Compositional Changes in the Vaginal Bacterial Microbiome of Healthy Pregnant Women across the Three Gestational Trimesters in Ismailia, Egypt

The composition of the vaginal microbiome may lead to adverse pregnancy outcomes. Normal pregnancy is associated with changes in the vaginal bacterial community composition, which tend to be more enriched with one or two Lactobacillus species promoting a healthy vagina and favorable birth outcomes. The aim of the current study was to determine compositional changes in the healthy vaginal microbiome composition during the three trimesters of pregnancy in Ismailia, Egypt using Illumina MiSeq sequencing of the V3–V4 region of the 16S rRNA. The phylum Firmicutes and the genus Lactobacillus dominated across the three trimesters of pregnancy. L. iners was the most abundant species. However, L. coleohominis and L. reuteri represented the least dominant vaginal lactobacilli. Core microbiome analyses showed the Lactobacillus genus and L. iners species to have the highest prevalence in all the samples of our study groups. The phylum Firmicutes was found to be negatively correlated with almost all other vaginal phyla during pregnancy. Likewise, a negative correlation between Lactobacillus and almost all other genera was detected, including significant negative correlations with Dialister and Prevotella. Furthermore, negative correlations of L. iners were detected with almost all other species, including a significant negative correlation with L. helveticus, G. vaginalis, S. anginosus, and S. agalactiae.

The vaginal microbiome of healthy pregnant women tends to be more enriched and dominated with one or two Lactobacillus species compared to the non-pregnant state [18,19]. The hormone levels in pregnancy favor the dominance of Lactobacillus species while Gardnerella vaginalis and other anaerobic bacteria decrease. Pregnancy is associated with high levels of circulating estrogen produced from the ovaries and the placenta [5,20]. The elevated estrogen levels promote glycogen deposition in the vaginal epithelial cells, which, in turn, promotes the proliferation of different Lactobacillus species. Due to these changes, the composition of the vaginal microbiome is stable with limited variations during pregnancy [5,17,21,22]. Such dominance of Lactobacillus species has been linked to the low production of proinflammatory cytokine, healthy vagina, and favorable birth outcomes [5,23,24].

The pH Measurement
Vaginal pH of non-pregnant women (control group) was measured using commercial pH strips in the pH range 4.5-10.0 (Sigma, Germany, Cat. No. P4536). The color of the test pH indicator was compared to a standard color chart provided with the pH applicator. Non-pregnant women with pH exceeding 4.5 were excluded from the study [21,28,37]. As recommended by Agaard et al., 2012 [28], vaginal pH was not recorded for pregnant participants; however, they were excluded if presenting any signs and symptoms for bacterial vaginosis.

DNA Extraction
Samples were released from −80 • C conditions and allowed to reach room temperature. Microbial DNA was extracted using Dneasy PowerSoil DNA extraction kit (Qiagen, Valencia, CA, USA, Cat. No. 12888-50) [38,39], according to manufacturer's instructions. Extraction steps were conducted after gentle vortex of the collection tubes containing swab tips to induce better dispersion of the trapped bacteria.
PCR reactions were performed in volumes of 100 µL using OnePCRTM Ultra Master Mix (GeneDirex ® , Taichung City, Taiwan, Cat. No. MB208-0100). PCR was performed in a thermal cycler (BIORAD T100 Thermal Cycler) with the following conditions: Initial denaturation at 95 • C for 3 min, followed by 25 cycles of denaturation at 95 • C for 30 s, annealing at 55 • C for 30 s, and extension at 72 • C for 30 s. Then, final extension at 72 • C for 5 min. Size and quality of PCR amplification products was verified with agarose gel electrophoresis. The primer pair sequences used for the V3-V4 region were employed to create a single amplicon of approximately 550 bp. Amplicons from negative controls were analyzed to detect false-positive reactions due to contaminants in swabs, extraction reagents, and PCR mixture. Library preparation and high-throughput Illumina MiSeq paired-end 2 × 300 base sequencing was performed at IGA Technology Services (Udine, Italy). Negative controls, including an empty swab for DNA extraction and amplified nuclease-free water without extracted DNA, were also sequenced and analyzed along with the samples.

Bioinformatic Analysis
Raw sequences were imported into Quantitative Insights into Microbial Ecology 2 (QIIME2 version 2021.4) [33] as Casava 1.8 paired-end demultiplexed fastq format. Analy-  Figure S1 panels A and B) by randomly subsampling 10,000 out of 5,930,186 sequences without replacement. This was done using the QIIME2 command "qiime demux summarize". Positional quality plots were then used to decide the best positions for trimming and truncation. Then, we used DADA2 [41] through qiime dada2 plugin for filtering out noisy sequences, correcting errors in marginal sequences, removing chimeric sequences, removing singletons, joining denoised paired-end reads, and then dereplicating these sequences producing unique amplicon sequence variants (ASVs) feature table. The following parameters were used for DADA2 pipeline: forward read sequences were truncated at 300, while reverse read sequences were truncated at 260. The first 24 bases at 5 end of each sequence were trimmed for both forward and reverse sequences [42]. Taxonomy assignment of ASVs was performed based on trained RDP's naive Bayesian classifier at 99% sequence similarity against the Greengenes [43] database (https://data.QIIME2 .org/2021.4/common/gg-13-8-99-nb-classifier.qza, accessed on 11 August 2022). Further taxonomic classification of unrecognized features in our data set was performed using NCBI-BLAST links, which were complementary to QIIME2 and provided with dada2rep-seq.qzv visualization file produced from DADA2 pipeline. Sequences were rarefied with QIIME2 script "qiime diversity alpha-rarefaction" command. Alpha diversity (within sample variation) and beta diversity (between sample variation) analysis were performed through running the QIIME2 scripts "qiime diversity core-metrics-phylogenetic" command. Statistical significance and visualization of violin plots of the alpha diversity indices (Observed, Chao1, Shannon, and Simpson) and Pielou's evenness was performed using "ggstatsplot" R package, using the function "ggbetweenstats"; the p value was measured by pairwise Kruskal-Wallis test and adjusted using Holm adjustment. Dunn pairwise test was also performed. Beta diversity distances, including Bray-Curtis, Jaccard, and unweighted and weighted UniFrac distances, were represented via generating principal coordinates analysis (PCoA) plots at feature level through MicrobiomeAnalyst [44,45] using analysis of group similarities (ANOSIM) statistical method. Pairwise permutational multivariate analysis of variance (PERMANOVA) test in QIIME2 using pseudo-F statistic was also done. This was applied to measuring the amount of variation in beta diversity between groups. A p value < 0.05 is considered statistically significant. Taxonomy bar plots, core microbiome, and discriminative taxa between study groups were analyzed using MicrobiomeAnalyst (https://www.microbiomeanalyst.ca/, accessed on 15 December 2022) [44,45] and "ggplot2" R package of R (v 4.2.0). Discriminative taxa between study groups were measured by linear discriminant analysis (LDA) effective size (LEfSe). LDA score ≥ 2 and a p value < 0.05 was considered statistically significant. Correlations between taxa across pregnancy trimesters and non-pregnancy state were analyzed by Spearman correlation analysis using the "rcorr" function of R package "Hamsic" (r ≥ ±0.6, p ≤ 0.05). The "ggpairs" function of the R package "GGally" was used for plotting correlation between pairs of the vaginal microbiota during each trimester of pregnancy at phylum, genus, and species levels.

Community State Types (CST) Assignment
Taxonomic profiles of our vaginal bacterial communities were classified into CSTs using VALENCIA (VAginaL community state typE Nearest CentroId classifier) [46]. This is a nearest centroid-based tool, which sorts samples according to their similarity to a set of reference centroids.

Data Availability
The raw sequences were submitted in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA), under the BioProject ID PRJNA877217 https://www.ncbi.nlm.nih.gov/sra/PRJNA877217 (accessed on 15 September 2022) involving accession numbers from SAMN30696179 to SAMN30696228.

Sequencing Data Profiles
A total of 4,051,265 read counts were maintained after the quality trimming of Illumina MiSeq raw sequencing data of 48 vaginal samples. The average read counts per sample was 84,401, while minimum and maximum counts per sample were 36,027 and 149,731, respectively. The total number of ASVs features that remained was 927. The plateau was reached at 20,000 within the rarefaction curves (Supplementary Figure S2 Table S2). The relative abundance of Firmicutes was the highest in the first trimester (97.7%), compared to the second trimester (96.2%), and non-pregnant control group (92.4%), while it was the least abundant in the third trimester (86.3%). However, the relative abundance of Actinobacteria and Tenericutes were the highest in the third trimester (6.4%, 4.4%, respectively) compared to the second trimester (1.4%, 0.9%, respectively) and non-pregnant control group (5.2%, 0.9%, respectively), while they were relatively the least abundant in the first trimester (1.2%, 0.4%, respectively). in the third trimester compared to 0.19% in the first trimester, while its relative abundance was the least in the second trimester and non-pregnant control group (0.09%, 0.085%, respectively). The relative abundance of L. reuteri was 0.49% the in third trimester, 0.42% in the second trimester, and 0.29% in the non-pregnant control group, while its abundance was the least in the first trimester (0.2%).  Figure 1D is re-represented as Supplementary Figure S4 with different y-scale to show the relative abundance of minor non-Lactobacillus species). Negative control represents samples with empty swab for DNA extraction and amplified nuclease-free water without extracted DNA.
Among the non-Lactobacillus species ( Figure 1D, Supplementary Table S5), Bifidobacterium breve, Peptostreptococcus anaerobius, Streptococcus anginosus, Ureaplasma parvum, and Streptococcus agalactiae were frequently detected. The relative abundance of B. breve, P. anaerobius, and S. anginosus was the highest in the non-pregnant control group. The relative abundance of B. breve was 2.5% in the non-pregnant control group compared to 0.04% in the first trimester, while it was nearly absent in the second and third trimesters. Similarly, the relative abundance of P. anaerobius was 0.29% in the non-pregnant control group compared to 0.28% in the first trimester, while its relative abundance was the lowest in the  Figure 1D is re-represented as Supplementary Figure S4 with different y-scale to show the relative abundance of minor non-Lactobacillus species). Negative control represents samples with empty swab for DNA extraction and amplified nuclease-free water without extracted DNA.
The most abundant genera ( Figure 1B, Supplementary Table S3) were Lactobacillus, Gardnerella, Ureaplasma, and Streptococcus. The relative abundance of Lactobacillus was the highest in the first trimester (95.4%) compared to the second trimester (94.7%) and non-pregnant control group (89.2%), while it was the least abundant in the in third trimester (81.8%). However, the relative abundance of Gardnerella and Ureaplasma was the highest in the third trimester (5.9%, 4.4%, respectively) when compared to the second trimester (1.3%, 0.9%, respectively) and non-pregnant control group (2.6%, 0.9%, respectively), while their relative abundance was the least in the first trimester (0.9%, 0.4%, respectively). The relative abundance of the genus Streptococcus was the highest in the third trimester (2.1%) when compared to the first trimester (0.5%) and non-pregnant control group (0.6%), while it was the least in the second trimester (0.05%).
On the species level ( Figure 1C, Supplementary Table S4), L. iners was the most abundant species in all study groups. However, it was relatively the most abundant species in the second trimester (65.3%) and non-pregnant control group (66.6%) when compared to the first trimester (52.2%), while it was relatively the least abundant in the third trimester (33%). Other detected Lactobacillus species included L. helveticus, L. jensenii, L. gasseri, L. coleohominis, and L. reuteri. The relative abundance of L. helveticus was the highest in the third trimester (25.4%) when compared to the second trimester (18.8%) and first trimester (8.4%), while its relative abundance was the lowest in the non-pregnant control group (2.9%). The relative abundance of L. jensenii was the highest in the first trimester (23.1%) when compared to the third trimester (16.1%) and non-pregnant control group (13.4%), while its relative abundance was the least in the second trimester (9.9%). The relative abundance of L. gasseri was the highest in the first trimester (11.3%) when compared to the third trimester (6.3%) and non-pregnant control group (5.9%), while its abundance was the least in second trimester (0.17%). L. coleohominis and L. reuteri were detected as minor Lactobacillus species. The relative abundance of L. coleohominis ranged from 0.22% in the third trimester compared to 0.19% in the first trimester, while its relative abundance was the least in the second trimester and non-pregnant control group (0.09%, 0.085%, respectively). The relative abundance of L. reuteri was 0.49% the in third trimester, 0.42% in the second trimester, and 0.29% in the non-pregnant control group, while its abundance was the least in the first trimester (0.2%).
Among the non-Lactobacillus species ( Figure 1D, Supplementary Table S5), Bifidobacterium breve, Peptostreptococcus anaerobius, Streptococcus anginosus, Ureaplasma parvum, and Streptococcus agalactiae were frequently detected. The relative abundance of B. breve, P. anaerobius, and S. anginosus was the highest in the non-pregnant control group. The relative abundance of B. breve was 2.5% in the non-pregnant control group compared to 0.04% in the first trimester, while it was nearly absent in the second and third trimesters. Similarly, the relative abundance of P. anaerobius was 0.29% in the non-pregnant control group compared to 0.28% in the first trimester, while its relative abundance was the lowest in the second and third trimesters (0.02%, 0.019%, respectively). The relative abundance of S. anginosus was 0.48% in the non-pregnant control group compared to 0.12% in the third trimester, while its relative abundance was as its lowest in the second and first trimesters (0.03%, 0.01%, respectively). However, the relative abundance of U. parvum and S. agalactiae was the highest in the third trimester. The relative abundance of U. parvum was 4.4% in the third trimester compared to 0.9% in the second trimester and the non-pregnant control group, while its relative abundance was the lowest in the first trimester (0.4%). The relative abundance of S. agalactiae was 1.9% in the third trimester compared to 0.5% in the first trimester, while its relative abundance was the lowest in the non-pregnant control group (0.005%) and it was absent in the second trimester.

Bacterial Biomarkers
Bacterial biomarkers of each study group were detected through Linear discriminant analysis effect size (LEfSe) using MicrobiomeAnalyst. At species level, B. breve was detected to be significantly abundant in the non-pregnant control group compared to the pregnant group. However, across pregnancy trimesters, L. gasseri was significantly abundant in first trimester and S. agalactiae in third trimester.

Alpha Diversity
Alpha diversity, as calculated by the Alpha diversity indices Observed (p = 0.62, Figure 3A), Chao1 (p = 0.62, Figure 3B), Shannon (p = 0.19, Figure 3C), and Simpson (p = 0.18, Figure 3D), revealed non-significant differences in the vaginal community composition within study groups. This was further confirmed by Pielou's evenness alpha diversity (p = 0.22, Figure 4, Supplementary Table S7). There was a relative homogeneity of the vaginal microbiome both across the three trimesters of pregnancy and, in comparison, to the non-pregnant status. II  III  IV-B  IV-C  V  Total  III-A  III-B  IV-C0  IV-C1  IV-

Bacterial Biomarkers
Bacterial biomarkers of each study group were detected through Linear discriminant analysis effect size (LEfSe) using MicrobiomeAnalyst. At species level, B. breve was detected to be significantly abundant in the non-pregnant control group compared to the pregnant group. However, across pregnancy trimesters, L. gasseri was significantly abundant in first trimester and S. agalactiae in third trimester.

Alpha Diversity
Alpha diversity, as calculated by the Alpha diversity indices Observed (p = 0.62, Figure  3A), Chao1 (p = 0.62, Figure 3B), Shannon (p = 0.19, Figure 3C), and Simpson (p = 0.18, Figure  3D), revealed non-significant differences in the vaginal community composition within study groups. This was further confirmed by Pielou's evenness alpha diversity (p = 0.22, Figure 4, Supplementary Table S7). There was a relative homogeneity of the vaginal microbiome both across the three trimesters of pregnancy and, in comparison, to the non-pregnant status.

Beta Diversity
The dissimilarities between vaginal microbial communities were d measuring beta diversity by calculating the Bray-Curtis and Jaccard in differences in the composition of the vaginal microbial communities amo were revealed by Bray-Curtis (R = 0.059326, p < 0.037, ANOSIM) and Jacca p < 0.037, ANOSIM) dissimilarity distances ( Figure 5 panels A and B). T were further confirmed by pairwise PERMANOVA tests. There was a s ence in the composition of the vaginal microbial communities between except between the vaginal microbial communities of the second and t pregnancy as they were revealed to be relatively similar when compare mester and the non-pregnant state. However, phylogenetic dissimilarity 5 panels C and D), unweighted UniFrac (R = 0.014871, p < 0.3, ANOSIM UniFrac tests (R = 0.01908, p < 0.151, ANOSIM) revealed non-significan tween study groups confirming the low diversity of the vaginal microbio nancy.

Beta Diversity
The dissimilarities between vaginal microbial communities were detected through measuring beta diversity by calculating the Bray-Curtis and Jaccard indices. Significant differences in the composition of the vaginal microbial communities among study groups were revealed by Bray-Curtis (R = 0.059326, p < 0.037, ANOSIM) and Jaccard (R = 0.059326, p < 0.037, ANOSIM) dissimilarity distances ( Figure 5 panels A and B). These differences were further confirmed by pairwise PERMANOVA tests. There was a significant difference in the composition of the vaginal microbial communities between all study groups except between the vaginal microbial communities of the second and third trimester of pregnancy as they were revealed to be relatively similar when compared to the first trimester and the non-pregnant state. However, phylogenetic dissimilarity distances ( Figure 5 panels C and D), unweighted UniFrac (R = 0.014871, p < 0.3, ANOSIM), and weighted UniFrac tests (R = 0.01908, p < 0.151, ANOSIM) revealed non-significant differences between study groups confirming the low diversity of the vaginal microbiome during pregnancy.

Correlation of Vaginal Microbiota during Pregnancy
Correlation analysis revealed that Firmicutes was negatively correlated with almost all other phyla ( Figure 6A, Supplementary Table S8). Likewise, Lactobacilli were negatively correlated with almost all other genera ( Figure 6B, Supplementary Table S9) showing significant negative correlation with the genera Dialister and Prevotella. Similarly, L. iners was negatively correlated with almost all other species ( Figure 6C, Supplementary Table S10). There was a significant negative correlation between L. iners and other Lactobacillus species, in particular, L. helveticus. There was also a negative correlation between L. iners and other non-Lactobacillus species, demonstrating significant negative correlations with G. vaginalis, S. agalactiae, P. anaerobius, and S. anginosus versus a significant positive correlation between the non-lactobacillus species pairs G. vaginalis and S. agalactiae, G. vaginalis and P. anaerobius, S. aginosus and P. anaerobius. Further positive correlations were found between Finegoldia magna and the species S. agalactiae, P. anaerobius, and S. aginosus.

Correlation of Vaginal Microbiota during Pregnancy
Correlation analysis revealed that Firmicutes was negatively correlated with almost all other phyla ( Figure 6A, Supplementary Table S8). Likewise, Lactobacilli were negatively correlated with almost all other genera ( Figure 6B, Supplementary Table S9) showing significant negative correlation with the genera Dialister and Prevotella. Similarly, L. iners was negatively correlated with almost all other species ( Figure 6C, Supplementary Table S10). There was a significant negative correlation between L. iners and other Lactobacillus species, in particular, L. helveticus. There was also a negative correlation between L. iners and other non-Lactobacillus species, demonstrating significant negative correlations with G. vaginalis, S. agalactiae, P. anaerobius, and S. anginosus versus a significant positive correlation between the non-lactobacillus species pairs G. vaginalis and S. agalactiae, G. vaginalis and P. anaerobius, S. aginosus and P. anaerobius. Further positive correlations were found between Finegoldia magna and the species S. agalactiae, P. anaerobius, and S. aginosus.

Discussion
A deep investigation of the vaginal environment can provide new perspectives for future health through understanding the pathophysiology of pregnancy and identifying women at risk for adverse pregnancy outcomes. In the current study, we have characterized the vaginal microbiome during normal pregnancy at the three gestational trimesters compared to non-pregnant status, in Ismailia, Egypt through profiling the vaginal microbiota using 16S rRNA next-generation sequencing. We were able to describe fluctuations in the composition of the vaginal microbiota during the three trimesters of pregnancy compared to non-pregnant state.
In general, irrespective to the period, pregnancy was dominated by members of the phylum Firmicutes. As expected, Lactobacillus-dominant profiles were a characteristic feature of healthy vaginal microbiomes during pregnancy. The relative abundance of the genus Lactobacillus was in a range from 81% to 95%, confirming the low biodiversity of healthy vaginal bacterial communities. This is in agreement with other populations worldwide [6,9,10,12,[17][18][19]28,30,47] despite differences in methodologies and definitions. The growth-inhibiting ability of vaginal lactobacilli is well-documented and primarily attributed to their role in lowering the vaginal pH by lactic acid production [3,24,48]. Other reported growth-inhibiting strategies included hydrogen peroxide production and release of antimicrobial substances, such as bacteriocins [3,24,[48][49][50].
L. iners was the most abundant species in both healthy pregnant and non-pregnant women in our study. L. iners has been frequently reported as the most Lactobacillus-dominant member of the vaginal microbiome of women with African ancestry. A PRISMAcompliant review conducted by Juliana and co-workers [5] in seven sub-Saharan African countries identified L. iners among the most prevalent species. Similarly, Gautam et al., 2015 [51] displayed that L. iners-dominated vaginal microbiome clusters among the most

Discussion
A deep investigation of the vaginal environment can provide new perspectives for future health through understanding the pathophysiology of pregnancy and identifying women at risk for adverse pregnancy outcomes. In the current study, we have characterized the vaginal microbiome during normal pregnancy at the three gestational trimesters compared to non-pregnant status, in Ismailia, Egypt through profiling the vaginal microbiota using 16S rRNA next-generation sequencing. We were able to describe fluctuations in the composition of the vaginal microbiota during the three trimesters of pregnancy compared to non-pregnant state.
In general, irrespective to the period, pregnancy was dominated by members of the phylum Firmicutes. As expected, Lactobacillus-dominant profiles were a characteristic feature of healthy vaginal microbiomes during pregnancy. The relative abundance of the genus Lactobacillus was in a range from 81% to 95%, confirming the low biodiversity of healthy vaginal bacterial communities. This is in agreement with other populations worldwide [6,9,10,12,[17][18][19]28,30,47] despite differences in methodologies and definitions. The growth-inhibiting ability of vaginal lactobacilli is well-documented and primarily attributed to their role in lowering the vaginal pH by lactic acid production [3,24,48]. Other reported growth-inhibiting strategies included hydrogen peroxide production and release of antimicrobial substances, such as bacteriocins [3,24,[48][49][50].
L. iners was the most abundant species in both healthy pregnant and non-pregnant women in our study. L. iners has been frequently reported as the most Lactobacillus-dominant member of the vaginal microbiome of women with African ancestry. A PRISMA-compliant review conducted by Juliana and co-workers [5] in seven sub-Saharan African countries identified L. iners among the most prevalent species. Similarly, Gautam et al., 2015 [51] displayed that L. iners-dominated vaginal microbiome clusters among the most dominant clusters in African women from four sub-Saharan countries. Another study by Jespers et al., 2015 [52] in African women from three sub-Saharan regions demonstrated the higher relative abundance of L. iners in women with normal Nugent score compared to non-African studies. Likewise, Serrano et al., 2019 [10] showed pregnant women of African ancestry with significantly higher representation of L. iners compared to non-African ancestry.
Another species that was commonly detected as a characteristic feature of African vaginal microbiomes is Ca. Lachnocurva vaginae or BVAB1 [15,27]. This bacterium is consistently correlated with pregnancy complications, such as preterm birth [10,35]. However, Ca. Lachnocurva vaginae was not detected among our vaginal microbial communities even when compared to studies using similar approaches for investigating the vaginal microbiome profiles [53,54].
In consistence with others [8,15,38], our correlation analysis demonstrated that the phylum Firmicutes, the genus Lactobacillus, and the species L. iners were negatively correlated with almost all other vaginal taxa. According to Ng et al., 2021 [8], this emphasizes their well-documented exclusionary behavior in the vaginal ecosystem. Furthermore, we found a relative increase in the abundance of Gardnerella and Ureaplasma in the third trimester compared to the first and second trimesters, but this increase was in accordance with a high Lactobacillus abundance. Gardnerella and Ureaplasma are well-known pathogens associated with adverse pregnancy outcomes [55]. However, recent microbiome reports [7,55] point out that vaginal colonization with Ureaplasma and Gardnerella during a pregnancy with Lactobacillus abundance seems to be protective against pregnancy complications. Park et al., 2022 [55] demonstrated that Ureaplasma and Lactobacillus coexistence with high Lactobacillus abundance leads to term birth. In addition, Park et al., 2022 [55], Liu et al., 2022 [7], and DiGuilio, 2015 [16] reported that preterm birth and premature membrane ruptures are only predictive with low or depleted Lactobacillus abundance in the presence of Ureaplasma and Gardnerella. This suggests that the balanced co-colonization of Ureaplasma and Gardnerella in the vaginal ecosystem under Lactobacillus dominance may promote full-term birth and prevent pregnancy complications.
We also found L. iners was negatively correlated with L. helveticus, which was common among our vaginal samples. This could be attributable to the overlapping ecological functions in the vaginal econiche. On the other hand, we detected multiple positive correlations between bacteria associated with bacterial vaginosis or vaginal dysbiosis. We found positive correlations between the pairs of G. vaginalis and S. agalactiae, G. vaginalis and P. anaerobius, and S. aginosus and P. anaerobius. Further positive correlations were found between F. magna and S. aginosus, as well as S. agalactiae and P. anaerobius. This is in line with previous reports which highlighted the importance of the synergistic interplay between non-lactobacillus vaginal taxa leading to Lactobacillus displacement, overgrowth of anaerobic bacteria, rise in vaginal pH, and ultimately vaginal dysbiosis or bacterial vaginosis [56][57][58][59][60][61].
Microbiota profiling through 16S rRNA gene next-generation sequencing is primarily dependent on selecting suitable primer pairs and hypervariable regions of the bacterial 16S rRNA gene [62]. Primer selection may have significant influence on obtaining different results [62]. In the current study, we used universal primers, as per instructions of the standard Illumina protocol [40] spanning the V3-V4 hypervariable region, which is widely used for investigating the human vaginal microbiome [1,2,4,7,12,13,38]. This region has been reported to provide intense discrimination between most microbiota of the urogenital tract [7,13]. However, alternative hypervariable regions of the 16S rRNA were also utilized, such as the V3-V5 [16,28,30] and the V1-V2 or V1-V3 hypervariable regions [62,63]. Prince et al., 2015 [63], indicated primers spanning the V1-V3 region are more suitable for detecting Lactobacillus, Prevotella, and Clostridium species, whereas primers spanning the V3-V5 region are better when considering Enterobacteriaceae and Bifidobacteriaceae. According to Graspeuntner et al., 2018 [62], a greater number of taxa in the vaginal microbiota are detectable under the V3-V4 region compared to V1-V2, which is inappropriate for identifying G. vaginalis. The authors indicated that when considering well-defined vaginal microbiota with proven influence in both health and disease, such as G. vaginalis, Bifidobacterium bifidum, and Chlamydia trachomatis, the V3-V4 region is recommended for 16S rRNA microbiome studies. These taxa are underestimated in V1-V2 16S rRNA vaginal community-based profiles, resulting in failure to correctly assess the bacterial diversity and overestimation of the abundance of other taxa.
The selection of tools and databases for appropriate taxonomic assignation is another important methodological issue that should be considered in 16S rRNA microbiome studies. Similar to others [1,8,38], we used QIIME 2 software and Greengenes database. QIIME 2 has been demonstrated to facilitate the reproducible and comprehensive analysis of diverse microbiome data [64]. Greengenes is a popular taxonomic database, which is widely used in profiling the human vaginal microbiome [65][66][67][68][69][70]. Nonetheless, microbiota profiling by 16S rRNA next-generation sequencing is the preferred approach when compared to traditional culture. Currently, it is the quickest and most accurate means for investigating the human vaginal microbiota.
The present study has several limitations that should be avoided in the future. The small sample size of enrolled participants may not give an accurate demonstration of the composition and diversity of the vaginal microbiomes. In addition, we were unable to collect detailed clinical data that would have been valuable in determining potential correlations based on pregnancy state and associated vaginal community composition. Moreover, 16S rRNA next-generation sequencing was used for profiling vaginal bacterial communities as it allows the simultaneous detection of many taxa. However, this technique has limited taxonomic power, particularly on the species level and is not precise enough when considering taxa with a minor abundance [71][72][73]. Some authors may use species-specific quantitative real-time PCR assays to confirm the identity and relative abundance [74]. However, this approach is hampered by the limited multiplex capability of real-time PCR [75,76]. Alternative approaches, such as whole genome shotgun metagenomic sequencing (WGS), should be considered in future studies. WGS is more precise in revealing bacterial species, diversity, and abundance with added the advantage of providing direct information on functional genes [71][72][73].

Conclusions
In conclusion, we investigated the vaginal microbiome composition of healthy pregnant women in Ismailia, Egypt during the first, second, and third trimesters compared to healthy non-pregnant women. As expected, phylum Firmicutes and genus Lactobacillus dominated during the three trimesters of pregnancy. L. iners-dominant vaginal communities CST-III prevailed as L. iners was the most abundant species in both healthy pregnant and non-pregnant women in our study. L. iners was almost negatively correlated with all other species in the vagina, in particular, with G. vaginalis, S. anginosus, and S. agalactiae. In contrast, we found significant positive correlations between the species pairs G. vaginalis and S. agalactiae, P. anaerobius and F. magna, S. aginosus and F. magna, S. agalactiae and P. anaerobius. This highlights the protective role of lactobacilli in promoting vaginal health and emphasizing the potential synergistic interplay among vaginal microbiota to induce vaginal dysbiosis. However, a major limitation of the current study is the small sample size and minimal demographic data. Further studies recruiting larger cohorts with detailed demographics of participants should be considered.
Supplementary Materials: The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/microorganisms11010139/s1, Figure S1: Positional quality plots generated in QIIME2 by random subsampling 10,000 out of 5,930,186 sequences without replacement. (A) Positional quality plot for forward reads. (B) Positional quality plots for reverse reads. Figure S2: Alpha rarefaction curves of vaginal microbial communities in our study groups. The lines in this rarefaction curve indicate a sampling depth nearly about 20,000. Therefore, additional sequences above that sampling depth would not lead to the observation of new additional features. Good's coverage was 100 suggesting the full identification of the majority of vaginal bacterial communities in our samples. Figure S3: Relative taxa abundance in vaginal bacterial communities of Egyptian pregnant women across the three trimesters of pregnancy compared to non-pregnant samples. (A) Stacked bar charts representing relative abundance of genus. (B) Stacked bar charts representing relative abundance of species. Each bar represents one sample. Negative-control represents samples with empty swab for DNA extraction and amplified nuclease-free water without extracted DNA. Figure S4: Relative abundance of non-Lactobacillus species across the three trimesters of pregnancy in comparison to non-pregnant state. Negative-control represents samples with empty swab for DNA extraction and amplified nuclease-free water without extracted DNA. Table S1: Age of participants; Table S2: Relative abundance of taxa at phylum level. Table S3: Relative abundance of taxa at genus level. Table S4: Relative abundance of Lactobacillus species. Table S5: Relative abundance of non-Lactobacillus. Table S6: VALENCIA CST classification output revealed similarity of vaginal microbiota profile of individual sample to each of 13 reference centroids producing 13 similarity scores ranging from 0.0 to 1.0 for each sample. Table S7: Alpha diversity using Pielou's evenness by measuring Kruskal-Wallis test (pairwise) revealing non-significant difference. Table S8: Spearman correlation between the relative abundances of vaginal microbial communities at phylum level and their p values. Table S9: Spearman correlation between the relative abundances of vaginal microbial communities at genera level and their p values. Table S10: Spearman correlation between the relative abundances of vaginal microbial communities at species level and their p values. Software code S1: R scripts used for downstream analysis. Funding: This research received no external funding.

Data Availability Statement:
The data presented in this study are openly available in the National Center for Biotechnology Information (NCBI) Sequence Read Archive (SRA), under the BioProject ID PRJNA877217 https://www.ncbi.nlm.nih.gov/sra/PRJNA877217 (accessed on 15 September 2022) involving accession numbers from SAMN30696179 to SAMN30696228.