Application of metagenomic analysis to the detection of EsβL- and carbapenemase-producing enteric pathogens recovered from different hosts

The project's main objective was to understand the spread of Escherichia coli and Klebsiella pneumoniae, carriers of encoding resistance genes of extended-spectrum β-lactamases (ESβL) and carbapenemases in fecal samples from different hosts in Brazil´s five major geographic regions. Metagenomic sequencing of the samples revealed a total of 21,029 distinct bacterial species, with 12,388 in humans and 16,779 in poultry. The bacteriome consisted mainly of species belonging to phyla Firmicutes, Proteobacteria, Bacteroidetes and Actinobacteria. Resistome analysis detected a total of 405 antimicrobial resistance genes (ARG) that encode resistance to 12 different classes of antimicrobials. Genes encoding antimicrobial-modifying enzymes were the most frequent. Interestingly, the study showed acquired carbapenemase-encoding genes such as blaAIM, blaCAM, blaHMB, and blaGIM that had not been reported previously in Latin America. The presence of blaKPC-2 and blaNDM-1 genes was noted in the bacterial culture of the fecal samples included in the study using selective pressure to detect ESβL and carbapenemases. These genes were not detected in the metagenomic analysis. Finally, Monte Carlo simulation showed that humans, not cattle, are the main reservoir of blaTEM-like, blaSHV-like, and blaCTX-M-like

Fecal samples were collected in three rural properties in Brazil's five major geographic regions. Five centers cooperated with this project: School of Biological and Environmental Sciences of the Federal University of Grande Dourados (UFGD) (Midwest Region), Regional University of Blumenau (FURB) (South Region), Evandro Chagas Institute (IEC) (North Region), Federal University of Ceara (UFC) (Northeast Region), and São Francisco University (USF) (Southeast Region). 

Fecal samples were collected using Amies with charcoal transport swabs and transferred to the Alert Laboratory of the Federal University of São Paulo (UNIFESP), where total DNA extraction and bacterial culture were performed. The extracted DNA was then forwarded to the National Laboratory for Computational Science (LNCC) and metagenomic sequencing was performed with the NextSeq 500 system using NextSeq 500/550 high-output v2.5 (300 cycles) kit (Illumina, USA), with the system set to produce 2 × 150 bp readings. Subsequently, bacteriome and resistome composition analyses were conducted based on the total operational taxonomic units (OTUs) of selected species (Salmonella spp., Escherichia spp., Enterobacter spp., and Klebsiella spp.).

The study thus generated the Relative Abundance Index (RAI) of resistance markers potentially related to the production of ESβLs. For each host group (humans, swine, cattle, and poultry), the relative abundance of ESβL-encoding genes in the total OTUs found from bacterial index species was used to apply stochastic simulation by Markov chains (Monte Carlo).

Microbial compositions were determined in 107 samples collected from poultry (n = 30), cattle (n = 30), swine (n = 15), and humans (n = 32) from all five Brazilian geographic regions. 1). The average number of readings per sample ranged from 1,474,376 to 49,909,522, totaling 1.62 billion bp. The mean number of N50 contigs was 4,160, and there were 112,267 gene coding sequences (CDS). About 5.4 million reads had taxonomic signatures down to the species level.

21,029 unique bacterial species were identified, ranging from 12,388 in humans to 16,779 in poultry. The bacteriome was composed mainly of species belonging to phyla Firmicutes (41%), Proteobacteria (29.4%), Bacteroidetes (14.8%), and Actinobacteria (8.7%). When the diversity and richness of human bacterial composition were compared to those of animals, the Chao1 richness estimator showed that the number of species was significantly higher in cattle, poultry, and swine (P ≤ 0.05). Based on the Shannon index, the estimated diversity was similar across the different hosts, except for swine from the South of Brazil and cattle from the Midwest, which showed greater diversity (P ≤ 0.05). Furthermore, the bacterial composition varied in cattle between South and Northeast Brazil. The observed diversity was not affected by the dominance of some species, as estimated by the Simpson index, which showed high similarity for all hosts, with no statistically significant difference.

Resistome analysis detected a total of 405 ARGs in 12 antimicrobial classes. The genes that encode antimicrobial-modifying enzymes (n = 231; 57%) were the most frequent, followed by genes related to target site alteration (n = 95; 23.5%) and efflux systems (n = 79; 19.3%). Among genes that encode antimicrobial-modifying enzymes, those encoding β-lactamases and aminoglycoside-modifying enzymes (AMEs) were the most frequent. The genes encoding intrinsic and acquired β-lactamases were by far the most frequent and diverse group of enzymes found (n = 122; 52.8%). Following Ambler’s classification, 42 genes encoding β-lactamases belonged to molecular class C (n = 42; 34.4%), compared to class A (n = 39; 32.0%), class D (n = 25; 20.5%), and class B (n = 16; 13.1%). The analysis also showed the presence of acquired carbapenemase-encoding genes such as blaAIM, blaCAM, blaHMB, and blaGIM, which had not been reported previously in Latin America. The occurrence of blaAIM was observed in cattle from the South of Brazil, in poultry from the Southeast, and in cattle and poultry from the Northeast. Thus, blaSME were found in cattle from the South and Midwest and in humans from the North. Some of these ARGs were identified in specific locations; for example, blaGIM and blaCAM were recovered from different hosts in the Southeast and blaHMB and blaVEB in the Midwest and Northeast, respectively. In addition, a total of 59 distinct AME coding genes were observed, among which acetyltransferases (AACs) were the most frequent enzymes (n = 27 variants), including aac(6′)-Ib-cr, followed by adenyl-transferases (ANTs) (n = 18 variants) and phosphotransferases (APHs) (n = 14 variants).

Resistome data obtained from metagenomic sequencing were used to perform the Monte Carlo simulation to understand the occurrence of the blaTEM-like, blaSHV-like, and blaCTX-M-like genes in humans, cattle, swine, and poultry in Brazil.  Monte Carlo simulation was chosen as a tool to allow predicting the possible distribution of certain genes in the population starting from a smaller sample group. The data generated revealed with 97.80% confidence that, in humans, the probability of relative abundance of blaTEM-like, blaSHV-like, and blaCTX-M-like can vary from 0.006% to 1.179%. The simulation also indicated that the relative abundance of these ARGs in poultry ranged from 0.001% to 0.349%, with 96.9% confidence. In cattle, the relative abundance of selected genes ranged from 0.00% to 0.155% (96.97% confidence). Finally, estimated abundance in swine ranged from 0.001% to 0.703%. The number of total species was significantly higher in cattle, poultry, and swine than in humans. However, we can conclude that humans must be the main reservoir of blaTEM-like, blaSHV-like, blaCTX-M-like, while cattle have the lowest rate for the selected genes. These data indicate that the greater the diversity of species in a host, the lower the relative abundance of blaTEM-like, blaSHV-like, and blaCTX-M-like resistance genes. 480 bacteria were recovered from swab cultures, with Escherichia (22.5%) and Pseudomonas (13%) as the most frequent genera. The study also showed that ESβL-producing strains in hosts were most common in humans and poultry. Our findings showed that seven hosts were colonized by carbapenemase-producing strains, of which KPC-producing strains were the most frequent, present in five hosts.

Several Brazilian studies have reported phenotypic and genotypic frequencies in multidrug-resistant bacteria in different ecological niches. However, no previous study had used metagenomic analysis to assess the microbiome and resistome of human and farm animal samples collected in Brazil´s five geographic regions during the same period.

Currently, the antimicrobial resistance phenomenon is not limited to the hospital environment. The colonization and/or infection of individuals by resistant bacteria in the community has expanded steadily in recent years, thus showing that indiscriminate use of antimicrobials has a global impact. A key attribute of an integrated surveillance program for antimicrobial resistance in bacteria transmitted through the food chain is the ability to monitor and detect the emergence and spread of resistant bacteria in animal products and their correlation with the composition of the human microbiota. The growing global trade in farm animals and their derived products underscores the importance of sharing global data on pathogens and foodborne diseases, including data on antimicrobial resistance. The information produced in this project can serve as a model for the future implementation of a national program for integrated surveillance of antimicrobial resistance based on One Health.

We intend to expand the number of samples analyzed in Brazil, drawing on a partnership with the Brazilian Agricultural Research Corporation (EMBRAPA) and the Ministry of Agriculture, Cattle, and Supply (MAPA) to select the cattle, swine, and poultry in farm properties.