The fetuses were harvested from nulliparous Angus-cross heifers on day 83 ± 0.27 of gestation via ovariohysterectomy. The detailed information with respect to the experimental design, diet, feeding and animal husbandry has been described previously (Diniz et al., 2021; Menezes et al., 2021). Briefly, 35 angus-cross heifers (initial BW 359.5 ± 7.1 kg) were used in a randomized complete block design with a 2 × 2 factorial treatment arrangement. The main factors included vitamin and mineral supplementation (VTM vs. NoVTM), and the rate of gain [low gain (LG) vs. moderate gain (MG)]. Heifers received a diet of 0.45 kg ground corn/heifer/day and 113 g/heifer/day of VTM premix (Purina Wind and Rain Storm All-Season 7.5 Complete, Land O’Lakes, Inc., Arden Hills, MN, United States) for at least 71 days before artificial insemination (AI) and continuing through the first 83 days of gestation. On the day of AI, within each VTM or NoVTM group, heifers were randomly assigned to one of two rate of gain groups: LG or MG. The LG group heifers were targeted to gain 0.28 kg/d during gestation and were fed a basal total mixed ration [consisting of triticale hay, corn silage, modified dried distillers’ grains plus solubles (DDGS), and ground corn]. The MG group heifers were fed the basal total mixed ration plus the starch-based protein/energy supplement (a blend of ground corn, DDGS, wheat midds, fish oil, urea, and ethoxyquin) to achieve the targeted MG of 0.79 kg/d. Heifers receiving the VTM supplement (LG-VTM and MG-VTM) continued to receive the same concentration of VTM as during the pre-insemination period until the day of fetal collection. Further details on the composition of the VTM supplement, and nutrient composition of TMR and supplements provided to the heifers during the first trimester of gestation are provided in our previous publication (Menezes et al., 2021). Heifers were bred by AI using female-sexed semen from a single sire. Pregnancy diagnosis was performed 35 days after AI, and fetal sex was determined on day 65 using transrectal ultrasonography. Only heifers confirmed pregnant with a female fetus continued on the experiment.

On day 83 ± 0.27 of gestation, gravid reproductive tracts (n = 33) were collected via ovariohysterectomy as described previously (McLean et al., 2016). Immediately upon removal of the uterus, 20 ml of allantoic and amniotic fluid samples (Supplementary Figure 1A) were collected using a sterile 20-mL Luer-lock syringe with a sterile 18-gauge needle as shown in Supplementary Figure 1B and were transferred into 50-mL Falcon tubes and immediately snap frozen using liquid nitrogen. In addition, the fetal intestine (Supplementary Figure 1C) was removed and one third portion of the intestine was dissected. The fetal (cotyledon) portion of the placenta was also dissected. Fetal intestinal and placental tissue samples were wrapped in aluminum foil and snap frozen. All samples were stored at −80°C until analyses.

DNA from the allantoic and amniotic fluids, and fetal intestinal and placental tissue samples was extracted using cetyltrimethylammonium bromide buffer (CTAB) and a phenol:chloroform based method as described previously (Fujimura et al., 2016; Rackaityte et al., 2020) with some modifications. Ten milliliters of allantoic or amniotic fluid was centrifuged at maximum speed (16,000 × g) for 30 min at 4°C using a floor ultra centrifuge (Beckman Optima L-60 Ultracentrifuge). After removal of the supernatant, the pellet was resuspended with 500 μl prewarmed CTAB buffer (10% CTAB, 0.7 M NaCl, 240 mM potassium phosphate buffer pH 8), and transferred to a 2 mL Lysing Matrix E tube (containing 1.4 mm ceramic spheres, 0.1 mm silica spheres and one 4 mm glass bead) (MP Biomedical LLC, Irvine, CA, United States). The cell suspension was incubated in a ThermoMixer (Eppendorf, Enfield, CT, United States) for 15 min at 65°C and 800 rpm followed by mechanical cell lysis with bead beating (5.5 m/s, 30 s) in a FastPrep-24 (MP Biomedical LLC). Then, 500 μl of phenol:chloroform:isoamyl alcohol (25:24:1) was added into the cell lysate and mixed by manual shaking for 20 s before centrifugation at 16,000 × g for 5 min. The supernatant was carefully transferred into a heavy phase-lock gel tube (QuantaBio, Beverly, MA, United States), and mixed with an equal volume of chloroform and centrifuged at 16,000 × g for 5 min. The aqueous phase was then transferred into a sterile 1.5 mL tube and 1 μl linear acrylamide solution was added and vortexed. Then, 2 volumes of PEG-NaCl solution were added to the mixture and vortexed followed by incubation at 21°C for 2 h. The isolated DNA was pelleted by centrifugation at 14,000 × g for 30 min and then washed twice with 70% ethanol. Finally, the isolated DNA was resuspended with 30 μl of 10 mM Tris–Cl pH 8, and the concentration was measured using a NanoDrop ND-1000 spectrophotometer and stored at −20°C.

For DNA extraction from the intestinal and placental tissue samples, approximately 50 mg of tissue was dissected on a sterile empty Petri dish in a biosafety cabinet and placed in a 2 mL Lysing Matrix E tube. The remaining DNA extraction procedures were the same as described for fetal fluids except for an additional cycle of bead beating for the intestinal and placental tissue samples. DNA extraction from the nuclear free water (Corning Molecular Biology Grade Water, Corning, Glendale, AZ, United States) was performed to capture any potential microbial DNA contamination associated with the CTAB-based method.

Amplification and sequencing of the 16S rRNA gene were performed at MR DNA (Shallowater, TX, United States) as described previously (Santiago-Rodriguez et al., 2016). A total of 138 samples which include 8 negative extraction control samples (two per sample type) was sequenced. Briefly, the V4 region of the 16S rRNA gene was amplified using the 515F (5′-GTGCCAGCMGCCGCGGTAA-3′) and 806R (5′-GGACTACHVGGGTWTCTAAT-3′) primers. A PCR clamp was also included to block amplification of bovine host DNA (Lundberg et al., 2013). PCR amplification was conducted using the HotStarTaq Plus Master Mix Kit (Qiagen Inc., Germantown, MD, United States). After amplification, PCR products were electrophoresed on a 2% agarose gel to ensure the correct size and band intensity. The 16S rRNA gene amplicons were indexed and pooled in equimolar concentrations and then purified using calibrated Ampure XP beads. Then the pooled and purified PCR product was used to prepare an Illumina DNA library. These 16S rRNA gene libraries were then sequenced on an Illumina MiSeq instrument (Illumina Inc., San Diego, CA, United States) using the MiSeq reagent kit v3 (2 × 300 bp) following manufacturer’s instructions.

The 16S rRNA gene sequences were processed using DADA2 v. 1.18 (Callahan et al., 2016) in R. 4.0.3. Briefly, the forward and reverse reads were truncated at 210 bp, merged, chimeric sequences removed, and taxonomy assigned to each merged sequence, referred to here as operational taxonomic units (OTUs) at 100% similarity, using the naïve Bayesian RDP classifier (Wang et al., 2007) and the SILVA SSU database release 138 (Quast et al., 2013). All OTUs that were found in the negative (extraction) control samples and likely to be contaminants were removed prior to analyses as were those OTUs classified as chloroplasts, eukaryote, or mitochondria. The number of OTUs per sample (richness), the Shannon and inverse Simpson’s diversity indices, and Bray-Curtis dissimilarities were calculated in R using Phyloseq 1.34.0 (McMurdie and Holmes, 2013) and vegan 2.5–7 (Oksanen et al., 2013). To account for uneven sequence depths, samples were randomly subsampled to 14,000 prior to the calculation of Bray-Curtis dissimilarities and diversity measures for comparisons among different sample types. For assessment of treatment effects within each sample type, the allantoic and amniotic fluid, and intestinal and placental tissue samples were randomly subsampled to 28,000, 20,000, 14,000, and 21,500 sequences, respectively. All 16S rRNA gene sequences are available in the NCBI’s sequence read archive under BioProject PRJNA731173.

The bacterial concentration was estimated for the fetal intestine and placental tissue (n = 33 each) samples using qPCR as previously described (Amat et al., 2020) with some modifications. Briefly, The primers 515Fq (5′-GTGYCAGCMGCCGCGGTAA-3′) (Parada et al., 2016) and 806Rq (5′-GGACTACNVGGGTWTCTAAT-3′) (Apprill et al., 2015) were used to amplify the V4 region of the 16S rRNA gene. Each qPCR mixture contained 1 SsoAdvanced Universal SYBR Green Supermix (Bio-Rad Laboratories, Inc., Hercules, CA, United States), 0.4 μM (each) primer, 0.1 μg/μl bovine serum albumin (New England Biolabs, Pickering, ON, Canada), and 30 ng of DNA, in a total volume of 25 μl. A CFX96 Touch Real-Time PCR Detection system (Bio-Rad Laboratories Ltd.) was used with the following conditions: an initial denaturation at 95°C for 3 min, followed by 40 cycles at 95°C for 25 s, 50°C for 30 s, and then 72°C for 45 s. Standard curves (10²–10⁸ gene copies) were generated using the pDrive cloning vector (Qiagen Inc.) containing the PCR product from 16S rRNA gene. All qPCRs were performed in duplicate with standards (1 μl) and no-template controls (1 μl of nuclease-free H₂O), as well as a positive control (DNA from a bovine deep nasopharyngeal swab). A melt curve analysis was performed following qPCR amplification to ensure that only target genes were amplified. The copy number was normalized by 30 ng of input DNA.

Scanning electron microscopy (SEM) imaging was performed on some of the allantoic and amniotic fluid samples that had greatest DNA concentrations. The SEM imaging was carried out by the NDSU Electron Microscopy Center core facility. Allantoic and amniotic fluids (4 ml) were centrifuged for 10 min at 14,000 × g to produce a pellet, then the pellet was resuspended in deionized water. This process was repeated twice to remove buffer salts. Pelleted material was then applied to round glass coverslips affixed to aluminum mounts with silver paint (SPI Supplies, West Chester, PA, United States), air dried, and sputter coated with carbon (Cressington 208c, Ted Pella Inc., Redding, CA, United States). Images were obtained using a JEOL JSM-7600F scanning electron microscope at an accelerating voltage of 2 kV or a JEOL JSM-6490LV scanning electron microscopy at an accelerating voltage of 15 kV (JEOL USA Inc., Peabody, MA, United States).

Differentially abundant genera within the 20 most relatively abundant genera among the four sample types were identified using an ANOVA with sample type and treatment in the model. The Benjamini-Hochberg procedure was used to correct all P-values for multiple comparisons. The effect of fetal sample type and the treatment within each sample type on the microbial community structure was assessed using Bray-Curtis dissimilarities and PERMANOVA (adonis2 function). Pairwise comparisons of the Bray-Curtis dissimilarities among sample types and treatment groups were done using the R package pairwise Adonis v. 0.01 with the Benjamini-Hochberg procedure used to correct P-values for multiple comparisons.

The number of OTUs (richness), diversity indices, relative abundance of the most relatively abundant phyla and genera between sample types (allantoic and amniotic fluids, and intestinal and placental tissues) were compared using the generalized liner mixed model estimation procedure (PROC GLIMMIX) in SAS (ver. 9.4, SAS Institute Inc., Cary, NC, United States). The data regarding the comparison among different dietary treatment groups within each sample type were analyzed as 2 × 2 factorial treatment arrangement using the PROC GLIMMIX. The model included main effects of VTM supplementation (VTM or NoVTM), rate of gain (LG or MG) and the respective interaction. Means among different sample types or treatment groups within each sample type were compared using the LSMEANS statement and significance was declared at P < 0.05.

An average of 105,458 ± 77,752 (SD) 16S rRNA gene sequences per sample (min. = 14,029; max. = 514,084) were obtained from 130 allantoic and amniotic fluid, and intestinal and placental tissue samples from the 83-day-old calf fetuses. From these sequences, a total of 8,544 archaeal and bacterial OTUs (at 100% sequence similarity) were identified and classified into 39 unique phyla (3 archaeal and 36 bacterial phyla), and 921 unique genera.

A relatively diverse and unique microbial community was detected in the four fetal sample types (Figure 1). The microbial community structure was significantly different between the allantoic and amniotic fluid, and intestinal and placental tissue-associated microbiota (0.047 ≥ R² ≥ 0.019, P ≤ 0.031) (Figure 1A). Alpha diversity metrics also differed by fetal sample type (Figures 1B,C). Allantoic fluid had a greater (P < 0.05) microbial richness (number of OTUs, at 97% sequence similarity) (122 ± 10) compared with amniotic fluid (84 ± 6) and the intestinal (63 ± 7) and placental (66 ± 6) tissue. The number of OTUs in the amniotic fluid was similar to the placenta (P > 0.05) but greater than in the intestine (P < 0.05). Microbial diversity (Shannon index) was similar for the intestinal and placental samples, and both were less diverse compared to the fetal fluid (P < 0.05).

Among the 39 different phyla detected across all fetal samples, Proteobacteria (54.8%) was the most relatively abundant phylum followed by Firmicutes (16.3%), Actinobacteriota (13.7%), Bacteroidota (5.0%), and Acidobacteriota (2.6%) (Figure 1D). Archaeal phyla (Crenarchaeota, Euryarchaeota, and Nanoarchaeota) accounted for only 0.34% of the 16S rRNA gene sequences (Supplementary Table 1). The relative abundance of some of the bacterial phyla also differed by fetal sample type (Figure 1D). Proteobacteria was less relatively abundant in amniotic fluid (49.4%) compared to allantoic fluid (61.6%), and intestinal (55.7%) and placental (53.6%) tissues (P < 0.05). The fetal intestine (5.5%) and placenta (3.2%) harbored a greater relative abundance of Acidobacteriota compared to the allantoic (0.5%) and amniotic (1.0%) fluids (P < 0.05).

At the genus level, Acidovorax, Acinetobacter, Stenotrophomonas, Brucella, Anoxybacillus, Sphingomonas, Allorhizobium-Neorhizobium-Pararhizobium-Rhizobium (ANPR), Salinisphaera, and Lactobacillus were the most relatively abundant genera among all samples (Figure 2). Although between sample variance was relatively high, Acidovorax, Acinetobacter, Brucella, Corynebacterium, Enterococcus, Exiguobacterium, and Stenotrophomonas were differentially abundant between different sample types (P < 0.05). Acidovorax was equally abundant across amniotic fluid, and intestinal and placental tissues (P > 0.05) but less abundant in allantoic fluid (P < 0.05). Placental samples had a greater relative abundance of Acinetobacter compared to other sample types. ANPR and Brucella were more relatively abundant in allantoic fluid, whereas Salinisphaera and Aphanizomenon were enriched in amniotic fluid (P < 0.05).

No single OTU was identified in 50% of all the allantoic and amniotic fluid, and intestinal and placenta samples. As shown in the heatmap of the 100 most abundant OTUs (Figure 3), there was considerable inter-individual variation in both prevalence and abundance of most of these taxa. Certain OTUs including OTU96 (Pseudomonas), OTU22 (Acinetobacter), OTU78 (Achromobacter), OTU18 (Brucella), OTU26 (ANPR), OTU25 (Caulobacteraceae) and OTU47 (Saccharopolyspora) were present in the majority of the allantoic fluid samples with greater abundance. From these OTUs, OTU18 (Brucella), OTU26 (ANPR), and OTU47 (Saccharopolyspora) were also more prevalent in the amniotic fluid samples. Relatively more prevalent taxa detected from fetal intestinal samples are OTU17 (Anoxybacillus) and OTU70 (Microbacteriaceae). In placental tissue, two taxa (OTU19 and OTU14) within the genus Acinetobacter were more frequently detected.

A Venn diagram displaying the number of unique and shared OTUs among different sample types is displayed in Figure 4. Overall, the proportion of unique OTUs identified in amniotic and allantoic fluid, and intestinal and placental microbiota accounted for 33, 23, 16, and 17% of the OTUs, respectively. A total of 223 (3%) OTUs were shared by allantoic and amniotic fluids, and 2% of the OTUs were shared between intestine and placental tissues. Less than 1% taxa were shared between fetal fluid and tissue samples or between the three of the four sample types. Only 55 OTUs (<1% of the total OTUs) were shared among the four different microbial communities.

To further confirm the 16S rRNA gene sequencing results we quantified the total bacterial concentration in the fetal intestinal and placental tissue samples using qPCR. The total bacterial concentration in placental tissue samples was estimated to be 46% greater than that of intestinal tissue (P < 0.0001). Next, to provide visual evidence of the presence of microbial cells in the fetal fluid samples, SEM imaging was performed on select amniotic (n = 2) and allantoic fluid (n = 3) samples. Microbial cells with prokaryotic cell structure (cocci, spherical) were detected in two allantoic fluid samples (Figure 5).

We also evaluated whether the fetal microbiota in early gestation would diverge in response to in utero environment resulting from altered maternal nutrition. To this end, we analyzed the sequencing data by maternal nutritional treatment groups within each fetal sample type. Microbial community structure did not differ between treatment groups in any of amniotic and allantoic fluid, and intestinal and placental microbiota in 83-day-old bovine fetuses (0.0152 ≥ R² ≥ 0.0762, P ≥ 0.18) (Figure 6).

No effects of VTM supplementation, rate of gain or their respective interaction were observed for microbial richness (number of OTUs) of amniotic fluid- and placenta-associated microbiota (P ≥ 0.122) (Figure 7 and Supplementary Table 2). However, the number of OTUs observed in allantoic fluid samples derived from the fetuses born from the LG-NoVTM heifers was greater than that of MG-NoVTM fetuses (P < 0.05). The interaction effect was significant for the microbial richness of intestinal microbiota (P = 0.05), with MG-VTM fetuses had greater microbial richness in their intestine compared to MG-NoVTM fetuses (P < 0.05). Microbial diversity (Shannon and Inverse Simpson’s diversity indices) was not impacted by main effects of VTM, rate of grain, or their respective interactions in any of the fetal sample type (P > 0.10).

When the microbial composition was compared at the phylum level, no significant differences were noted between treatment groups for any of the five most relatively abundant phyla in allantoic fluid (P > 0.05) (Figure 7 and Supplementary Table 2). However, the relative abundance of some of the five most relatively abundant phyla in amniotic, and intestinal and placental tissues was affected by treatment (VTM supplement or rate of gain or the interactions). For example, the effect of gain and VTM interaction on Proteobacteria in both amniotic fluid and intestinal tissue was significant (P ≤ 0.04). The relative abundance of Cyanobacteria in amniotic fluid and Firmicutes in intestine was affected by VTM supplementation (P ≤ 0.014). Rate of gain tended to affect the relative abundance of Bacteroidota (P = 0.057) in amniotic fluid and Firmicutes (P = 0.066) in placental tissue.

At the genus level, among the eight most relatively abundant genera listed in Table 1, the relative abundance of Sphingomonas in allantoic fluid, Salinisphaera in the intestine and Streptococcus in the placenta was affected by the rate of gain (P ≤ 0.037) (Table 1). A significant effect of VTM supplementation was observed on the relative abundance of Aphanizomenon (P = 0.010). The effect of VTM supplementation (P = 0.072), and an interaction between VTM and rate of gain (P = 0.067) tended to be significant for the relative abundance of Anoxybacillus in placenta. The relative abundance of the remaining eight most predominant genera was not influenced by the treatment (P > 0.10).