Nine healthy gestational donors were recruited for the study. They were screened and selected based on specific criteria, which included being between 30 and 40 weeks of gestation, not having used antibiotics or any other medicine within the last 3 months, and not experiencing any gastrointestinal symptoms, such as constipation, diarrhea, or discomfort. Stools were collected from these donors for 15 consecutive days and preserved in glycerol prior to freezing. The preparation of the fecal bacterial solution was performed under anaerobic conditions at different phases of fecal microbiome transplant (FMT) and was described in detail previously (12).

Germ-free (GF) C57BL/6 J mice (8-week-old, female, n = 10, body weight 18–22 g) were obtained from the GemPharmatech Co., Ltd., located in Jiangsu, China. The GF mice were housed in six separate isolators with a 12-h light/dark cycle in a temperature-controlled (22°C ± 1°C) and humidity-controlled (55% ± 10%) room. Humanized GF mouse models were generated by performing fecal microbiome transplant (FMT) twice a week for three consecutive weeks, using a 0.2 mL of fecal sample for each time. In order to prevent environmental microorganisms and operational contamination, all experimental operations were performed under strict aseptic conditions. Subsequently, the models were then evaluated using 16S rRNA sequencing, as described previously in reference (12). Protocols were approved by the Animal Care and Use Committee of GemPharmatech Co., Ltd. (No. GPTAP20220321-1). Experiments were performed following the Guide of Experimental Animals of the Ministry of Science and Technology (Beijing, China).

Lactiplantibacillus plantarum HM-P2 (Patent No. CN115491329A) was isolated from healthy breast milk, and the strain was stored in the China General Microbiological Culture Collection Center.¹ Its probiotic properties have been disclosed in our granted patents “CN115491329A” and summarized in Supplementary material S1.

Lactiplantibacillus plantarum HM-P2 was selected based on its evaluated probiotic functions. This strain exhibits high bile salt tolerance, enabling it to withstand the high bile salt environment of the human body (such as in the small intestine). It also demonstrates strong inhibitory effects against pathogenic Escherichia coli and/or Staphylococcus aureus. Therefore, HM-P2 holds promising potential for development and application in the food industry. Furthermore, this strain has received granted both a domestic invention patent in China (CN115491329A) as well as an international PCT (Patent Cooperation Treaty) patent (WO2024060768). Therefore, this study has selected HM-P2 for a detailed investigation into its mechanisms for promoting maternal health during pregnancy.

This study aims to determine the direct effects of HM-P2 on maternal health through intervention in humanized pregnant mice. FMT GF mice (n = 10) were randomly assigned to two experimental groups, named the control and L. plantarum HM-P2 groups. Both groups received standard chow feeding (100 μ L phosphate buffered saline [PBS] solution) and standard chow + L. plantarum HM-P2, separately. Probiotic was pre-tested with three doses (low-1*10⁸, middle-1*10⁹, and high-1*10¹⁰ CFU/mL), and based on in vitro safety evaluations and literature references (see Supplementary material S2), the middle dose (1 × 10⁹) was ultimately selected for the FMT experiment. Probiotic was fed at a final dose of 1*10⁹ CFU/mL in PBS solution by oral gavage (2 days/week, p.o., 3 weeks). The GF mice were housed in six separate isolators with a 12-h light/dark cycle in a temperature-controlled (22°C ± 1°C) and humidity-controlled (55% ± 10%) room. The mice were allowed ad libitum access to germ-free water throughout the experimental period. Health status was monitored daily throughout the experiment; feces were collected at 0 d, 1 d, 7 d, 14 d, and 18 d of the experiment, and then stored at −80°C for short-chain fatty acids (SCFAs), bile acids, tryptophan analysis, and metagenome analysis. After delivery, the number of offspring was recorded. At the end of the experiment, mice were fasted overnight and sacrificed, and the fresh weight of tissues (total and left liver, spleen, and kidney) was measured. Blood was collected and centrifuged at 3,000 rpm for 10 min and stored in a − 80°C environment until further analysis.

Fecal samples were collected and shotgun metagenomic sequenced to verify the colonization of the L. plantarum strain at 0 d, 1 d, 7 d, 14 d, and 18 d.

Small portions of the middle cecum and colon were washed, placed in a 10% phosphate-buffered formalin solution, and stored at room temperature for histological analysis. The evaluated morphological indicators were villi height, top crypt depth, and middle crypt depth. Morphologic analysis was performed on 10 well-oriented and intact villi and 10 crypt foci from each section of the cecum and colon.

Multiplexed bead-based immunoassays (Luminex 200 system, Thermo Fisher Scientific, United States) were performed to detect serum levels of cytokines IFN- γ , TGF- β -1, TNF- α , IL-10, IL-8, IL-6, IL-4, TGF- α , IL-22, IL-17, IL1- β , and IL1- α , as well as immune factors such as IgM and IgG, using mice ELISA kits (Cloud-Clone Corp., Katy, TX, United States) according to the manufacturer’s instructions.

Bacterial genomic DNA was extracted from 100 mg of homogenized fecal samples using the QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) following the manufacturer’s protocol. DNA was examined by 0.8% agarose gel electrophoresis, and the OD value of 260/280 was determined by spectrophotometry. A HiSeq 2,500 (Illumina, CA, United States) with 150 bp paired-end reads was used to perform shotgun metagenomic sequencing of all the DNA samples. FastQC v0.11.9² was used to check the quality of raw sequencing reads. KneadData v0.7.2³ was used to trim and remove host sequences. MetaPhlAn2 v2.2.0 (13) was performed for taxonomy annotation and profiling, and HUMAnN2 v0.11.2 (14) was used to perform functional annotation in default settings. After we get the abundances of microbiomes, genes, and metabolic pathways, MEGAHIT v1.2.9 (15) was used to construct metagenome-assembled genomes, Prodigal v2.6.3 (16) was used to predict open reading frames, MetaBAT v2.12.1 (17) was used for genome binning, genome quality was checked by CheckM v1.1.3 (18), and taxonomy was identified by GTDB-Tk v1.0.2 (19).

For fecal samples at 0 d, 1 d, 7 d, 14 d, and 18 d, and cecum content samples at the end of the experiment (18 d), seven SCFAs (acetic acid, butyric acid, caproic acid, propionic acid, isovaleric acid, isobutyric acid, and valeric acid) were determined using a Trace 1,310 gas chromatograph coupled with an ISQ LT (Thermo Fisher Scientific, United States) (20). Briefly, 100 mg of sample was extracted in 50 μL of 15% phosphoric acid with 100 μL of 125 μg/mL 4-methyl valeric acid solution as an internal standard and 400 μL of ether and was centrifuged at 4°C for 10 min at 12000 rpm after vortexing for 1 min, and the supernatant was transferred into the vial before GC–MS analysis. An Agilent HP-INNOWAX (30 m × 0.25 mm ID × 0.25 μm) column was used. Helium was used as the carrier gas at 1 mL/min. Injection was made in split mode at 10:1 with an injection volume of 1 μL and an injector temperature of 250°C. The temperatures of the ion source and interface were 300°C and 250°C, respectively. The column temperature was programmed to increase from an initial temperature of 90°C, followed by an increase to 120°C at 10°C/min, to 150°C at 5°C/min, and finally to 250°C at 25°C/min, which was maintained for 2 min (total run time of 15 min). Mass spectrometric detection of metabolites was performed on ISQ LT (Thermo Fisher Scientific, United States) with electron impact ionization mode. Single ion monitoring (SIM) mode was used with an electron energy of 70 eV. Quantification was made using external standard curves.

For fecal samples at 0 d, 1 d, 7 d, 14 d, and 18 d, and cecum content samples at the end of the experiment (18 d), 30 tryptophan catabolites were determined using a Vanquish UHPLC (Thermo Fisher Scientific, Haverhill, MA) coupled with ABI Sciex 5,000 mass spectrometer (ABI Sciex, Concord, ON, Canada) according to previous work (21). Briefly, the sample was extracted in 100 μL of 80% methanol aqueous solution, grinding for 60 s, then 900 μL of 10% methanol aqueous solution, grinding for 120 s. Centrifugation was carried out at 12,000 rpm at 4°C for 10 min, and the supernatant was filtered; an equal amount of the internal standard solution was added and then transferred into the vial for analysis. An ACQUITY UPLC HSS T3 column (150 × 2.1 mm, 1.8 μm) (Waters, Milford, MA, United States) was used for chromatographic separation at 40°C. The mobile phase was composed of solvent A (0.1% formic acid in water) and solvent B (0.1% formic acid in methanol). The gradient was: 0 ~ 2 min, 1% B; 2 ~ 3 min, 1 ~ 30% B; 3 ~ 3.5 min, 30% B; 4.5 ~ 8 min, 30 ~ 50% B; 8 ~ 10 min, 50 ~ 95% B; 10 ~ 11 min, 95% B; 11 ~ 17 min, 95 ~ 1% B. The injection volume was 3 μL. The mass parameters were set as the following: 45 arbitrary units (AU), 13 AU, 1 AU, 350°C, and 350°C for sheath gas, aux gas, sweep gas, ion transfer tube, and vaporizer temperature, respectively. The ion source was operated using heated electrospray ionization (ESI) with an ion spray voltage set at 5500 V in positive ion mode. Polarity switching and scheduled selected reaction monitoring (SRM) were employed. Quantification was made using external standard curves.

For fecal samples at 0 d, 1 d, 7 d, 14 d, and 18 d, and colonic content samples at the end of the experiment (18 d), 39 bile acids were determined using an LC-30 (Shimadzu Corporation, Kyoto, Japan) tandem ABI Sciex 6,500 Plus mass spectrometry (ABI Sciex, Concord, ON, Canada) according to previous work (22). Briefly, the sample was extracted in 400 μL of methanol, vortexing and shaking for 60 s, adding 100 mg of glass beads, and grinding at 55 Hz for 60 s. Repeat the above operation at least twice: ultrasonic at room temperature for 30 min, centrifuged at 12,000 rpm at 4°C for 10 min, and 200 μL of supernatant mixed with 400 μL of water. Take 200 μL of supernatant and add 400 μL of water; vortex for 30 s; take 300 μL of supernatant and filter it through a 0.22 μm membrane; vortex for 30 s. After centrifugation, take 20 μL of supernatant and dilute it 10 times with 180 μL of 30% methanol solution, and add it to the assay vial. An ACQUITY UPLC® BEH C18 column (2.1 × 100 mm, 1.7 μm) (Waters, Milford, MA, United States) was used for chromatographic separation at 40°C. The injection volume was 5 μL. The mobile phase was composed of solvent A (0.01% formic acid in water) and solvent B (100% acetonitrile [ACN]). The gradient was 0 ~ 9 min, 30% B; 9 ~ 14 min, 30 ~ 36% B; 14 ~ 18 min, 36 ~ 38% B; 18 ~ 24 min, 38 ~ 50% B; 24 ~ 32 min, 50 ~ 75% B; 32 ~ 33 min, 75 ~ 90% B; 33 ~ 35.5 min, 90 ~ 30% B. The flow rate was 5 μL/min. The mass parameters were set as the following: the temperature of the ion source was 500°C, curtain gas (CUR) 30 psi, source gas 1 (GAS1) 30 psi, and source gas 2 (GAS2) 35 psi. The ion source was operated using heated electrospray ionization (ESI) with an ion spray voltage set at −4,500 V in negative ion mode. Multiple reaction monitoring (MRM) was employed. Quantification was made using external standard curves.

For serum samples at the end of the experiment (18 d), five choline derivatives (choline, betaine, trimethylamine N-oxide [TMAO], creatinine, and L-carnitine) were determined using a Jasper HPLC with SCIEX 4500MD Triple Quadrupole mass spectrometry (ABI Sciex, Concord, ON, Canada) according to previous work (23). Briefly, samples were added to a 2 mL centrifuge tube; 10 μL of internal standard solution, then add 750 μL of 1% formic acid-acetonitrile solution, vortex, centrifuge for 10 min at 12,000 rpm at 4°C, and then resolve and filter the supernatant by 0.22 μm membrane for analysis. An ACQUITY UPLC® BEH HILIC column (2.1 × 100 mm, 1.7 μm) (Waters, Milford, MA, United States) was used for chromatographic separation at 40°C. The injection volume was 5 μL. The mobile phase was composed of solvent A (10 mM ammonium formate) and solvent B (100% ACN). The gradient was 0 ~ 1 min, 80% B; 1 ~ 2 min, 80 ~ 70% B; 2 ~ 2.5 min, 70% B; 2.5 ~ 3 min, 70 ~ 50% B; 3 ~ 3.5 min, 50% B; 3.5 ~ 4 min, 50 ~ 80% B; 4 ~ 6 min, 80% B. The flow rate was 0.4 mL/min. The mass parameters were set as the following: The temperature of the ion source was 500°C, CUR 30 psi, GAS1 5 psi, and GAS2 50 psi. The ion source was operated using heated ESI with an ion spray voltage set at 5000 V in positive ion mode. Multiple reaction monitoring (MRM) was employed. Quantification was made using external standard curves.

For serum samples at the end of the experiment (18 d), untargeted metabolome was determined using a Vanquish UHPLC System with Orbitrap Exploris 120 MS (Thermo Fisher Scientific, United States) (24). Briefly, samples were thawed at 4°C, vortexed for 1 min, and mixed evenly; then they were transferred into a centrifuge tube and vortexed for 1 min after adding 0.4 mL methanol; then they were centrifuged for 10 min at 12,000 rpm at 4°C, and the supernatant was dried and resolved using 0.15 mL 2-chloro-l-phenylalanine (4 ppm) solution, and then filtered by 0.22 μ m membrane and detected by LC–MS. An ACQUITY UPLC HSS T3 column (150 × 2.1 mm, 1.8 μm) (Waters, Milford, MA, USA) was used for chromatographic separation at 40°C. The flow rate and injection volume were set at 0.25 mL/min and 2 μL, respectively. For LC-ESI⁺-MS analysis, the mobile phases consisted of (B2) 0.1% formic acid in acetonitrile (v/v) and (A2) 0.1% formic acid in water (v/v). Separation was conducted under the following gradient: 0 ~ 1 min, 2% B2; 1 ~ 9 min, 2% ~ 50% B2; 9 ~ 12 min, 50% ~ 98% B2; 12 ~ 13.5 min, 98% B2; 13.5 ~ 14 min, 98% ~ 2% B2; 14 ~ 20 min, 2% B2. For LC-ESI-MS analysis, the analytes were carried out with (B3) acetonitrile and (A3) ammonium formate (5 mM). Separation was conducted under the following gradient: 0 ~ 1 min, 2% B3; 1 ~ 9 min, 2% ~ 50% B3; 9 ~ 12 min, 50% ~ 98% B3; 12 ~ 13.5 min, 98% B3; 13.5 ~ 14 min, 98% ~ 2% B3; 14 ~ 17 min, 2% B3. The MS system with a heated electrospray ionization (ESI) source was operated in both positive and negative ion modes. The parameters were as follows: 3.50 kV and −2.50 kV for ESI⁺ and ESI⁻ source voltage and 325°C capillary temperature. Capillary voltages for negative and positive ionization modes were set at 2500 and 3,500 V, respectively. MS1 range was 100–1,000 m/z, and the number of data-dependent scans per cycle was 4; MS/MS resolving power was 15,000 FWHM, normalized collision energy was 30%, and dynamic exclusion time was automatic.

The raw MS data were first converted to mzXML format by MSConvert in the ProteoWizard software package (v3.0.8789) (25). and processed using XCMS (26) for feature detection, retention time correction, and alignment. The metabolites were identified by accurate mass (< 30 ppm) and MS/MS data, which were matched with HMDB,⁴ Massbank,⁵ LipidMaps,⁶ mzCloud,⁷ and KEGG.⁸

Data were analyzed by R language (Version 4.3.1) (27). Nonparametric tests of independent variables (tissue weights, immune factors, and morphological and SCFAs) between the two groups were performed using the Wilcoxon test.

Relative abundance of gut microbial genera, α -diversity, β -diversity, LEfSe differential analysis, and correlation analysis were performed in the R library “microeco” (version 0.5.1) (28). Functional profiling of metagenomes was analyzed using MicrobiomeAnalyst 2.0 (29).⁹ Untargeted metabolome, SCFAs, bile acids, and tryptophan were analyzed and visualized for differential analysis and PLS-DA, and enrichment analysis of targeted metabolites was performed in MetaboAnalyst 5.0 (30).¹⁰ Only metabolites present in >50% of the samples were kept for further analysis and then log2 normalized and Pareto-scaled before further statistical analysis. Association analysis of microbial abundance, contents of SCFA, bile acids, tryptophan metabolites, and morphological indexes was performed and visualized in the Hmisc package and heatmap package in R using Spearman rank correlation. All statistical analyses were considered significant at an adjusted p value less than 0.05 level unless specifically noted.

Diagrams of the experiment design and mechanism were visualized using icons from Bioicons¹¹ and Flaticon,¹² and graphs were paneled using Inkscape 1.3.¹³

Lactiplantibacillus plantarum HM-P2 treatment significantly increased (p = 0.032) the relative abundance of fecal L. plantarum of mice at 7 d of gestation, compared with the control group (Figure 2A). This showed that L. plantarum HM-P2 successfully colonized the gut of GF mice at this stage of gestation.

Germ-free mice were fecal microbiota transplanted with stool from healthy pregnant women and were fed a diet supplemented with L. plantarum HM-P2 from gestation to delivery (0–18 d). The composition and dynamic changes of the stool microbiome after HM-P2 treatment were studied using shotgun metagenomic sequencing. A total of 13,636,188,396 bp raw sequences were obtained from 50 samples. The average raw data of 272,723,767.9 bp was obtained, and then the data were assigned to 5 kingdoms, 104 phyla, 85 classes, 163 orders, 348 families, 1,315 genera, and 5,786 species.

Composition analysis showed that Firmicutes (~50%), Bacteroids (~20%), Verrucomicrobia (~15%), Proteobacteria (<5%), Actinobacteria (<5%), Fusobacteria, and so on, were the dominant phyla in all samples (Figure 2B). At the genus level, Akkermansia, Bacteroids, Lactobacillus, Clostridium, Bifidobacterium, Roseburia, Eubacterium, Phascolarctobacterium, Butyricimonas, and Blautia were the dominant genera (Figure 2C).

𝛼-diversity indexes (Shannon) from 0 to 18 d of gestation showed that no significant difference was found between the HM-P2 and control groups (Figure 2D). Principal coordinate analysis (PCoA) showed that there was a moderate separation of the HM-P2 group and control group, and PC1 and PC2 together explained 57.1% of the total difference (Figure 2E). Bray-Curtis distance was not significantly different between the HM-P2 group and the control group (Figure 2F).

Differential analysis by linear discriminant analysis Effect Size (LefSe) (“Group” as the main effect and “Phase” as the sub-effect) showed that L. reuteri, Anaerofilum sp. An201, and Gemmiger were significantly higher (adjusted p < 0.05) in the HM-P2 group, while Bacteroides thetaiotaomicron sp. CAG:40, Bacteroides sp. 1_1_14, Bacteroides fragilis, Subdoligranulum sp. 4_3_54A2FAA, and Ruthenibacterium lactatiformans were significantly higher in the control group (adjusted p < 0.05) (Figure 2G).

A total of 89 metagenomic genes were differentially expressed in the HM-P2 group compared with the control group at 18 d. These differentially expressed genes were significantly (p < 0.05) enriched in starch and sucrose metabolism (see Supplementary Table S1).

The influence of human breast milk-derived L. plantarum HM-P2 on the morphometric indices of the cecum and colon was measured (Figure 3A). In the colon, top crypt depth was significantly higher in the HM-P2 group than the control group (Figure 3B); colon middle crypt depth showed no significant difference between the two groups (Figure 3C). In the cecum, crypt depth was significantly higher in the control group compared with the HM-P2 group (Figure 3D).

In fecal samples, the contents and changes of microbial bile acids are shown in Figure 4A. Among them, the contents of primary bile acids (unconjugated, ursocholic acid [UCA; Figure 4B), allo-cholic acid (ACA; Figure 4C)] and secondary bile acids [unconjugated, isolithocholic acid (isoLCA; Figure 4D)] showed significant differences between the HM-P2 and control groups. Other bile acids were not significantly different between the two groups.

In fecal samples, the contents and changes of 21 tryptophan metabolites are shown in Figure 4E. Among them, the concentration of serotonin (5-hydroxytryptamine, 5-HT) revealed a significant difference between the HM-P2 and control groups (Figure 4F). The other tryptophan metabolites were not significant between the two groups.

The contents of short-chain fatty acids (SCFAs) in cecum contents are shown in Figure 5A. Compared to the mice in the control group, caproic acid significantly increased in the HM-P2 group at 18 d of gestation (Figure 5B), and other SCFAs (acetic acid, propionic acid, isovaleric acid, valeric acid, and isobutyric acid) showed no significant difference between the two groups at 18 d of gestation.

The contents of bile acids in the cecum contents are shown in Figure 5C. Among 30 identified bile acids, allolithocholic acid was significantly higher in the HM-P2 group at 18 d of gestation compared with the control group (Figure 5D). Other bile acids were not significantly different between the two groups.

Untargeted metabolites of serum samples at 18 d of gestation were analyzed in both positive and negative modes. A total of 448 (287 positive modes and 161 negative modes) serum metabolites were identified and quantified, with the top 30 most abundant metabolites presented in Figure 6A. Sparse partial least-squares discriminant analysis (sPLS-DA) showed that there were clear differences between these two groups (Figure 6B). There were 17 positive and 6 negative metabolites significantly different in the control and HM-P2 groups. Enrichment analysis of significant difference metabolites showed that they were enriched mainly in fatty acid biosynthesis, glycerophospholipid metabolism (Figure 6C).

In addition, compared with the control group, oral administration of L. plantarum HM-P2 during gestation of humanized GF mice showed no significant differences on the offspring numbers (See Supplementary Figure S1), and the fresh weight of five types of tissues (liver, left lobe of liver, left and right kidney, and spleen; See Supplementary Table S2), and SCFAs of fecal (See Supplementary Figure S2), and tryptophan of cecum contents (See Supplementary Figure S3), and serum immune factors, such as immunoglobulin M (IgM), and immunoglobulin G (IgG), and inflammatory cytokine factors, such as interferon-gamma (IFN-γ), transforming growth factor-beta-1 (TGF-β-1), tumor necrosis factor alpha (TNF-α), interleukin-10 (IL-10), interleukin-8 (IL-8), interleukin-6 (IL-6), interleukin-4 (IL-4), transforming growth factor-alpha TGF-α, interleukin-22 (IL-22), interleukin-17 (IL-17), interleukin 1-beta (IL1-β), interleukin 1-alpha (IL1-α; See Supplementary Table S3), and the serum choline derivatives (choline, trimethylamine N-oxide (TMAO), betaine, creatinine, and carnitine; See Supplementary Figure S4).

Spearman correlation showed that microbial genes encoding SCFA biosynthesis were significantly correlated with SCFA contents. Fecal microbial acyl-CoA dehydrogenase (acdA) was negatively correlated with the content of fecal acetic acid, butyrate kinase (buk) was positively correlated with the content of fecal butyric acid (p < 0.05, rho = 0.685) and isobutyric acid (p < 0.05, rho = 0.77), while 3-sulfolactaldehyde reductase (yihU) showed an opposite relationship (p < 0.05, rho = −0.72) with butyric acid; propionate catabolism operon regulatory protein (prpR) was negatively correlated with the content of fecal propionic acid (p < 0.05, rho = −0.673; Figure 7A).

To further confirm probiotic effects on gut microbiome, microbial metabolites, and intestinal morphometric indices, Spearman correlation analysis was performed. Results showed that gut microbes were associated with those significantly different bile acids and tryptophans. Subdoligranulum sp. 4–3-54A2FAA and Ruthenibacterium lactatiformans were negatively correlated with cecum caproic acid (p < 0.05, both rho = −0.74), and B. bifidum was positively correlated with fecal serotonin (p < 0.05, rho = 0.48; Figure 7B).

Due to the limited number of GF mice used in this study, the effects of L. plantarum HM-P2 on the offspring’s health were not investigated. Future studies may include examining the effects of L. plantarum HM-P2 on offspring gut microbial metabolism and health of humanized GF mice. Additionally, we may increase the sample size and explore the effects of L. plantarum HM-P2 in clinically controlled trials of gestational and gastrointestinal diseases such as gestational diabetes mellitus (GDM) and irritable bowel syndrome, functional gastrointestinal disorders, and necrotizing enterocolitis.