The impact of maternal oral microbiota on the risk of small vulnerable newborns: a nested case-control study

Abstract

Background

Small vulnerable newborns (SVNs) account for most neonatal deaths worldwide. Though maternal periodontal disease has been shown associated with an increased risk of preterm birth (PTB) and low birth weight (LBW), little evidence shows the potential mechanism. Our study aimed to explore the association between maternal oral microbiota and SVNs before and during pregnancy.

Methods

A nested 1:4 case-control study was undertaken. Women delivering SVNs, including spontaneous PTB, LBW, and small-for-gestational-age (SGA) newborns, were selected as cases, while women delivering normal newborns were randomly selected as controls. 480 unstimulated saliva samples were collected from 240 women (48 cases and 192 controls) in preconception and late pregnancy. 16 S rRNA gene sequencing was used for analysis.

Results

Women with SVNs showed lower richness index (p = 0.032) in oral microbiota during preconception, lower shannon (p = 0.028) and simpson (p = 0.023) index in late pregnancy compared to the control group. Granulicatella and Streptococcus were significantly enriched in saliva both before and during pregnancy in women delivering SVNs. The two evaluated genera were positively correlated with enriched metabolic pathways like lactose and galactose degradation. These genera and their species were also enriched among women in the PTB and SGA sub-groups.

Conclusions

Women with SVNs exhibited significantly lower diversity in oral microbiota, with two enriched genera Granulicatella and Streptococcus in both before and during pregnancy.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12903-025-06779-5.

Introduction

Newborns who are preterm, small-for-gestational-age (SGA), or with low birthweight (LBW) are collectively defined as small vulnerable newborns (SVNs) and account for most neonatal deaths worldwide [1]. Neonates born as SVNs are also faced with an increased risk for metabolic disorders [2], developmental delay [3], and lifelong health adversities [4]. Identification of risk factors of SVN plays a crucial role in promoting child health and achieving Sustainable Development Goals.

Large amount of studies found that periodontitis and periodontal pathogens during pregnancy may cause adverse pregnancy outcomes such as delivering SVNs [5–8]. However, lessons from several clinical trials with large sample sizes suggested periodontal therapy during pregnancy may be too late to reduce the inflammation relevant to SVN [9, 10]. Alternatively, preconception might be an ideal time window to prevent adverse pregnancy outcomes [10]. The dysbiosis of the oral microbiota is a primary etiological factor of oral diseases, such as periodontal disease [11]. Evidence also indicates that oral microbiota dysbiosis is linked with systemic diseases [12], such as cardiovascular disease [13], diabetes mellitus [14], and adverse pregnancy outcomes [15]. However, the association between oral microbiota and the risk of SVNs during preconception and pregnancy has been rarely evaluated [16, 17].

The objectives of this study were to (1) examine the associations between maternal oral microbiota and SVN, in preconception and pregnancy respectively; and (2) identify the differential oral bacterial taxa and explore potential metabolic pathways related to SVN. The findings will be able to provide evidence for developing prediction and prevention strategy of SVNs.

Methods

Study design and participant involvement

The study was a nested case-control (1:4) design based on a preconception cohort in Shanghai [18, 19]. Women who attended the preconception examination were recruited from the preconception care clinic of the Maternal and Child Health Hospital of Jiading District in Shanghai. Women were eligible for the cohort study if they: (1) had intention to conceive; (2) were aged 20–49 years; and (3) were willing to be followed up through pregnancy until childbirth. In this nested case-control study, the case group included 48 women who delivered SVNs, while the control group included 192 women randomly sampled among those who delivered normal newborns excluding macrosomia, or large-for-gestational-age newborns. Women were also excluded in this nested case-control study if they were: (1) multiple pregnancies; (2) taking antibiotics or antifungal drugs within 30 days before biological sample collection; (3) medical-induced preterm delivery. Saliva samples from preconception to late pregnancy were acquired among all participants.

A self-administered questionnaire survey at recruitment was carried out to collect women’s demographic information, disease history, and dietary intake [20]. Factor analysis was used to summarize preconception dietary patterns based on the dietary intake data measured by the food frequency questionnaire. The data of the experience of oral health care, including scaling, root planning and scaling, tooth extraction, and other possible oral care methods, after recruitment and during pregnancy were collected in the third trimester during pregnancy. Information on spontaneous birth, gestational age at birth and birth weight was subsequently extracted from women’s health records.

The research was approved by the Ethics Committee of the School of Public Health, Fudan University, Shanghai, China (IRB#2016-10-0601, IRB#2019-07-0770, IRB#2020-01-0794). All participants were informed about the study procedure and had provided written informed consent.

Case definition

SVNs in the case group were defined with at least one of the following conditions: spontaneous PTB (gestational age less than 37 weeks, not induced by medical interventions [21]), or LBW (birthweight less than 2,500 g, regardless of gestational age [21]), or SGA (birthweight less than the 10th percentile of the gender-specific gestational age [21], according to Chinese neonatal birthweight curve [22]). Newborns in the control group were term with 37 to 40 gestational weeks and with normal birthweight which was 2,500 g or above, and less than 4,000 g, falling > 10th and < 90th percentile of the gestational age and specific gender.

Sample collection

In this study, whole-mouth unstimulated saliva samples of women were collected in both the preconception and the third trimester of pregnancy. Saliva samples were collected according to a standardized method: (1) no food, water, cigarettes or gum 30 min before sampling; (2) to rinse their mouths with water to remove food debris 30 min before sampling but do not brush teeth; (3) to keep saliva in mouths for at least 1 min; (4) to spit 3-5 ml saliva (not sputum) directly into sterilized centrifuge tubes without stabilizing buffer. All saliva samples aliquoted after standing at room temperature for 30 min after collection, and were kept frozen at −80°C no later than 4 h after collection.

DNA extraction, amplification, and sequencing

The QIAampDNA Mini Kit (Qiagen, MD) was used for DNA extraction from saliva samples following the manufacturer’s protocol. For detecting bacterial 16SrRNA gene sequence, polymerase chain reaction amplification of the V3-V4 region was performed using the barcode-specific primers 338 F and 806R. All amplicons were purified by a QIAquick PCR Purification Kit (Qiagen) and pooled with equal concentrations. Then the pooled amplicons were sequenced on an Illumina MiSeq instrument (Illumina, USA) with a 2 × 300 cycle run.

Bioinformatics analysis and statistical analysis methods

Raw sequencing data were processed by VSEARCH (2.13.6). We spliced the PE reads, removed the primers, performed quality control, and used Unoise3 to cluster sequence into amplicon sequence variants (ASVs). Taxonomy was assigned to ASVs by referring to Silva (V132) and eHOMD (V3.1) database. QIIME2 (2019.4) and R (3.6.3) were used to compute and analyze alpha diversity, including richness (observed species), Shannon, and Gini-Simpson index and beta diversity containing bray-curtis, unweighted and weighted UniFrac distance.

We compared the differences in diversity, bacterial taxonomic profile, and predicted functional profile between the case and the control groups in preconception and late pregnancy, respectively. After rarefaction to an even sequencing depth per sample, we used t-test for alpha diversity indicies conforming to a normal distribution with a chi-square variance, and used the Mann-Whitney U-test for other indicies. Boxplot was also used to compare the differences of alpha diversity between the groups. Multivariate dispersion homogeneity was assessed using permutational analysis of multivariate dispersions (PERMDISP) with 9,999 permutations, and then the principal coordinate analysis (PCoA) and permutational multivariate analysis of variance (PerMANOVA) were used to analyze the differences in microbiota composition between the two groups based on beta diversity. We used logistic regression within the generalized linear model and PerMANOVA to perform multivariable analyses controlled with preconception BMI, preconception dietary pattern, oral health care during pregnancy, and infant’s gender as covariates.

Linear discriminant analysis (LDA) effect size (LEfSe) [23] was performed to identify differential bacterial taxa between groups, with the LDA threshold of 2.0. We also used Mann-Whitney U-test to identify differential bacterial species and adjusted p values by the Benjamini-Hochberg false discovery rate (FDR). Species with q value (adjusted p value) < 0.05 was considered significantly differential. The changes in bacterial taxonomic profile would directly alter its metabolic function. Therefore, we used Phylogenetic Investigation of Communities by Reconstruction of Unobserved States 2 (PICRUSt2) [24] to predict metagenome function from 16 S rRNA data based on the MetaCyc database [25]. The DESeq2 package [26], FDR adjustment, and bubble plots were then used to explore differential metabolic pathways between the groups. Correlation analysis and heatmap were performed among differential taxa, and differential predicted metabolic pathways, using Pearson correlation coefficient.

Post hoc analyses

Post hoc analyses were conducted to test whether those SVN-related genera and species were associated with specific type of SVNs, including PTB and SGA. Outcomes assessed only included relative abundances of Granulicatella, Streptococcus, Streptococcus lactarius, Granulicatella adiacens, and Streptococcus oralis. For PTB sub-group analyses, we compared the difference in these taxa between women with and without spontaneous preterm newborns (n = 15 vs. n = 60), whereas the controls were randomly selected from those women with term newborns (n = 240 − 15 = 225). For SGA analyses, we compared the difference in these taxa between women with and without SGA newborns (n = 34 vs. n = 136), whereas the controls were randomly extracted from those women with AGA newborns (n = 240 − 34 = 206). The Mann-Whitney U-test and boxplot were used for post hoc analyses.

Results

Participant characteristics

In this study, we analyzed a total of 480 saliva samples of 240 women, including 48 SVN cases and 192 controls, collected from October 2016 to December 2022. The preconception and gestational characteristics of the two groups are listed in Table 1. Among the 48 cases, 15 (31.25%) women had preterm birth, 13 (27.08%) gave birth to infants with LBW, and 34 (70.83%) had newborns classified as SGAs. The average birthweight of infants delivered by women in the case and the control group was 2,620.21 g and 3,325.47 g, respectively, with the average gestational age of 37.98 and 39.02 weeks. According to factor analysis, four preconception dietary patterns were identified, including “non-staple food pattern” (characterized by intaking dark vegetable, mushroom, meat, animal offal, fish, and shrimp), “balanced low-sugar pattern” (characterized by intaking rice, vegetable, and fruit, and low intake of sugary drinks), “processed food pattern” (characterized by intaking candy or chocolate, instant noodle, processed meat product, and pastry), and “coarse food pattern” (characterized by intaking wheat, legume grain, and tuber), with an accumulated variance of 47.0%. The case and the control group did not differ significantly in terms of age, BMI, education level, family annual income, dietary pattern, parity, and thyroid-related diseases before pregnancy, oral health care during pregnancy, other pregnancy complications including gestational diabetes mellitus (GDM) and pregnancy-induced hypertension (PIH), and infants’ gender (p > 0.05).

Table 1.

Characteristics of participants between the case and the control groups (n = 240)

Characteristics	Cases(n = 48)	Controls (n = 192)	p value	Characteristics	Cases(n = 48)	Controls (n = 192)	p value
Preconception age (Year)			0.388a	Preconception thyroid related disease			0.161a
< 30	37(77.08%)	136(70.83%)		Non-staple food pattern	16(33.34%)	55(28.65%)
≥ 30	11(22.92%)	56(29.17%)		Balanced low-sugar pattern	18(37.50%)	55(28.65%)
Preconception BMI (kg/m 2)			0.222a	Processed food pattern	10(20.83%)	39(20.31%)
< 18.5	8(16.67%)	19(9.90%)		Coarse food pattern	4(8.33%)	43(22.40%)
18.5–24	36(75.00%)	143(74.48%)
> 24	4(8.33%)	30(15.62%)		Preconception thyroid related disease
Education level			0.307a				0.905a
Below college	10(20.83%)	54(28.13%)		No	44(91.67%)	177(92.19%)
College and above	38(79.17%)	138(71.88%)		Yes	4(8.33%)	15(7.81%)
Family annual income (yuan)			0.112a	PIH			0.092a
< 100,000	20(41.67%)	57(29.69%)		No	45(93.75%)	162(84.38%)
≥ 100,000	28(58.33%)	135(70.31%)		Yes	3(6.25%)	30(15.63%)
Parity			0.329a	GDM			0.623a
0	44(91.67%)	166(86.46%)		No	40(83.33%)	154(80.21%)
≥ 1	4(8.33%)	26(13.54%)		Yes	8(16.67%)	38(19.79%)
Gestational oral health care			0.487a	Infants’ gender			0.051a
No	9(18.75%)	45(23.44%)		female	33(68.75%)	102(53.13%)
Yes	39(81.25%)	147(76.56%)		male	15(31.25%)	90(46.88%)
Gestational age (week)	37.98 ± 2.21	39.02 ± 1.10	< 0.001b	Birthweight (gram)	2,620.21 ± 283.01	3,325.47 ± 282.04	< 0.001b

PIH Pregnancy-induced hypertension syndrome, GDM Gestational diabetes mellitus

^aχ² test

^bt-test

Altered microbial diversity both before and during pregnancy in SVNs group

We compared the differences of alpha diversity and beta diversity between the SVN and the control group (Tables 2 and 3; Fig. 1A-B). During preconception, the oral microbiota of women delivering SVNs showed significantly lower alpha diversity indices, compared with that of women with non-SVNs (observed species:363.00(330.50, 413.00) vs. 399.00(340.50,456.00), p = 0.037). The beta diversity analyses also detected significant differences in preconception oral microbiota composition among women with and without SVNs (bray-curtis: p = 0.012; weighted UniFrac: p = 0.002). Similar group differences were also detected in late pregnancy analysis [Gini-Simpson: 0.91(0.84,0.94) vs. 0.93(0.89,0.96), p = 0.034; bray-curtis: p = 0.001; weighted UniFrac: p = 0.001].

Table 2.

Comparison of alpha diversity of oral microbiota of participants between case and control group (n = 240)

Index	Case group	Control group	P valuea	βvalued	Standard errord	P valued
Preconception
Richness	363.00(330.50,413.00)	399.00(340.50,456.00)	0.037 c	−0.004	0.002	0.046
Shannon	5.41 ± 0.63	5.61 ± 0.68	0.051 b	−0.359	0.235	0.126
Simpson	0.94(0.91,0.95)	0.94(0.92,0.96)	0.063 c	−2.153	2.600	0.408
Late Pregnancy
Richness	359.50(304.25,413.00)	387.50(332.00,456.75)	0.079 c	−0.002	0.002	0.231
Shannon	5.08 ± 1.01	5.39 ± 0.83	0.055 b	−0.379	0.185	0.040
Simpson	0.91(0.84,0.94)	0.93(0.89,0.96)	0.016 c	−3.951	1.843	0.032

^aP value describing the differences between groups by univariate analysis

^bt-test

^cU test

^dβ value, standard error, and P value describing the differences between the groups by multivariate analysis, with preconception BMI, dietary pattern, gestational oral health care, and infants’ gender as covariates

Table 3.

Comparison of beta diversity of oral microbiota of participants between case and control group (n = 240)

Distance Matrix	Model 1a			Model 2b
	R 2	F	P value	R 2	F	P value
Preconception
Bray-curtis	0.008	1.965	0.012	0.008	1.966	0.012
Unweighted UniFrac	0.005	1.085	0.268	0.005	1.092	0.263
Weighted UniFrac	0.016	3.974	0.002	0.016	3.969	0.002
Late Pregnancy
Bray-curtis	0.016	3.768	0.001	0.016	3.772	0.001
Unweighted UniFrac	0.006	1.514	0.057	0.006	1.523	0.057
Weighted UniFrac	0.028	6.881	0.001	0.028	6.869	0.001

^athe results of PerMANOVA analysis without any covariates

^bthe results of PerMANOVA analysis, with preconception BMI, dietary pattern, gestational oral health care, and infants’ gender as covariates

Fig. 1.

Differential oral microbiota diversity and composition between the case and the control group (n = 240). A Distribution boxplots of alpha diversity indices, the boxes represent median and interquartile range; red boxes represent the control group, and pink boxes represent the SVN group. T-test was used for comparisons of shannon index, and U-test was used for comparisons of richness and simpson indices. B PCoA of beta diversity based on Bray-Curtis distance, unweighted unifrac distance, and weighted unifrac distance, respectively. Blue and yellow dots represent the control and the SVN group, respectively

Increased granulicatella, streptococcus and their species in SVNs group

The LEfSe analyses showed that at the genus level (Fig. 2A-B), two genera including Granulicatella and Streptococcus were significantly enriched in preconception oral microbiota of women delivering SVNs, while four genera including Lautropia, Dialister, Aggregatibacter, and Haemophilus were significantly depleted. Similarly, the gestational oral microbiota of the SVN group demonstrated enriched genera Granulicatella, Streptococcus, and Gemella, and decreased Oribacterium, Fusobacterium, Haemophilus, Peptostreptococcaceae_[G-5], Absconditabacteria_[SR-1]_[G-1], Neisseria and Ruminococcaceae_[G-2].

Fig. 2.

Differential bacterial taxa identified by LEfSe analysis and Mann-Whitney U-test. Green and red color indicates enrichment in the case and the control group, respectively. A Cladogram of differential bacterial taxa (n = 240). Taxa at the level of phylum, class, order, family, and genus are shown from inside to outside. Circles’ diameter is proportional to the taxon’s abundance. B LDA score distribution bar plot of differential bacterial taxa (n = 240). The length of the histogram represents the LDA score. C Post hoc analyses of differential bacterial taxa among women with spontaneous preterm newborns by Mann-Whitney U test (n = 75). D Post hoc analyses of differential bacterial taxa among women with SGA newborns by Mann-Whitney U test (n = 170)

At the species level (Table 4), we found five species including Streptococcus cristatus, Streptococcus lactarius, Granulicatella adiacens, Streptococcus oralis, and Prevotella nanceiensis were significantly enriched in the oral microbiota of the SVN group during pregnancy. However, no different bacterial species were detected in the preconception analyses.

Table 4.

Differential bacterial species identified by univariable non-parametric test (n = 240)

Bacterial species	Enriched in	p value	q value
Preconception (12 species with p < 0.05)
Granulicatella adiacens	SVN	0.002	0.154
Streptococcus oralis	SVN	0.004	0.154
Aggregatibacter sp._HMT_458	Control	0.010	0.203
Streptococcus australis	SVN	0.010	0.203
Abiotrophia defectiva	SVN	0.016	0.205
Porphyromonas endodontalis	Control	0.016	0.205
Dialister invisus	Control	0.017	0.205
Prevotella scopos	Control	0.020	0.210
Haemophilus parainfluenzae	Control	0.027	0.247
Campylobacter concisus	Control	0.031	0.260
Streptococcus cristatus	SVN	0.036	0.277
Neisseria subflava	Control	0.049	0.344
Late pregnancy (15 species with p < 0.05)
Streptococcus lactarius	SVN	< 0.001	0.002*
Streptococcus oralis	SVN	< 0.001	0.003*
Granulicatella adiacens	SVN	< 0.001	0.006*
Prevotella nanceiensis	Control	0.001	0.027*
Streptococcus cristatus	SVN	0.002	0.032*
Fusobacterium periodonticum	Control	0.005	0.069
Peptostreptococcaceae_[G-5] bacterium_HMT_493	Control	0.007	0.084
Ruminococcaceae_[G-2] bacterium_HMT_85	Control	0.012	0.128
Veillonella rogosae	Control	0.018	0.165
Streptococcus australis	SVN	0.023	0.180
Dialister invisus	Control	0.024	0.180
Streptococcus salivarius	SVN	0.030	0.207
Oribacterium parvum	Control	0.036	0.233
Streptococcus sanguinis	SVN	0.046	0.236
Campylobacter rectus	Control	0.049	0.236

Mann-Whitney U-test and Benjamini-Hochberg FDR were used

*q < 0.05 was considered significant

Post hoc analyses showed that the Granulicatella genus, Streptococcus genus, and Streptococcus oralis were all significantly enriched in the preconception oral microbiota of women with spontaneous preterm newborns (Fig. 2C), compared to women with term newborns. In addition, women with SGA newborns demonstrated more Streptococcus both before and during pregnancy, and more Granulicatella only during pregnancy (Fig. 2D), compared to women with AGA newborns.

Overexpression of microbial metabolic function in SVNs group

To infer metabolic pathways associated with SVNs, we used PiCRUSt2 and DESeq2 to map microbial genes by metabolic databases and to identify microbial functions differentially expressed (Fig. 3A). A total of 38 and 137 differential predicted metabolic pathways were detected before and during pregnancy, respectively (Supplementary file 2). Among those top 10 distinct pathways, four overlapped pathways were found enriched including lactose and galactose degradation I, superpathway of purine deoxyribonucleosides degradation, superpathway of geranylgeranyldiphosphate (GGPP) biosynthesis I (via mevalonate), and mevalonate pathway I both in preconception and late pregnancy. The subsequent correlation analyses (Fig. 3B, Supplementary file 3) exhibited significant positive correlations between the above three predicted metabolic pathways (except mevalonate pathway I) and those previously detected differential bacterial species such as Granulicatella adiacens, Streptococcus lactarius, and Streptococcus oralis both before and during pregnancy.

Fig. 3.

Differential metabolic pathways identified by DESeq2 package and correlation with differential taxa (n = 240). A Top 10 differential metabolic pathways identified by DESeq2 package (q < 0.05). B Heatmaps of Pearson correlation between differential bacterial taxa and metabolic pathways. *p < 0.05, **p < 0.01, ***p < 0.001. The color shows the Pearson correlation coefficient, with red representing negative correlation (r < 0) and blue representing positive (r > 0)

Discussion

To our knowledge, this is the first study to explore the association between maternal oral microbiota in preconception and late pregnancy and the risk of SVN. Overall, our study demonstrated significant differences in oral microbiota diversity and community structures between mothers with and without SVNs both before and during pregnancy. Three overlapped differential bacterial genera and four overlapped predicted metabolic pathways were found in preconception and late pregnancy. Post hoc analyses revealed that genera including Granulicatella and Streptococcus and their species were also associated with PTB and SGA.

Compared to the control group, women with SVNs exhibited a lower oral microbiota diversity both in preconception and late pregnancy. Lower microbiota diversity was usually caused by increased interactions between oral microorganisms, such as competition and antagonism [27]. Usually in a balanced ecosystem, functionally-related microbes compensate for each other and make up the functions of missing species [28]. People with lower gut microbiota diversity were more likely to develop obesity [29] and inflammatory bowel disease [30] etc. The decreased diversity implies the possibility of microbiota function dysbiosis and a species-rich ecosystem plays a robust role against environmental influences. Therefore, lower diversity could be considered as an underlying indicator of dysbiotic ecosystem and thus causing diseases, such as SVNs.

Considering the violation of assumptions of parametric statistical methods in microbiota studies, we used nonparametric methods including LEfSe analyses and U-test for differential abundance analyses. At the genus level, increased abundance of Granulicatella and Streptococcus in maternal oral microbiota both before and during pregnancy correlated with a higher risk of delivering SVNs. At the species level, women with SVNs presented relatively higher abundances of Granulicatella adiacens, Streptococcus lactarius, and Streptococcus oralis, consistent with existing evidence [15]. The Granulicatella genus had been linked to high inflammatory status in human, particularly in preterm infants with extrauterine growth restriction [31]. Several studies suggested that Granulicatella adiacens was positively linked with severe oral diseases such as refractory periodontitis, possibly worsening and inflammation in peripheral circulation through oral-systemic link [32]. As for Streptococcus, several species were identified as participants in metabolic pathways associated with oral inflammation and periodontitis, and were also found as predicting pathogens linked to low birth weight [33, 34]. Specifically, Streptococcus oralis was positively associated with poorer oral health and dental plaque formation [35]. However, the potential association between Streptococcus lactarius and SVN has been rarely studied and the underlying mechanism needs to be further investigated.

Our study disclosed that the enrichment of the Haemophilus genera both during preconception and pregnancy might promote normal fetal growth. Little evidence previously existed to explain the influence of the genera on SVN. The Haemophilus genus was thought to be associated with healthier periodontal status [36], consistent with the findings of our study. Additionally, the Neisseria genus was also enriched in pregnant women without SVN deliveries. Consistent with this finding, studies had aligned Neisseria species with healthy periodontal conditions and identified it as part of the healthy “core microbiota” of the human oral cavity [37, 38]. Furthermore, a prospective clinical trial reported a reduced Neisseria in oral microbiota during pregnancy in women with preterm-LBW [5].

The predicted function analyses indicated that the association between maternal oral microbiota and SVN deliveries might be mostly attributed to upregulation of three metabolic pathways, including lactose and galactose degradation I, superpathway of purine deoxyribonucleosides degradation, and superpathway of geranylgeranyldiphosphate biosynthesis I (via mevalonate). Typically, lactose metabolism occurs through the Leloir pathway among most species, playing critical role for fetal growth, especially in late pregnancy and during myelinogenesis development [39]. However, bacterial species with the upregulation of the pathway of lactose and galactose degradation I, such as Streptococcus oralis [35], would degrade lactose through the tagatose-6-phosphate pathway [25], depleting lactose as a carbon source without accumulating galactose extracellularly [40].

The superpathway of purine deoxyribonucleosides degradation contains metabolism pathways of various nucleosides and nucleotides to serve as sources of nutrients and energy donors for bacterial biological processes [41]. Another differential pathway found in this study, the superpathway of geranylgeranyldiphosphate biosynthesis I (via mevalonate), exists among human and most bacteria and is related to the biosynthesis of GGPP [42]. Bacteria with the upregulation of the GGPP biosynthesis pathway would compete against their human host for the raw materials and thus leading to human’s GGPP depletion, which might initiate chronic low-grade inflammation [43].

Moreover, the correlation analyses in our study in both preconception and late pregnancy showed that these three metabolic pathways were positively correlated with those bacterial species enriched in the SVN group such as Granulicatella adiacens, while negatively with bacterial genera enriched in the control group such as Neisseria. The findings suggested that these specific maternal oral bacterial taxa in both preconception and pregnancy could be the potential predictive biomarkers and intervention targets of SVN deliveries.

The findings of this study could provide evidence for developing early screening and intervention strategies of SVNs, highlighting the preconception period as a critical window for gamete function and fertilization. Chronic preconception infection from (conditional) pathogens in preconception might arouse systematic inflammation, leading to adverse pregnancy outcomes. Our results indicated consistent microbiota diversity, composition, bacterial features in genus, and differential metabolic pathways from preconception to late pregnancy, suggesting preconception oral microbiota could serve as potential predictors and intervention targets of adverse pregnancy outcomes.

Interestingly, although the abundance of species including Granulicatella adiacens and Streptococcus oralis in preconception was higher in the SVNs than non-SVNs group without statistical significance, they became significantly enriched in the SVNs group in pregnancy, indicating worsened dysbiosis due to hormonal alterations. Therefore, preconception care has the potential for enhancing pregnancy outcomes. Previous longitudinal studies found that preexisted periodontitis enhanced inflammatory response to oral microbiota in pregnant women [44–46]. Future studies are suggested to involve larger sample size and track microbiota profile from preconception to childbirth.

Our study used a 1:4 nested case-control design based on the prospective cohort to assess the association between oral microbiota and birth outcome, which might contribute to causal inference. We focused on the preconception stage which could be a potential ideal intervention window for improving pregnancy outcomes. The oral microbiologic features of the SVN group identified in our study can serve as potential non-invasive predictive biomarkers for discriminating risky population and inform the intervention development for preventing SVNs.

However, there are some Limitations in this study. First, we did not collect data regarding probiotics use, which could be the possible confounding factor. Second, prediction of metabolic pathways based on 16 S rRNA sequencing only indicated possible potential functions of the oral microbiota which needs to be validated through metagenomics, transcriptomics, etc. Additionally, the small sample size of the study, particulary for PTB, could limit statistical power for detecting associations and increase susceptibility to Type II errors in the post hoc analyses.

Conclusions

Women with SVNs exhibited significantly lower diversity in oral microbiota both before and during pregnancy, compared to women with non-SVNs. Two genera including Granulicatella and Streptococcus were enriched in women with SVNs in both preconception and pregnancy in women who delivered SVNs.

Abbreviations

SVN

Small vulnerable newborn

SGA

Small-for-gestational-age

PTB

Preterm birth

LBW

Low birth weight

ASV

Amplicon sequence variant

PCoA

Principal coordinate analysis

PerMANOVA

Permutational multivariate analysis of variance

LDA

Linear discriminant analysis

LEfSe

Linear discriminant analysis effect size

FDR

False discovery rate

PICRUSt2

Phylogenetic investigation of communities by reconstruction of unobserved states 2

Authors’ contributions

HJ conceived and obtained funding of this study. HJ, QX, TZ, AY, XL, and HW coordinated and supervised study implementation and data collection. YS and XC performed data collection and quality control. YW and HZ conducted laboratory testing on the samples. QX and XL conducted the data analysis. QX, XL and HJ drafted the manuscript. XX and HZ provided critical comments and revisions. All authors have approved to submit the manuscript for publication.

Funding

This study was supported by the National Natural Science Foundation of China (81973057, 82373579), the National Key Research and Development Program (2022YFC2704605), Shanghai “Science and Technology Innovation Action Plan” International Intergovernmental Science and Technology Cooperation Project (22410712700), the Sixth Round of the Three-Year Public Health Action Plan of Shanghai (GWVI-11.1-32), and the Key Discipline and Project of High⁃Quality Development of Public Health of School of Public Health, Fudan University-Jiading District Health Commission (GWGZLXK⁃2023⁃04).

Data availability

The datasets presented in this study can be found in online repositories. The names of the repository/repositories and accession number(s) can be found below: National Omics Data Encyclopedia (NODE) - OEP00006494.

Declarations

Ethics approval and consent to participate

This study was approved by the Ethics Committee of the School of Public Health, Fudan University, Shanghai, China (IRB#2016-10-0601, IRB#2019-07-0770, IRB#2020-01-0794) in accordance with the Declaration of Helsinki. All participants were informed about the study procedure and had provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.