Circulation of Extended-Spectrum Beta-Lactamase-Producing Escherichia coli of Pandemic Sequence Types 131, 648, and 410 Among Hospitalized Patients, Caregivers, and the Community in Rwanda

Multi-drug resistant (MDR), gram-negative Enterobacteriaceae, such as Escherichia coli (E. coli) limit therapeutic options and increase morbidity, mortality, and treatment costs worldwide. They pose a serious burden on healthcare systems, especially in developing countries like Rwanda. Several studies have shown the effects caused by the global spread of extended-spectrum beta-lactamase (ESBL)-producing E. coli. However, limited data is available on transmission dynamics of these pathogens and the mobile elements they carry in the context of clinical and community locations in Sub-Saharan Africa. Here, we examined 120 ESBL-producing E. coli strains from patients hospitalized in the University Teaching Hospital of Butare (Rwanda), their attending caregivers as well as associated community members and livestock. Based on whole-genome analysis, the genetic diversification and phylogenetics were assessed. Moreover, the content of carried plasmids was characterized and investigated for putative transmission among strains, and for their potential role as drivers for the spread of antibiotic resistance. We show that among the 30 different sequence types (ST) detected were the pandemic clonal lineages ST131, ST648 and ST410, which combine high-level antimicrobial resistance with virulence. In addition to the frequently found resistance genes blaCTX–M–15, tet(34), and aph(6)-Id, we identified csg genes, which are required for curli fiber synthesis and thus biofilm formation. Numerous strains harbored multiple virulence-associated genes (VAGs) including pap (P fimbriae adhesion cluster), fim (type I fimbriae) and chu (Chu heme uptake system). Furthermore, we found phylogenetic relationships among strains from patients and their caregivers or related community members and animals, which indicates transmission of pathogens. Also, we demonstrated the presence and potential transfer of identical/similar ESBL-plasmids in different strains from the Rwandan setting and when compared to an external plasmid. This study highlights the circulation of clinically relevant, pathogenic ESBL-producing E. coli among patients, caregivers and the community in Rwanda. Combining antimicrobial resistance with virulence in addition to the putative exchange of mobile genetic elements among bacterial pathogens poses a significant risk around the world.

Multi-drug resistant (MDR), gram-negative Enterobacteriaceae, such as Escherichia coli (E. coli) limit therapeutic options and increase morbidity, mortality, and treatment costs worldwide. They pose a serious burden on healthcare systems, especially in developing countries like Rwanda. Several studies have shown the effects caused by the global spread of extended-spectrum beta-lactamase (ESBL)-producing E. coli. However, limited data is available on transmission dynamics of these pathogens and the mobile elements they carry in the context of clinical and community locations in Sub-Saharan Africa. Here, we examined 120 ESBL-producing E. coli strains from patients hospitalized in the University Teaching Hospital of Butare (Rwanda), their attending caregivers as well as associated community members and livestock. Based on whole-genome analysis, the genetic diversification and phylogenetics were assessed. Moreover, the content of carried plasmids was characterized and investigated for putative transmission among strains, and for their potential role as drivers for the spread of antibiotic resistance. We show that among the 30 different sequence types (ST) detected were the pandemic clonal lineages ST131, ST648 and ST410, which combine high-level antimicrobial resistance with virulence. In addition to the frequently found resistance genes bla CTX−M−15 , tet(34), and aph(6)-Id, we identified csg genes, which are required for curli fiber synthesis and thus biofilm formation. Numerous strains harbored multiple virulence-associated genes (VAGs) including pap (P fimbriae adhesion cluster), fim (type I fimbriae) and chu (Chu heme uptake system). Furthermore, we found phylogenetic relationships among strains from patients and their caregivers or related community members and animals, which indicates transmission of pathogens. Also, we

INTRODUCTION
The versatility of Escherichia coli (E. coli) is based on the diversity of genetic substructures within this species (Whittam et al., 1983). In addition to commensal strains, which are an essential part of the non-anaerobic intestinal microflora of humans, other mammals and birds, pathogenic variants occur. The dissimilarity of these pathotypes depends also on their virulence attributes, resulting in a wide range of pathologies in both humans and animals. The intestinal pathogenic E. coli (InPEC) express characteristic virulence factors that allow to adhere and invade intestinal cells, causing specific enteric and diarrheal diseases. While InPEC are obligate pathogens, extraintestinal pathogenic E. coli (ExPEC) are part of the intestinal microbiome but exhibit a heterogeneous composition of virulence factors to colonize niches such as the urinary tract (Kaper et al., 2004). They can thus cause infections in almost any organ or non-intestinal site, regardless of the state of the host's immune system (Russo and Johnson, 2000). However, a strict differentiation of pathogenic and commensal E. coli is difficult, provided by their rapid geno-and phenotypic adaptation to changing environmental conditions, for example through horizontal gene transfer (Pallen and Wren, 2007). Despite the plasticity of the genome, phylogenetic studies have shown some clonality within the population structure of E. coli, from which seven distinct phylogenetic groups were derived (Jaureguy et al., 2008;Touchon et al., 2009;Clermont et al., 2013). Usually, commensal strains and obligatory pathogens belong to the phylogroups A and B1, whereas strains with extended virulent attributes (mainly ExPEC) are part of the phylogroups B2, D, and F, with the latter as a sister group of B2 (Escobar-Páramo et al., 2004;Clermont et al., 2013). Multi-locus sequence typing (MLST) allows additional classification and several phylogenetic studies suggest the spread of pandemic high-risk clonal lineages including primarily sequence type (ST) 131 (Nicolas-Chanoine et al., 2008;Ewers et al., 2010;Hussain et al., 2012), ST648 (Ewers et al., 2014;Schaufler et al., 2019), ST410 (Schaufler et al., 2016b;Zurita et al., 2020), putatively ST405 (Manges et al., 2019), and others.
The management of zoonotic infections caused by antibioticresistant bacteria has become a multidisciplinary challenge for all modern healthcare systems and is nowadays often approached in a holistic One Health context. Bacterial pathogens spread through direct contact among humans and animals, indirectly by (environmental) pollution and also through non-living and living vectors (Rahman et al., 2020). One example for the latter are houseflies, which have been demonstrated to carry antibiotic-resistant pathogens including extended-spectrum beta-lactamases-(ESBL)-producing E. coli (Rahuma et al., 2005;Heiden et al., 2020b;Tufa et al., 2020) non-susceptible to third-generation cephalosporins (e.g., cefotaxime) and monobactams (e.g., aztreonam) (Bevan et al., 2017). Notably, ESBL enzyme production is often accompanied by cross-and co-resistances (Cantón and Coque, 2006;Hidron et al., 2008;Pitout, 2012) resulting in multi-drug resistant (MDR) representatives (Beceiro et al., 2013).
The One Health concept-addressing human, animal and environmental health-encounters some challenges, especially in low-income countries like Sub-Saharan Africa/Rwanda. On the one hand, the lack of surveillance systems may result in inadequate establishment and implementation of hygienic strategies and therapy guidelines (Muvunyi et al., 2011;Ntirenganya et al., 2015;Carroll et al., 2016). On the other hand, uncontrolled over-the-counter sale of partially counterfeit and substandard antibiotic drugs (Kayumba et al., 2004;Carroll et al., 2016) as well as close human-livestock contact and household crowding might contribute to the broad occurrence and interspecies transmission of MDR bacteria in Sub-Saharan Africa.
This study aimed to investigate whether (i) ESBL-producing E. coli circulate among patients, caregivers, the community, and animals in Rwanda, (ii) some of these belong to pandemic highrisk clonal lineages and how they are phylogenetically related, (iii) they demonstrate virulence-associated features, (iv) their mobile genetic elements contribute to the spread of antibiotic resistance.

Bacterial Strains
The E. coli strains investigated in this study were sampled over a time period of 8 weeks at the University Teaching Hospital of Butare (Rwanda) in 2014 (previously described by Kurz et al., 2017). Rectal swabs (Sarstedt AG & Co. KG, Nümbrecht, Germany) were collected from patients and caregivers at admission and discharge as well as from several community members and animals. Each patient had their own caregiver, who were usually relatives accompanying the patient upon admission. They stayed in the patient's room and were involved in personal care of the patient and food preparation. This is a common practice in African hospitals (Hoffman et al., 2012;Ugochukwu, 2013). Sample groups consisting of a patient and related caregiver, and associated family members, neighbors and/or pets were included in the same study-ID. The samples were plated onto chromogenic agar (CHROMagar-ESBL, Mast Diagnostica GmbH, Reinfeld, Germany) supplemented with 2 µg/mL cefotaxime (Cayman Chemical Company, Ann Arbor, United States) and incubated at 37 • C. For putative ESBL-positive colonies, the production of ESBL and/or ampicillinase (AmpC) was verified (ESBL-AmpC-Detection Test, Mast Diagnostica GmbH, Reinfeld, Germany) and all strains positive for AmpC only were excluded. The strains were stored at -80 • C in LB broth (Carl Roth GmbH & Co. KG, Karlsruhe, Germany) supplemented with 20% (V/V) glycerol (Merck KGaA, Darmstadt, Germany). Originally, we have obtained overall 289 ESBL-producing E. coli strains (from patients, caregivers, community members, and animals), with 120 selected strains (based on related study-IDs) that were whole-genome sequenced. Additionally, flies caught with fly traps at different wards of the hospital examined in a previous study (Heiden et al., 2020b) were partly included in this study (Supplementary Table 1).

Whole-Genome Sequencing
One single E. coli colony was cultured in LB broth supplemented with 2 µg/mL cefotaxime overnight and the total DNA was extracted using the MasterPure TM DNA Purification Kit for Blood, Version II (Lucigen, Middleton, United States) according to the manufacturer's instructions. DNA was puritycontrolled and quantified using NanoDrop TM 2000 (Thermo Fisher Scientific Inc., Waltham, United States). WGS was performed in collaboration with LGC (LGC Genomics GmbH, Berlin, Germany) with 150 bp paired-end-reads using Illumina NextSeq 500/550 V2.

Genomic Analysis
Raw reads were quality-trimmed, adapter-trimmed and contaminant-filtered using BBDuk from BBTools v. 38.86 1 . After de novo assembly (at a maximum coverage of 100×) using shovill v. 1.1.0 2 in combination with SPAdes v. 3.14.1 (Bankevich et al., 2012), draft genomes were polished by mapping all trimmed reads back to the contigs with bwa v. 0.7.17 (Li and Durbin, 2009), processing SAM/BAM files marking optical duplicates with Samtools v. 1.10 ) and calling variants with Pilon v. 1.23 (Walker et al., 2014) (Supplementary Table 2).

Minimum Inhibitory Concentration of Colistin
When the genotype was positive for colistin resistance (presence of mcr genes), we evaluated the resistance phenotype by determining the minimum inhibitory concentration (MIC) using MICRONAUT MIC-Strip Colistin (Merlin Diagnostika GmbH, Bornheim, Germany) according to the manufacturer's instructions and interpreted the results according to the published breakpoints of EUCAST (The European Committee on Antimicrobial Susceptibility Testing, 2021). Experiments were performed thrice.

Phylogenetic Relationships
To elucidate phylogenetic relationships, we constructed a tree in an alignment-free manner for all investigated genomes (Supplementary Figure 1). Additionally, we inferred phylogenies, which are based on SNPs in the core genome of strains belonging to the five predominant STs of this study to assess potential transmission scenarios (Figure 2). For comparative reasons, we included 13 genomes of ESBLproducing E. coli isolated from houseflies (Heiden et al., 2020b) originating from the same Rwandan hospital.
The phylogenetic analysis (Supplementary Figure 1) shows that the E. coli strains were distributed among six distinct phylogroups and grouped into several clades according to their sequence type. Within these ST-associated phylogenies, several sub-clades were defined with genomes interspersed in patients, caregivers, related community members as well as animals and flies, which suggests common phylogenetic backgrounds (Clermont et al., 2011) potentially based on interspecies transmission. For example, PBIO458 (study-ID 60) and PBIO459 [(study-ID 133)-both ST38 isolated from two animals-clustered with four different strains from patients, family members and neighbors, PBIO272 (patient admission, study-ID 60), PBIO451 (neighbor, study-ID 60), PBIO455 (family member, study-ID 60), and PBIO467 (patient follow-up, study-ID 60)]. These findings are corroborated by results of two of our previous studies, where we demonstrated the likely transmission of ESBL-producing E. coli ST38 among humans and animals (Guenther et al., 2017;Schaufler et al., 2018). Note, however, that in this current study, only one representative of ST38 [PBIO302 (caregiver admission, study-ID 131)] carried a chromosomally encoded bla CTX−M−15 gene and the before mentioned strains carried plasmid-encoded ESBLs (Supplementary Figure 1), which is contrary to our previous findings. We then compared the bla CTX−M−15 gene-carrying chromosomal contig of PBIO302 to two of the plasmid-encoded bla CTX−M−15 sequences of ST38 (PBIO272 and PBIO459; Supplementary Figure 2). The chromosomal sections of PBIO302, PBIO272, and PBIO459 were highly similar, except the chromosomal insertion of bla CTX−M−15 in PBIO302. This resistance gene was flanked by transposable elements, as described below.
In addition, strains with the numbers PBIO1939, PBIO1942, PBIO1945, PBIO1946, and PBIO1947 from houseflies were in the same sub-clade as strains isolated from different human sources indicating the potential role of living vectors in the spread of pathogenic bacteria.
For the phylogenetic trees of the five predominant STs of this study (Figure 2) it is interesting to notice that some genomes stemming from different sources were more closely related than genomes from the same source. For example, PBIO283 [(study-ID 92) Figure 2A, ST131, sub-clade 1], which originates from a caregiver at admission differed in 0.2 ± 0.0003 SNPs/Mbp with strains isolated from the related patient at admission [PBIO286 (study-ID 92)] as well as discharge [PBIO285 (study-ID 92)] and unrelated patients at discharge [PBIO293 (study-ID 114) and PBIO296 (study-ID 117)]. Moreover, PBIO440 [(study-ID 133) Figure 2A, ST131, sub-clade 2], isolated from a community member, only varied in 0.2 SNPs/Mbp compared to a follow-up strain of an already discharged patient [PBIO405 (study-ID

434)]
. Notably, all strains belonging to ST354 (Figure 2B, subclade 3) only differed in 1.0 ± 0.4 SNPs/Mbp including one strain carried by a housefly (PBIO1945). Also interesting was the difference between PBIO368 [(study-ID 335) Figure 2E, ST648, sub-clade 4], originating from a patient at discharge, and strains of distinct sources [PBIO354 (caregiver admission, study-ID 265), PBIO355 (patient discharge, study-ID 265), and PBIO367 (caregiver admission, study-ID 288)], differing in 0.9 ± 0.1 SNPs/Mbp. These numbers of SNPs were up to 10-fold lower than described for clonal EHEC strains during an outbreak in Germany (1.8 SNPs/Mbp) (Grad et al., 2012;Been et al., 2014), suggesting the circulation of only a handful of sequence types in this African setting, which interestingly happen to mostly be international high-risk clonal lineages. In addition, some strains from identical STs were carried by both flies and humans, for example PBIO1945 and PBIO374 (caregiver discharge, study-ID 401; Figure 2B), again indicating the role of flies in the spread of antibiotic-resistant pathogens.
In addition to the resistances described, genes encoding for efflux pumps were found frequently, with mdfA in all (120/120) and acrB in 85.8% (103/120) of all genomes.
The strains belonging to the five predominant STs carried several VAGs, mainly associated with adherence, antiphagocytosis, biofilm formation, invasion, iron uptake and bacterial secretion. The ability to attach to surfaces/cells and form biofilms is a common strategy used by bacterial populations to resist antibiotic treatment and host defense mechanisms as well as cause infection (Moser et al., 2017;Amanatidou et al., 2019). In particular, genes for the P fimbriae adhesion cluster [pap operon (70.5%; 43/61)], Dr family of adhesins (4.9%; 3/61) and type I fimbriae [fim (100%; 61/61)], which are necessary for uroepithelia cell adhesion and invasion, and, thus, for causing urinary tract infection (Mulvey, 2002), were frequently found. Notably, strains belonging to ST354 and ST410 showed a lack of pap genes, which is consistent with previous findings (Vangchhia et al., 2016;Zogg et al., 2018;Schaufler et al., 2019). Furthermore, we detected several members of the csg gene family in all genomes (100%; 120/120). These genes encode curli fibers, which are essential components of bacterial biofilms (Hammar et al., 1995;Evans and Chapman, 2014).

Mobile Genetic Elements
We next investigated the occurrence and circulation/transfer of ESBL-plasmids among strains and thus their contribution to the spread of antibiotic resistance in the Rwandan setting.
To better assess similar plasmid backgrounds, we compared the plasmid sequences of all strains that carried a plasmidborne bla CTX−M−15 gene against the plasmid sequence of PBIO241 (ST405; patient discharge; study-ID 159; Figure 4) as a representative for the most prevalent ST. Keep in mind, however, that the selection of the reference biases the visual representation when a large number of plasmid sequences is absent in the query sequences. On the other hand, plasmid sequences, not present in the reference but in queries, are missing in this approach. The plasmid backbones of other ST405 FIGURE 3 | Presence/absence of virulence factors (VAGs) among ESBL-producing E. coli belonging to the five most prevalent sequence types (STs). Color-filled boxes show the presence of genes (coverage and identity ≥ 65%) encoding for different VAGs.
We also compared some of our plasmid sequences of representative strains (criteria: top BLAST-hits) against a publicly available reference plasmid pIV_IncHI2_CTX_M_15 (Marchetti et al., 2020 ; Supplementary Figure 3). This bla CTX−M−15containing plasmid has been originally obtained from an E. coli ST58 strain causing a deadly puppy infection in Italy in 2019. Interestingly, the plasmid backbone of PBIO251 (ST155, patient admission, study-ID 153) was nearly identical when compared to this external Italian plasmid. The plasmids of PBIO275 (ST540, caregiver admission, study-ID 78), PBIO385 (ST3494, patient discharge, study-ID 434), and PBIO1943 (ST5474, housefly) also showed high similarities to the reference and thus among each other.
The presence of similar resistance plasmids in strains of distinct sequence types and sample groups indicates the potential transmission of mobile genetic elements and the growing prevalence of successful plasmid families (Carattoli, 2011).

DISCUSSION
In contrast to Europe, the United States of America and Australia, only little information is available concerning the exact characteristics and distribution of ESBL-producing E. coli in Africa. This study reveals the broad occurrence and potential circulation of several international high-risk clonal lineages in a hospital and associated locations in Rwanda, following up on a publication by Kurz et al. (2017). Despite the potential bias due to the analysis of both admission and discharge strains of the same patient and/or caregiver, it is remarkable that our sample set was dominated by ST131, ST648, and ST410, which have been frequently reported from humans, animals and the environment (Nicolas-Chanoine et al., 2008;Ewers et al., 2010Ewers et al., , 2014Hussain et al., 2012;Schaufler et al., 2016bSchaufler et al., , 2019 and which are "classic" pandemic, high-risk clonal lineages. In addition to these, we found ST354 to be the third most detected sequence type in this study. Its emergence has been previously described globally except for the African continent (Manges et al., 2019). This locally restricted accumulation of one single ST in combination with a small number of other STs indicates re-entering and circulation of Interestingly, within the ST-associated clades, some genomes without genetic differences were interspersed in humans (hospitalized patients and caregivers as well as community members) and animals. This putative lack of host adaptation and the close phylogenetic relationships indicate the colonization and rapid transmission of several clones within the community and the potential transmission into the clinical setting and vice versa, underlined by the high acquisition rates of ESBLproducing E. coli during hospitalization as described previously (Kurz et al., 2017).
The resistance genes found in this study confer resistances to antibiotics frequently used in veterinary medicine and/or in sub-therapeutic doses as food supplements and growth promotors in Africa (Eagar et al., 2012;Adesokan et al., 2015;Mainda et al., 2015;Manishimwe et al., 2017). When also considering the zoonotic character of ESBL-producing E. coli, it is not surprising that we found clonal strains with similar patterns of resistance features in the different sample groups. Transmission likely occurred among patients and caregivers/family members and was also influenced by livestock animals due to close human-animal contact (Klous et al., 2016).
Two strains of this study carried the mcr-9 gene but were phenotypically susceptible to colistin. This phenomenon was first reported in 2019 (Carroll et al., 2019;Kieffer et al., 2019). Due to the structural heterogeneity compared to other mcr genes (65% amino acid identity with the most closely related mcr-3 gene) and the weak inactivation of colistin, the clinical importance of mcr-9 is unknown (Tyson et al., 2020).
Notably, some of the strains showed extensive, chromosomally encoded virulence-associated features. The CNF-1-encoding gene (cnf1) detected in strain PBIO350 (ST648), for example, is associated with causing neonatal meningitis (Khan et al., 2002). In addition to the major virulence factors of meningitisassociated and uropathogenic E. coli (like P fimbriae adhesion cluster, K1 capsule, heme utilization systems, and the secreted autotransporter toxin), the strains showed VAGs typical for InPEC (especially the various bacterial secretion systems) underlining the clinical relevance of these pathogens.
Finally, we demonstrate that similar plasmid sequences were present in strains from different sample groups, thus likely indicating mobile genetic element transmission, and underlining the importance of plasmid-driven spread of antimicrobial resistance independent of the host's phylogenetic background (Schaufler et al., 2016a;Ranjan et al., 2018). Interesting in addition to similarities among strains from the Rwandan setting is in particular the close relationship to an external plasmid, which has been obtained only recently (Marchetti et al., 2020). This highlights the sometimes global spread of such mobile genetic elements and their bacterial hosts.

CONCLUSION
In this study, we investigated and identified the presence of clinically relevant ESBL-producing E. coli that circulate among patients, caregivers, the community and animals in a Rwandan setting. The findings contribute to the understanding of the global dissemination of bacterial high-risk clonal lineages, their virulence features as well as plasmid transmissions. They also underline the potential role of houseflies in this harmful dynamic.

DATA AVAILABILITY
The data for this study have been deposited in the European Nucleotide Archive (ENA) at EMBL-EBI under accession number PRJEB42795 (https://www.ebi.ac.uk/ena/browser/view/ PRJEB42795).

AUTHOR CONTRIBUTIONS
KS and EE designed the study. EE and JM performed the laboratory and phenotypic experiments. SH, SS, and KK performed the bioinformatics analyses. KS, EE, SH, KK, CB, JN, AS, JG, MK, FM, and SS analyzed the data. KS, EE, and SH wrote the manuscript and prepared the tables and figures. All authors read and approved the final version of the manuscript.