffered both spontaneous preterm birth at 34 gestational weeks and histologic chorioamnionitis;
3. Subject 98 suffered spontaneous preterm birth at 35 gestational weeks;
4. Subject 105 suffered both spontaneous preterm birth at 34 gestational weeks and histologic chorioamnionitis;
5. Subject 106 suffered histologic chorioamnionitis;
6. Subject 128 suffered histologic chorioamnionitis;
7.Subject 132 were monochorionic-diamniotic twins, suffered late abortion at 27 gestational weeks, and Ureaplasma was found in both two amniotic fluid samples;
8. Subject 138 suffered histologic chorioamnionitis;
9. Subject 162 suffered spontaneous preterm birth at 36 gestational weeks
10. Subject 175 suffered histologic chorioamnionitis;
11. Subject 191 suffered histologic chorioamnionitis;
12. Subject 208 was monochorionic-diamniotic twins, and suffered fetal death at 20 gestational weeks
|
Lactobacillus
|
68
|
10, 41.6%
|
1. Subject 58 suffered histologic chorioamnionitis;
2. Subject 112 suffered both spontaneous preterm birth at 33 gestational weeks and histologic chorioamnionitis;
3. Subject 128 suffered histologic chorioamnionitis;
4. Subject 162 suffered spontaneous preterm birth at 36 gestational weeks
5. Subject 132 were monochorionic-diamniotic twins, suffered late abortion at 27 gestational weeks, and Lactobacillus was found in both two amniotic fluid;
6. Subject 171 suffered histologic chorioamnionitis;
7. Subject 175 suffered histologic chorioamnionitis;
8. Subject 179 suffered histologic chorioamnionitis;
9. Subject 209 was monochorionic-diamniotic twins, and suffered fetal death at 20 gestational weeks;
|
Turicibacter
|
23
|
2, 8.3%
|
1.Subject 112 suffered both spontaneous preterm birth at 33 gestational weeks and histologic chorioamnionitis;
2. Subject 132 were monochorionic-diamniotic twins, suffered late abortion at 27 gestational weeks, and Ureaplasma was found in both two amniotic fluid samples;
|
Bradyrhizobium
|
13
|
5, 20.8%
|
1. Subject 171 suffered histologic chorioamnionitis;
2. Subject 175 suffered histologic chorioamnionitis;
3. Subject 187 suffered histologic chorioamnionitis;
4. Subject 196 suffered iatrogenic preterm at 32 gestational weeks and histologic chorioamnionitis;
5. Subject 208 was monochorionic-diamniotic twins, suffered fetal death at 20 gestational weeks and Bradyrhizobium found in one of amniotic fluid;
|
Streptococcus
|
3
|
3, 12.5%
|
1.Subject 132 were monochorionic-diamniotic twins, suffered late abortion at 27 gestational weeks, and Streptococcus was found in one of amniotic fluid;
2. Subject 171 suffered histologic chorioamnionitis
3. Subject 196 suffered iatrogenic preterm at 32 gestational weeks and histologic chorioamnionitis;
|
Gardnerella
|
2
|
2, 8.3%
|
1. Subject 105 suffered both spontaneous preterm birth at 34 gestational weeks and histologic chorioamnionitis;
2. Subject 208 was monochorionic-diamniotic twins, suffered fetal death at 20 gestational weeks, and Gardnerella was found in one of amniotic fluid;
|
Ruminococcus
|
2
|
2, 8.3%
|
1.Subject 112 suffered both spontaneous preterm birth at 33 gestational weeks and histologic chorioamnionitis;
2. Subject 132 were monochorionic-diamniotic twins, suffered late abortion at 27 gestational weeks, and Ruminococcus was found in one of amniotic fluid;
|
Anaeroplasma
|
2
|
2, 8.3%
|
1.Subject 112 suffered both spontaneous preterm birth at 33 gestational weeks and histologic chorioamnionitis;
2. Subject 132 were monochorionic-diamniotic twins, suffered late abortion at 27 gestational weeks, and Ruminococcus was found in one of amniotic fluid;
|
Mucispirillum
|
2
|
2, 8.3%
|
1.Subject 112 suffered both spontaneous preterm birth at 33 gestational weeks and histologic chorioamnionitis;
2. Subject 132 were monochorionic-diamniotic twins, suffered late abortion, and Ruminococcus was found in one of amniotic fluid;
|
The second trimester AF in cohort 2 has relative high concentrations of inflammatory cytokines
The sequencing data revealed different profiles of bacterial signals among the second trimester and delivery AF in cohort 1, as well as the second trimester AF in cohort 2, however, no evidence supported the relationship between these bacterial signals and clinical complications. To determine potential inflammatory response in the amniotic cavity, the profile of 6 cytokines, including tumor necrosis factor-α (TNF-α), interleukin (IL)-6, IL-8, IL-10, IL-1β and IL-2, were further investigated. Generally, in all AF samples, the levels of all cytokines were considerably low, and the data did not meet the measurement threshold was removed. In cohort 1, no significant difference was found between the second trimester AF and delivery AF regarding the 6 cytokines. These results suggested that the bacterial signals might be non-biologically functional for the absence of the maternal inflammation responses. When compared with the second trimester AF in cohort 1, the levels of TNF-α, IL-10, and IL-1β were significantly higher in the second trimester AF in cohort 2, while the levels of IL-2, IL-6 and IL-8 were comparable between them (Table 3).
Table 3 Cytokines concentrations in AF samples
|
|
|
Second trimester AF in cohort 1
|
Delivery AF in cohort 1
|
P
|
Second trimester AF in cohort 1
|
Second trimester AF in cohort 2
|
P
|
Pro-inflammatory (pg/ml)
|
|
TGF-α
|
3.74
(1.42–5.36)
|
2.91
(2.1–9.24)
|
0.08
|
3.74
(1.42–5.36)
|
5.87
(3.77–194.8)
|
< 0.0001
|
IL-6
|
286.5
(33.34–1316)
|
301.8
(43.77–1643)
|
0.27
|
286.5
(33.34–1316)
|
241.9
(43.61–3605)
|
0.16
|
IL-1β
|
3.37
(2.14–8.07)
|
3.67
(2.45–4.31)
|
0.07
|
3.37
(2.14–8.07)
|
5.2
(3.91–18.21)
|
0.01
|
IL-2
|
19.63
(8.01–41.22)
|
21.2
(8.4-36.85)
|
0.96
|
19.63
(8.01–41.22)
|
24.18
(14.77–62.74)
|
0.08
|
Anti-inflammatory (pg/ml)
|
|
IL-10
|
2.29
(1.77–7.26)
|
2.73
(2.45–4.31)
|
0.56
|
2.29
(1.77–7.26)
|
28.17
(11.02–272.7)
|
< 0.0001
|
Chemokines (pg/ml)
|
|
IL-8
|
333.8
(115.2–3049)
|
499.6
(199.7-837.4)
|
0.05
|
333.8
(115.2–3049)
|
393
(120.2–1961)
|
0.2
|