2.1. Plasmid Selection

The plasmids for both Prevotella and Bifidobacterium were collected from the plasmid database PLSDB [17] and from NCBI [18]. The term “Bifidobacterium” was used to search plasmids of Bifidobacterium and the term “Prevotella” was used to search plasmids of the Prevotella genus in the search browser given. 61 plasmid sequences were collected for Bifidobacterium [Table A]; they included Bifidobacterium longum DJO10A and B. longum subsp. infantis 157F, Bifidobacterium choerinum strain FMB-1 and other strains of Bifidobacterium. The size of the plasmids ranged from 0.0026 Mb to 0.198 Mb and their GC content ranged between 55% and 68%.

For Prevotella, a total of 32 plasmid sequences were collected [Table B]; they included different strains of Prevotella like Prevotella scopos JCM 17725, Prevotella nigrescens, and Prevotella copri DSM 18205 and others. The size of the plasmid ranged between 0.004 Mb and 1 Mb and their GC content ranged between 32% and 46%.

2.2. Phylogenetic Tree Construction with Plasmids

The phylonium tool was used to determine the evolutionary distances between genomes that are closely related. Compared to alignment-based methods of phylogeny reconstruction, it is faster and more accurate in comparison to other methods that are alignment-free. Using these evolutionary distances, a phylogenetic tree was constructed [19].

2.3. Screening Plasmids for AMR Genes

The collected plasmids were subjected to screening to check the prevalence of AMR genes using the ResFinderFG 2.1 web server [20]. The database contains fifteen antibiotic drug classes and all of them were used for screening. Resfinder will identify the presence of AMR genes in a given sequence by contrasting as well as locating genes that confer antibiotic resistance. They included various drug classes, including beta-lactam (BL), fluoroquinolone, macrolide (ML)-lincosamide (LS)-streptogramin B (SGB), tetracycline (TC), and more.

Bifidobacterium, is a gram-positive bacteria so drugs exclusively used to cure gram-negative bacteria do not have any significant negative effect on them, in the same way, the gram-positive drug does not have any significant negative effect on Prevotella as it is a Gram-negative bacteria. So, the gram-negative and gram-positive drug classes are used as a control respectively.

The sequences were uploaded to the web server, and the AMR gene criteria were set to detect the presence of resistance genes to all drug classes available on the server. The percentage identity was determined at 90%, and the percentage of perfect alignment was set to 100%. The nucleotide number that is similar to the gene providing the best-matching resistance in the database and the respective sequence in the plasmid were compared to determine the percentage identity. By default, a percentage above 60% in nucleotide sequence overlap with the resistant gene was considered to be a hit. The plasmid that contained the most AMR genes in the database was evaluated using a minimum identity setting of 30 and a maximum identity value of 100%.

2.4. Potential Microbial Resistance Index

Potential multiple antibiotic resistance also known as the p-MAR index was computed for the plasmids screened for resistance, from the aforementioned results. Previously published methods were used to calculator the index [21,22]. The p-MAR index is defined as the ratio between genes that show antibiotic resistance and genes that show antibiotic susceptibility.

2.5. Verification of the Acquired AMR Genes

The results obtained from ResFinderFG 2.1 database are verified using another database called KmerResistance 2.2 [23,24]. The database was set to resistant genes, while the identity threshold is set to 70%, involving a depth correction threshold at 10%. The results outputted were recorded and cross checked with the outputs given by the ResFinderFG database. KmerResistance 2.2 also shows the template sequence and query coverage of the uploaded sequence in its output.

2.6. Determining Antibiotic Resistance Determinants (ARDs)

The plasmid that confers resistance genes was collected and they were screened to find the presence of ARDs using another software/database called ResfinderFG 1.0 [20].Using functional metagenomic ARDs, ResFinderFG identifies resistant phenotypes through the usage of a functional genomic database. Default parameters were used. Identity percentage was set to 98% and the minimum query length set to 60%. “Assembled contigs/genomes” was the read type preferred, and these sequences were cross checked for the 13 determinant families in the database [25,26].

2.7. Statistical Analysis

Using XLSTAT [27], a principal component analysis (PCA) was completed to show the relationships between the size of the plasmid and the GC% content, and the mean, standard deviation and Pearson correlation coefficient were calculated and recorded [28].

3.1. Acquired AMR Genes

To determine the prevalence of AMR genes in both Bifidobacterium and Prevotella in the ResFinder database. ResFinder finds the presence of AMR genes in the genome using the available known genes procured in the database. The query sequence uploaded by the user is aligned or compared using an alignment program called BLAST. The output received is recorded for further use.

Totally 61 plasmid sequences of Bifidobacterium were collected and screened for anti-microbial drug resistance. Out of which 9 plasmids confer resistance genes to the selected drug classes [Table C]. Bifidobacterium strains resisted drug classes like SGB, LS, ML and TC. The genes responsible for the resistance are erm(X), tet(W), and lnu(C). Among them, the most frquent gene was the erm(X), and is seen 4 times out of 10 while the tet(W) resistance gene, was seen 5 times on 10 observations. The erythromycin ribosome methylase (erm) resistance gene is an rRNA methyltransferase that projects the ribosome from inactivation due to antibiotic binding and is widely found in Bifidobacterium species and Corynebacterium species and is responsible for antibiotic resistance for SGB, LS and ML [29,30]. Traditional MLs such as erythromycin and azithromycin have limited clinical utility due to the increased spread and broad antibiotic resistance spectrum of erm [31]. The tet(W) gene codes for a protection protein that binds to the ribosome and modifies the ribosomal conformation that prevents TC from binding, enabling protein synthesis to continue. Of all TC resistance gene classes, this is the most common [32].

For Prevotella, 32 plasmids were collected and screened for AMR genes, out of which 5 genes confer resistance genes [Table D]. Strains of Prevotella were resistant to two main drug classes: BL and TC. The genes responsible for the resistance are cfxA6, cfxA3, cfxA4, cfxA5 and tet(Q). The cfxA6, cfxA3, cfxA4, and cfxA5 belong to class A beta-lactamase, where they catalyse the opening and hydrolysis of the BL ring. Some Prevotella strains exhibit resistance to penicillin by expressing the beta-lactamase genes, and they may have gained resistance to this drug class as penicillin is a systemic antibiotic prescribed to patients with dentoalveolar infections, where it involves anaerobic bacteria residing in the oral cavity like Prevotella and Fusobacterium [33,34]. Tet(Q) is a TC-resistant determinant that confers resistance by a ribosome protection mechanism.

3.2. Verification of the Acquired AMR genes

The data collected from the ResFinder software is once again verified using the database called KmerResistance 2.2. The output of KmerResistance 2.2 gives a detailed view of template coverage, Query coverage (query coverage is the length of the matching query sequence divided by the template length), Q_value (Q_value tests whether the current template is a significant hit) and P_value. The obtained outputs, shown in Tables 1 and 2, are compared with the output from ResFinder and no significant difference was noted. The verified results were noted, and they were used in the next step.

Table 1: Resistance to different classes of antimicrobial drugs and the p-MAR index of the 9 Bifidobacterium plasmids listed showing the 4 drug classes.

Table 2: Resistance to different classes of antimicrobial drugs and the p-MAR index of the 5 Prevotella plasmids listed showing the 2 drug classes.

3.3. Determining ARDs

To determine the presence of ARDs, the plasmids with the AMR genes were screened in the database ResFinderFG 1.0. ResFinderFG 1.0 contains 13 ARD families in its database they are 16S_rRNA_methyltransferase, aminoglycosides (AGs) acetyl-transferases, AGs nucleotidyltransferase, AGs phospho-transferases, beta-lactamases (all sub-families), chloramphenicol acetyl-transferases, dihydrofolate reductase, efflux pumps, quinolone resistance, spanning tet(A) and variants, spanning tet(M) and variants, TC monooxygenases, homologues of D-Ala-D-x ligases (selected on D-cycloserine).

Three ARD families were detected in Bifidobacterium. The most occurring ARD was dihydrofolate reductase (dfr) followed by spanning tet(M) and variants (tet_protection).

Five ARD families were detected in Prevotella. They are beta-lactamase (all sub-families), dihydrofolate reductase (dfr), homologues of D-Ala-D-x ligases (van_ligase) followed by spanning tet(M) and variants (tet_protection) and chloramphenicol acetyl-transferases (cat).

3.4. Size and GC content of the Bifidobacterium and Preovetlla Plasmids

To establish a correlation between the size and the GC content of the plasmid, PCA was carried out on the 9 plasmids of Bifidobacterium and 5 plasmids of Prevotella that contained the AMR genes. The following summarises the sizes of the plasmids and the GC content, where Bifidobacterium plasmids had a lowest of 0.003 and highest of 0.198 in size (MB) and the mean being 0.067 ± 0.091, and the lowest GC content was roughly 56% and highest 66%, and a mean of 60% ± 3.98. The Prevotella plasmid’s lowest size (MB) was 0.006 and highest of 0.167 in size (MB) and the mean being 0.061 ± 0.076, and the lowest GC content was approximately 37% and a highest of 42%, with a mean of 40% ± 1.96.

For Bifidobacterium, the negative Pearson correlation (r = −0.68, α = 0.95) indicates no linear correlation between the GC content and the plasmid size [Figure 1].

Figure 1: A principal component analysis (PCA) biplot of variable size (MB) and GC (%) content for 9 plasmids of Bifidobacterium (labelled in blue). Factor 1 and Factor 2 of the PCA show their percentage contribution in brackets. [Click here to view]

For Prevotella, the positive correlation (r = 0.24, α = 0.95) indicates that there is a linear relationship between the GC content and the plasmid size [Figure 2].

Figure 2: A principal component analysis (PCA) biplot of variable size (MB) and GC (%) content for five plasmids of Prevotella (labelled in blue). Factor 1 and Factor 2 of the PCA show their percentage contribution in brackets. [Click here to view]

Recent advancements in the study of GC content and analysis of newer plasmids are important, as they are involved in horizontal gene transfer, which can give more information on natural evolution and the increase of AMR genes [35]. The GC content gives extensive data regarding the evolutionary information and the genomic size of bacteria [36-38]. The range of GC content seen here falls in the average range reported for species, which is 13–75% [39], but higher than the average range reported for bacteria, which is 50.76% [40]. It may be inferred that plasmids with a greater GC content are introduced earlier than plasmids with a relatively lower GC content in comparison, which may be due to environmental differences or the phylogenetic composition [41].

3.5. Statistical Analysis of Sequences

Using phylonium, the sequences were analysed, and the pairwise distances were calculated. Using these distances, a phylogenetic tree was constructed, and standard error estimates were calculated by toggling to the bootstrap method (1000 replicates). Figures 3 and 4 showcase the phylogenetic trees for the Bifidobacterium and Prevotella species [42].

Figure 3: The phylogeny of 9 Bifidobacterium genomes; This phylogeny was computed using phylonium with bootstrap method (1000 replicates), neighbour-joining and then visualised with ETE [30]. [Click here to view]

Figure 4: The phylogeny of 5 Prevotella genomes; This phylogeny was computed using phylonium with bootstrap method (1000 replicates), neighbour-joining and then visualised with ETE. [Click here to view]

Various studies have been conducted to prove the presence of AMR genes in gut microbiomes. This study mainly focused on the presence of AMR in Bifidobacterium and Prevotella, and there have also been various studies that prove the presence of antimicrobial genes of other gut microorganisms such as Lactobacillus, Escherichia coli, and Bacteroides fragilis.

Lactobacillus is a genus of bacteria that is commonly found in the human gut and is known for its probiotic properties. Studies have proved the presence of AMR genes in Lactobacillus against antibiotics like TC, erythromycin, and clindamycin [43].

B. fragilis is a common anaerobic bacterium found in gut microbiota. Isolates of B. fragilis from clinical samples contain cfiA(C) genes resistant to carbapenem antibiotics [44].

Advancements in genomic sequencing technology have allowed for the identification and characterization of the antimicrobial genes. Lactobacillus reuteri forms antimicrobial substances such as reuterin and 3-hydroxypropionaldehyde [45]. Anti-microbial genes in E. coli are found to be involved in the synthesis of colicins and other antimicrobial peptides [46].

Understanding the mechanism of antimicrobial genes of different gut microbes can help researchers develop new strategies for promoting the growth of beneficial bacteria and preventing the spread of pathogenic bacteria in the gut.