Pan-Resistome Characterization of Uropathogenic Escherichia coli and Klebsiella pneumoniae Strains Circulating in Uganda and Kenya, Isolated from 2017–2018

Urinary tract infection (UTI) develops after a pathogen adheres to the inner lining of the urinary tract. Cases of UTIs are predominantly caused by several Gram-negative bacteria and account for high morbidity in the clinical and community settings. Of greater concern are the strains carrying antimicrobial resistance (AMR)-conferring genes. The gravity of a UTI is also determined by a spectrum of other virulence factors. This study represents a pilot project to investigate the burden of AMR among uropathogens in East Africa. We examined bacterial samples isolated in 2017–2018 from in- and out-patients in Kenya (KY) and Uganda (UG) that presented with clinical symptoms of UTI. We reconstructed the evolutionary history of the strains, investigated their population structure, and performed comparative analysis their pangenome contents. We found 55 Escherichia coli and 19 Klebsiella pneumoniae strains confirmed uropathogenic following screening for the prevalence of UTI virulence genes including fimH, iutA, feoA/B/C, mrkD, and foc. We identified 18 different sequence types in E. coli population while all K. pneumoniae strains belong to ST11. The most prevalent E. coli sequence types were ST131 (26%), ST335/1193 (10%), and ST10 (6%). Diverse plasmid types were observed in both collections such as Incompatibility (IncF/IncH/IncQ1/IncX4) and Col groups. Pangenome analysis of each set revealed a total of 2862 and 3464 genes comprised the core genome of E. coli and K. pneumoniae population, respectively. Among these are acquired AMR determinants including fluoroquinolone resistance-conferring genes aac(3)-Ib-cr and other significant genes: aad, tet, sul1, sul2, and cat, which are associated with aminoglycoside, tetracycline, sulfonamide, and chloramphenicol resistance, respectively. Accessory genomes of both species collections were detected several β-lactamase genes, blaCTX-M, blaTEM and blaOXA, or blaNDM. Overall, 93% are multi-drug resistant in the E. coli collection while 100% of the K. pneumoniae strains contained genes that are associated with resistance to three or more antibiotic classes. Our findings illustrate the abundant acquired resistome and virulome repertoire in uropathogenic E. coli and K. pneumoniae, which are mainly disseminated via clonal and horizontal transfer, circulating in the East African region. We further demonstrate here that routine genomic surveillance is necessary for high-resolution bacterial epidemiology of these important AMR pathogens.


Introduction
Antimicrobial resistance (AMR) has raised alarms as a global health threat. AMR is often fueled by misuse and abuse of antibiotics including self-medication [1,2] and unrestricted access to antimicrobial drugs [3][4][5], and is further accelerated by industrialization, poor waste disposal, and poor hygiene levels. AMR pathogens are frequently detected in food, clinical, and environmental settings in East Africa. Despite facing broad challenges, significant efforts have recently been put in place to curb AMR in East African countries. For instance, Kenya (KY) has adapted the National Action Plan that incorporates One Health measures to prevent AMR and is highly supported by multiple governmental policies (NAPCAR 2017) [6]. Similarly, an extensive evaluation of the AMR situation in Uganda (UG) was assessed by the Uganda National Academy of Sciences (UNAS) supported by the Global Antibiotic Resistance Partnership (GARP)-Uganda (UNAS 2015) [7]. High prevalence of multi-drug resistant bacteria particularly extended-spectrum beta-lactamase (ESBL)-producing strains is significantly recorded in both countries.
Urinary tract infection (UTI) develops after a pathogen's adherence to the inner lining of the urinary tract. UTIs occur among patients of all age groups and account for high morbidity in the clinical and community settings [8]. Following binding within the urinary tract, uropathogens either cause asymptomatic or commensal connection or severe disease. About 1% of the population have asymptomatic bacteriuria (ABU), wherein a pathogen (≥10 5 cfu mL −1 ) inhabits the tract without eliciting mucosal host response [9,10]. Infections in the lower urinary tract region (e.g., cystitis) are recognized by symptoms such as dysuria. Successful virulent strains can induce pyelonephritis where rapid immune response is mobilized via cytokine secretion and influx of immune cells. UTIs are either uncomplicated or complicated. Uncomplicated UTI cases are usually observed in patients who are otherwise healthy, while complicated UTIs are diagnosed in compromised patients (e.g., if they have anatomical or functional anomalies in their urinary tract or are under longterm catheterization) [11] Treatment of these complicated UTI cases is often confounded by AMR uropathogens usually caused by Gram-negative bacteria [12]. Uncomplicated UTIs are frequently caused by uropathogenic Escherichia coli (E. coli (UPEC)) while complicated cases might be caused by several pathogens such as Proteus mirabilis, Providencia stuartii, Morganella morganii, Klebsiella pneumoniae (K. pneumoniae), and Pseudomonas aeruginosa [8]. Recurrent UTI cases are also common, particularly when urinary tract anomalies linger, or treatment failed to kill resistant bacteria [13], leading to more severe type of infections. Due to the lack of active investigation of UTI cases in East Africa, particularly in the community, access to accurate data can be challenging.
An increasing number of studies have employed whole genome sequencing (WGS) and analyses for disease surveillance in both hospital and community settings [14][15][16]. The high-resolution genotyping that WGS provides allows one to investigate and describe the population structure and evolutionary history of the isolates, as well as tracing their spread. Outbreaks have been robustly detected and described using high-throughput methodologies designed for bacterial pathogens [17][18][19][20]. Comprehensive AMR gene databases and prediction tools are also available that help assess AMR gene content in whole genomes with high accuracy [21].
Here, we used WGS to investigate the prevalence of acquired AMR-conferring genes in E. coli and K. pneumoniae isolated from urine samples taken from patients in rural areas of KY and UG that presented UTI-like symptoms. Our analysis of AMR determinants was limited to those associated with the pan genome and mutation in core genes responsible for antibiotic resistance were not investigated. We further explored their phylogenetic relationships of the isolates collected with other currently circulating African and global strains. This study represents a pilot project of the HATUA consortium. HATUA stands for Holistic Approach to Unravel Antibiotic Resistance in East Africa and the team is comprised of researchers from different disciplines that aim to tackle the main drivers of AMR among uropathogens in East Africa.

Study Design and Patient Recruitment
A total of N = 150 bacterial isolates were obtained from patients in KY (n = 91) and UG (n = 59) presenting UTI-like symptoms, as part of a larger study. Ethical Review Board of University of St Andrews ethical approval, Approval code MD14548 and KY (KEMRI/SERU/P00112/3865) approved verbal consent taken from all the patients. Important patient data such as name, age, gender, location was recorded, and unique identification number were assigned to each patient.

Library Preparation and Whole Genome Sequencing
Bacterial genomic DNA for the isolates were extracted using the QIAxtractor (Qiagen, Valencia, CA, USA) according to the manufacturer's instructions. Library preparation was conducted according to the Illumina protocol and sequenced (96-plex) on an Illumina MiSeq platform (Illumina, San Diego, CA, USA) using 250 bp paired-end reads.

Read Library Quality Control, Mapping and De Novo Genome Assembly
Illumina MiSeq read libraries were rid of sequencing adapters and ambiguous bases using Fastp [22]. Sets that passed the quality filtering were de novo assembled using Unicycler v4.6 [23] pipeline in normal mode to merge contigs.
The read libraries were mapped to reference sequences using SMALT v7.6 (http: //www.sanger.ac.uk/resources/software/SMALT/ (accessed on 18 December 2019) [24] and the resulting SAM files were converted to BAM format, sorted and PCR duplicates removed using SAMtools v1.19 [25]. Strain TOP52_1721_U1 [26] was used the reference genome for the K. pneumoniae samples while the strain EC958 [27] was employed as the reference sequence for the E. coli population.

Pangenome Analyses
The resulting annotation files from Prokka v.1.10 [35] were used as the basis for generating a pangenome for each species set. This step was completed by running Roary v3.11.2 [45] with a 100% BLAST v2.6.0 identity threshold using the MAFFT v7.3 setting [46]. Pangenome outputs were also used to assess the accessory genome composition of each bacterial population and as basis for reconstructing core genome phylogenies.

Patient and Bacterial Strain Profiles
From the total of N = 150 strains, we collected from urine samples of patients, n = 81 were identified as E. coli and n = 19 were K. pneumoniae. The respondents were either to be admitted or visiting rural hospitals in KY and from clinics in the countryside of UG.

Prevalence of AMR Genes in E. coli and K. pneumoniae Uropathogens from KY and UG
All n = 55 E. coli and n = 19 K. pneumoniae isolates harbored type 1 fimbrin. Among the UPEC, fimH30 was the most common allele, followed by fimH41; n = 4/55 samples had type fimH22 and n = 2/55 singleton were found with fimH22.

Population Structure of KY and UG Uropathogens
The UPEC collection was polyclonal. Eighteen (18) different sequence types were identified in the UPEC population (Achtman scheme). The most prevalent MLST sequence types were ST131 (n = 17/55, 31%), ST335 and ST1193 (n = 6/55, 11%) and ST10 (n = 4/55, 7%). These sequence types were usually associated with UTI cases ( Other clones were also observed: n = 3 ST73, n = 2 each from ST155, ST410, ST6161 and ST162, and singletons from ST44, ST48, ST165, ST167, ST212, ST448, ST617, ST648 and ST2163; n = 2 strains from UG (BN2 and BN48) were unclassified (Figure 1a, Table 2). E. coli isolates from UG belong to 15 STs and were thus more diverse compared to those collected from KY, which belong to only 6 STs ( Figure 1a, Table 2). This difference in diversity is consistent with the number of serotypes found in UG relative to those from KY: Ugandan strains belong to 20 different O:H antigen combinations while the KYn ones were found to have 9 O:H types.
We compared our E. coli samples from the three most prevalent clones, ST131, ST335 and ST10, and our K. pneumoniae strains with previously published genomes listed in BacWGSTdb 2.0. Based on the metadata of the reference genomes, these strains were of different geographical origins (country/state) and were mostly isolated from human hosts and have caused disease (Supplementary Table S2). Computing for the pairwise SNP distances showed that strain CP023853 is the most closely related genome with our KY isolates with distances ranging from 910-1489; CP023853 was also sampled from a UTI patient in Sweden in 2009 (Supplementary Table S2; Supplementary Figure S2a). Our ST335 collection is solely composed of KY isolates, and all appeared to be genetically distant to the selected sequences in the database with a minimum of 4700 SNP differences between the two groups (Supplementary Figure S2a). In contrast, our E. coli ST10 strains were all from UG. The closest reference isolate was LSBS01 (isolated from a fecal sample; Supplementary Table S2), which was 2009 and 2070 SNPs apart from BN20 and BN70, respectively (Supplementary Figure S2a).
Our K. pneumoniae collection, which was dominated by ST11 showed~3500 SNP differences from strain 27 from KY while those that had no defined ST (e.g., BN14 and BN 16 from UG) appeared to be most closely related (minimum SNP distances of 3519 and 3255) to the human isolate references LXMM01 and VUBS01, respectively (Supplementary  Table S2; Supplementary Figure S2b).

Plasmid Characterization
Genome assemblies of the KY and UG uropathogens were screened for the presence and type of plasmids using PlasmidFinder v.2.1.1. N = 47/55 in the E. coli collection were found with at least one plasmid. IncFIA was consistently found in n = 10 had both IncFIA and IncFII, n = 9 contained IncFIA, IncFII, Col156 types, n = 1 was detected with IncFIA, IncFII and IncY only or IncI only and IncFII-IncFIA-IncX4 plasmid combinations.
All the samples from the K. pneumoniae collection were found with at least one plasmid type. IncFII-IncFIB-IncR is the most common combination and is found among n = 15 isolates, while n = 2/19 was found with IncR, IncFII, IncFIA, Col, and IncX4. Notably, the strain 90 from KY had the bla CTX-M-1 gene-carrying plasmid IncN and BN7 from UG had the bla NDM -associated IncR.

Discussion
We assessed the prevalence of acquired AMR characteristics among uropathogenic E. coli and K. pneumoniae circulating in East African region using WGS. We recruited out-patients that presented UTI-like symptoms from rural areas in KY and UG, which represents a limitation of our sample collection. The lack of point-of-care diagnostic tool such as the use of dipstick test also contributed to some difficulties in our screening. This is evidenced by a high level of contaminants that comprised of strains that do not contain UTI determinants. Nevertheless, our in silico predictions using whole genome analysis revealed alarming rates of ESBL-producing and MDR strains in both our UPEC and K. pneumoniae collections, which reiterates the great necessity for effective interventions to curb their spread.
Our results firmly indicate a high diversity among E. coli uropathogens, which was more evident in samples taken from UG rather than KY. Strains that belong to the same clonal group had <200 core SNPs from each other. This rich genetic diversity is consistent with those observed in other isolates collected from rural or semi-rural communities of low/middle-income countries [47][48][49]. The widely disseminated UTI-causing clones ST131, ST335 and ST10 were common among our E. coli strains and dominated our Kenyan collection. This is unsurprising as these STs are reported to be circulating globally [50][51][52]. What is remarkable is the detection of emerging clones such as ST1193 and ST617 that were unusually associated with UTI [53,54] albeit observed in hospital settings. UPEC strains from UG are even more alarming as they represent higher number of unusual or novel UTI clones (i.e., ST155, ST448, and ST162) with potentially higher virulence levels [55][56][57] compared to those globally-known STs.
Several Klebsiella species were known to have broad-spectrum resistance to common antibiotics [58]. K. pneumoniae strains particularly those belonging to the hypervirulent ST11 have been extensively reported to cause severe infections [59,60] and have led to dire disease outcomes in intensive care units [61]. This K. pneumoniae clone has an alarming antibiotic resistance profile [62] making it difficult to treat. The dominance of ST11 strains in our samples that were mainly collected from outpatients suggests the strong presence of this clinically important bacterial pathogen in the community and pose an apparent threat to public health.
Bla CTX-M genes were present in 40% of our UPEC collection and in all but one K. pneumoniae strain (95%). Notably, the bla CTX-M-15 gene that confers resistance to last-resort antibiotics was found in high levels in both countries. This gene was detected with other ESBL determinants, bla TEM and bla OXA-1 in E. coli and with bla NDM in K. pneumoniae, concordant with those in uropathogens found from the Middle East [63] and Asia [64], among others. Consistent with previous findings in other African regions, tet genes in this study were also detected alongside ESBL genes bla CTX-M-15 , bla OXA-1 and bla TEM in n = 30/55 E. coli and with bla LEN-3/4/5/6 among n = 3/19 K. pneumoniae [65,66] which stipulates their co-selection and co-transmission in KY and UG. The presence of these genes in the identified plasmid-associated contigs suggest that the mode of transfer may have been plasmid-mediated.

Conclusions
We underline in this pilot study the high frequency of AMR determinants associated with resistance to common antibiotic classes among E. coli and Klebsiella pneumoniae in East Africa, with specific focus on MDR and ESBL-producing strains from KY and UG. We further demonstrate that routine genomic surveillance is necessary for high-resolution investigation of bacterial epidemiology especially in less represented regions. Our findings have significant implications on improving interventions that aim to address the strong presence of AMR pathogens that cause UTI (particularly in low/middle-income countries).

Supplementary Materials:
The following are available online at https://www.mdpi.com/article/10 .3390/antibiotics10121547/s1. Table S1: Metadata of E. coli and K. pneumoniae strains isolated from urine samples including Antibiotic Sensitivity Test results of n = 16 isolates; Table S2: Comparison of host niche, disease implication, isolation source, collection year, antimicrobial and virulence gene contents between three most abundant E. coli clones, ST131, ST335 and ST10 (a) and K. pneumoniae ST11 (b) strains in our study and selected isolate genomes listed in BacWGSTdb 2.0. ND means not determined; Figure S1: Distribution of antimicrobial resistance genes (AMRGs; right panel) among E. coli (a) and K. pneumoniae (b) isolates from our HATUA Pilot collection. Left panel shows clustering of the strains in a phylogenetic tree according to the presence (green blocks) or absence (pink blocks) of AMRGs; Figure S2: Pairwise SNP distances in core genome multi-locus sequence type (cgMLST)-based alleles of the three most abundant E. coli clones, ST131, ST335, and ST10 (a) and K. pneumoniae ST11 (b) strains in our study and selected isolate genomes listed in BacWGSTdb. The assemblies of the reference sequences were downloaded from European Nucleotide Archive.