Background
The antimicrobial resistance (AMR) phenomenon has spread rapidly over the course of the past decades to establish itself as a major global public health threat [
1] in spite of the strides made by modern medicine to apply the use of antibiotics to ensure safe surgical procedures and improve the quality of medical care that have greatly reduced morbidity and mortality in public health [
2,
3]. The bacterial infections that have been able to exhibit AMR have become fatal and given birth to a possible post-antibiotic era which initially was thought to be an apocalyptic fantasy before the 21st century [
2,
4]. It goes without saying that the role played by humans in exacerbating the rate at which these infectious agents have developed resistance to antibiotics has been at the forefront of accelerating this phenomenon mainly through the rampant and inappropriate use of antibiotics [
4,
5]. A previous study predicted an alarming 10 million deaths per annum with $100 trillion dollars’ worth of efforts trying to combat AMR by 2050 if this is not tackled [
6]. Some regions in the world especially the African continent characterized by high infectious disease burdens and limited healthcare infrastructure have the least accurate and reliable statistical data on the epidemiology and impact of AMR on the public health sectors [
7‐
9]. The rapid recent evolution of genomics-based technologies applied in the diagnosis and surveillance of the epidemiology of drug-resistant bacteria has led to the generation of large amounts of genomic data that have given deeper insights into the nature and changes of AMR determinants using modern bioinformatics analysis pipelines [
10]. Application of next-generation sequencing technologies (NGS) alongside conventional microbiology procedures and antimicrobial susceptibility testing (AST) may be the key to understanding AMR [
11], accelerating knowledge generation, and deploying interventions tailored towards optimization of antimicrobial use in public health [
12].
Therefore, this study explored and sought to understand the epidemiology, and factors driving AMR in hospital settings in Uganda and Tanzania through the use of whole-genome sequencing (WGS) data combined with the socio-demographic metadata provided by the mother study.
Materials and methods
Study design and settings
This study utilized a laboratory-based and longitudinal study design approach. The laboratory-based design was used to undertake WGS to determine the AMR elements from the bacterial isolates.
Study sites
This study was carried out at the orthopedic units of Mulago National Referral Hospital, Kampala, Uganda with geographical coordinates (0°20’16.0"N, 32°34’32.0"E) and Bugando Medical Centre (BMC), Mwanza, Tanzania with geographical coordinates (2°31’41.0"S, 32°54’27.0"E).
Study population
The study population constituted of whole-genome sequence data obtained from a total of 142 multi-drug resistant
E. coli bacterial isolates provided by the mother study titled, “Understanding Transmission Dynamics and Acquisition of Antimicrobial Resistance at Referral Hospitals and Community Settings in East Africa using Conventional Microbiology and Whole-Genome Sequencing”. The multi-drug isolates were collected from both study sites in Uganda (n = 57) and Tanzania (n = 85) from patients, the immediate non-medical caretakers of these patients, the immediate health workers attending to these patients and the patients’ environment as previously discussed in detail by the recent publications from Uganda [
1] and Tanzania [
13].
Data collection and analysis tools
The WGS data used in this study was provided by a bigger mother study titled, “Understanding Transmission Dynamics and Acquisition of Antimicrobial Resistance at Referral Hospitals and Community settings in East Africa using Conventional Microbiology and Whole-Genome Sequencing (Grant number GCA/AMR/rnd2/058)”. The bacterial isolates were shipped and sequenced by the Earlham Institute, Norwich, located in the United Kingdom following the Low Input, Transposase Enabled (LITE) Illumina protocol using the Illumina NovaSeq 6000 System.
The analysis of WGS data was done using our previously published Linux command line-based bioinformatics workflow called “rMAP”, the Rapid Microbial Analysis pipeline [
10]. Briefly, the whole-genome raw sequences together with the GenBank
Escherichia coli str. K-12 substr. MG1655 with Accession NC_000913 reference was fed into the rMAP pipeline. Sequences in the format fastq.gz were used as the input for the pipeline. All sequences were inspected for quality in the rMAP pipeline [
10] before any subsequent processes using the embedded FastQC [
14] to generate individual sample reports and MultiQC [
15] for aggregating all the multiple reports into one report. Adapters were trimmed off the sequences using Trimmomatic [
16] with the selected parameters including minimum length and phred score set to 200 and 32 respectively.
The trimmed reads were loaded into the Shovill [
17]
de-novo pipeline using the Skesa as the assembler of choice.
K-mer sizes of 31, 55, 79, 103 and 127 were used to determine the optimum genome assembly. Pilon was used for checking assembly errors, correcting ambiguous gaps, insertions, deletions and finally polishing the genomes [
18]. Genome annotations were performed using the Prokka tool [
19].
Single nucleotide polymorphism (SNP) variant calling was performed using SAMtools, Burrows-Wheeler Aligner (BWA), SAMclip, Freebayes and SnpEff [
20‐
22]. The trimmed reads were aligned against an indexed reference in fasta format (GenBank reference
Escherichia coli strain K-12 sub strain MG1655 Accession: NC_000913.3) using Burrows-Wheeler aligner [
21] to produce Sequence Alignment Map (SAM) files. Soft and hard clipped alignments were removed from the SAM files using SAMclip(
https://github.com/tseemann/samclip). SAMtools [
20] then sorted, marked duplicates and indexed the resultant Binary Alignment Map (BAM) files. Freebayes [
23]was used to call variants using Bayesian models to produce variant call format (VCF) files containing SNP information which was filtered using BCFtools(
https://github.com/samtools/bcftools) and normalized of biallelic regions using Vt [
24]. The filtered VCF files were annotated using snpEff [
22]. Missense variants that were associated with resistance were identified from the VCF files according to their respective sites. Only true SNPs were considered for the downstream analysis; insertions, deletions, and complex SNPs were filtered out from the resistance-associated SNPs.
Phylogenetic inference by maximum likelihood was performed using MAFFT, IQtree, Vcf2phylip, and BMGE [
25‐
28]. The rMAP pipeline collated all the individual VCF files into a single VCF containing all the samples and their SNPs before being transposed by vcf2phylip [
26] into a multi-alignment fasta file. MAFFT software package was used to perform multiple sequence alignment [
27]; removal of ambiguously aligned reads as well as extraction of informative sites was performed to infer phylogeny using BMGE [
25]. IQtree [
28]was then used to test various substitution models and construct trees from the alignments using the maximum-likelihood method in 1,000 bootstraps. The resulting trees were visualized in the form of rectangular phylograms.
Mass screening for AMR genes against CARD [
29], ARG-ANNOT [
30], NCBI, ResFinder, and MEGARES [
31] databases was performed for each of the study isolates using the Abricate tool (
https://github.com/tseemann/abricate). For consistency purposes, we compared results from the two most commonly used well-annotated AMR databases across the
E. coli isolates (CARD and ResFinder) from both study sites. From our findings, we found more AMR genes conforming to > 90% cut-off for both coverage and identity being detected from the ResFinder database which were then presented in form frequencies and heatmaps in the
results section.
Multi-locus sequence typing was performed using MLST (
https://github.com/tseemann/mlst) from the
E. coli assembled contigs. The
E. coli isolates were determined against a set of seven (7)
E. coli housekeeping genes (
adk4,
fum26,
gyrB2,
icd25,
mdh5,
purA5, and
recA19).
Virulence factors were determined against the virulence factor database (VFDB) [
32] using the Abricate tool (
https://github.com/tseemann/abricate) to identify elements that conformed to > 90% cut-off for both coverage and identity.
Discussion
Control of multi-drug resistant infections is fundamental in reducing the disease burden and costs incurred while treating these pathogens in tandem with the Global Action Plan set by WHO [
33] to combat AMR. This study comes in at the right point in time where the scale at which global public health is threatened by the increasing infection rates. In this study, the aim was to explore the genetic determinants that confer AMR from isolates obtained at Mulago National Referral Hospital, Bugando Medical Centre, and their environmental settings. The findings from this study are provocative and inform the dire need to strengthen the existing infection-prevention controls (IPC) together with surveillance and monitoring systems.
This study was predominantly comprised of ESBL organisms; with 36/57 (63.1%) originating from Uganda and 67/85 (78.8%) originating from Tanzania. Previous findings from studies in Uganda reported ESBL
E. coli prevalence rates of 5.3% conducted between 2006 and 2007 [
34], 62% carried out in 2014 [
35] at Mulago National Referral Hospital, and between 2015 and 2016 at Kasese Regional Referral Hospital at 62% [
36]. Related meta-analysis and systematic review studies from East Africa carried out in hospitals and surrounding communities reported a similar predominance of ESBL-producing
Escherichia coli and
Klebsiella pneumoniae [
37‐
39]. The Enterobacteriaceae family has been reported to shape the nosocomial pathogen eco-system because of the plasticity of their genome and their ability to perform inter-species and intra-species incorporation and transfer of drug resistance mediating determinants like plasmids, transposons, insertion sequences, and virulence factors via horizontal gene transfer [
40‐
42].
Detection of SNP-associated mutations in the genes;
RarD,
yaaA, and
ybgl conferring resistances to chloramphenicol, peroxidase, and quinolones from the isolates from both Uganda and Tanzania further depicts how these organisms evolve resistance towards some of the most commonly used antibiotics and antiseptic used for the day-to-day management of clinical cases. A related study highlighted the roles played by these SNPs in the evolution of antimicrobial resistance and in shaping the genome of these organisms [
43].
Our results reported very high frequencies for
blaCTX−M−15 accounting for 11/18(61.1%), and
blaCTX−M−27 with 12/23 (52.1%),
blaTEM−1B with 13/23 (56.5%) of isolates originating from Uganda and Tanzania respectively. These genes are responsible for conferring resistance to penicillin, fluroquinolones, and third-generation cephalosporins (ceftazidime and cefotaxime) which are part of routinely prescribed antibiotics used in the treatment of medical cases within the two sites similar to other previous related studies that reported phenotypic AMR profiles of the same organisms [
44]. Tanzania had relatively higher tetracycline resistance gene
tet(A) with 12/23 (52.1%) as compared to the Ugandan isolates with
tet(A) with 7/18 (38.8%) prevalence which are in agreement with similar studies performed across six Tanzanian hospitals [
45]. We also reported relatively a high prevalence of sulphonamide-resistance conferring genes
sul1 8/18(44.4%) and
sul2 15/23 (65.2%) from Uganda and Tanzania respectively. Chloramphenicol resistance gene
mdfA(A) with 21/23 (91.3%) from Tanzanian isolates and trimethoprim resistance-conferring gene
dfrA17 with 8/23 (34.7%) in Ugandan isolates were also observed within the two cohorts. The authors propose that the high resistance observed in the majority of over-the-counter antibiotics can likely be explained by their affordability in the two countries. In contrast, expensive drugs such as piperacillin-tazobactam, amikacin, and carbapenems are less accessible in most drug shops within the two study sites. Consequently, these expensive drugs are less likely to be misused, leading to lower detected resistance levels. These findings sound a very big alarm about the potential dangers such pathogens can cause to the general public health and call for the need to scale up the microbiology laboratory capacity as a way of guiding antimicrobial agent prescription. Data from well-established laboratory facilities will shape and strengthen AMR surveillance, IPC protocols within community-based settings and healthcare facilities, and regulation of drug prescriptions from drug outlets and pharmacies.
The largest portion of sequence types isolated from Tanzania belonged to the ST131 accounting for 5/59 (8.4%) of the total isolates while Uganda was represented with 1/34 (2.9%) for ST131 of the total isolates. These sequence types associated with extra-intestinal infections have been reported to be rapidly spreading as high-risk clones in Europe and worldwide due to increased AMR [
46‐
49]. Some other sequence types like ST206 have been reported to be associated with colistin-resistance conferring isolates from a study in China [
50]. These findings reiterate the dangers these organisms impose on the public health system and call for immediate interventions [
51].
The presence of virulence factors like
Shigella dysenteriae Sd197 (
gspD, gspE, gspF, gspG, and
gspF),
Yersinia pestis CO92 (
irp1, ybtU, ybtX, and
iucA),
Salmonella enterica subsp. enterica serovar Typhimurium str. LT2 (csgF and csgG),
Shigella dysenteriae Sd197 (gspC, gspD, gspE, gspF, gspG, gspH, and gspI), and
Pseudomonas aeruginosa PAO1 (
flhA, fliG, and
fliM) in isolates from both Tanzania and Uganda reported by this study depict a classic case of inter-species genetic-determinant element transfer. Multiple studies have reported how the horizontal gene transfer, a process through which genetic information can be acquired from the environment to a bacterium or from one bacterium to another through other mechanisms other than chromosomal inheritance consequently shaping pathogen virulence evolution [
52‐
56]. This study provides strong evidence regarding the acquisition of a set of rather queer virulence factors like
Yersinia pestis CO92 among the
Escherichia coli isolates originating from a somewhat deadly plague-causing bacteria similar to what has been reported by a study in India [
57].
The ingenious advent of bioinformatics platforms like rMAP [
10] used to profile all these virulence elements within the isolates in one go provides a comprehensive way of analyzing WGS data because of its easy installation, usage, and applicability, especially in low-income settings where high-performance computing infrastructure is limited. In our opinion, it also bridges and fills the missing link between the rapidly embraced field of WGS and conventional microbiology while providing high-resolution, and shorter result-generating turnaround times for the genomes of MDR pathogens. It is on these grounds that we recommend this tool to be adopted as a continuous monitoring and surveillance software for monitoring the antimicrobial resistance gene trends, plasmids, virulence factors, and MLSTs within community and healthcare settings for Uganda, Tanzania, and Africa as a whole.
Acknowledgements
Ivan Sserwadda is a fellow of the Makerere University/Uganda Virus Research Institute (UVRI) Centre of Excellence for Infection & Immunity Research and Training (MUII) program. He is also supported by MUII. MUII is supported through the DELTAS Africa Initiative (Grant no. 107743). The DELTAS Africa Initiative is an independent funding scheme of the African Academy of Sciences (AAS), Alliance for Accelerating Excellence in Science in Africa (AESA) and supported by the New Partnership for Africa’s Development Planning and Coordinating Agency (NEPAD Agency) with funding from the Wellcome Trust (Grant no. 107743) and the UK Government.
We are equally grateful for the support of the NIH Common Fund, through the OD/Office of Strategic Coordination (OSC) and the Fogarty International Center (FIC), NIH award number U2RTW010672. Its contents are solely the responsibility of the authors and do not necessarily represent the official views of the supporting offices”.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.