Prediction of antimicrobial resistance based on whole-genome sequencing and machine learning.

MOTIVATION: Antimicrobial resistance (AMR) is one of the biggest global problems threatening human and animal health. Rapid and accurate AMR diagnostic methods are thus very urgently needed. However, traditional antimicrobial susceptibility testing (AST) is time-consuming, low throughput and viable only for cultivable bacteria. Machine learning methods may pave the way for automated AMR prediction based on genomic data of the bacteria. However, comparing different machine learning methods for the prediction of AMR based on different encodings and whole-genome sequencing data without previously known knowledge remains to be done. RESULTS: In this study, we evaluated logistic regression (LR), support vector machine (SVM), random forest (RF) and convolutional neural network (CNN) for the prediction of AMR for the antibiotics ciprofloxacin, cefotaxime, ceftazidime and gentamicin. We could demonstrate that these models can effectively predict AMR with label encoding, one-hot encoding and frequency matrix chaos game representation (FCGR encoding) on whole-genome sequencing data. We trained these models on a large AMR dataset and evaluated them on an independent public dataset. Generally, RFs and CNNs perform better than LR and SVM with AUCs up to 0.96. Furthermore, we were able to identify mutations that are associated with AMR for each antibiotic. AVAILABILITY AND IMPLEMENTATION: Source code in data preparation and model training are provided at GitHub website (https://github.com/YunxiaoRen/ML-iAMR). SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Ecology of Listeria monocytogenes and Listeria species in India: the occurrence, resistance to biocides, genomic landscape and biocontrol.

Listeria monocytogenes, the causative agent of listeriosis, has been implicated in increasing foodborne outbreaks worldwide. The disease is manifested in various forms ranging from severe sepsis in immune-compromised individuals, febrile gastroenteritis, still birth, abortions and meningoencephalitis. In India, data from studies on the detection and molecular epidemiological analysis of L. monocytogenes are only recently emerging. The presence of Listeria in different ecological niches has been recorded from India, including foods, soil, vegetables, mangrove swamps, seafood, freshwater fishes, clinical cases, and also insects. The organism has also been isolated from women with spontaneous abortions, miscarriage or recurrent obstetric history, aborted foetuses, animal clinical cases and wildlife samples. A novel species of Listeria has also been characterized. Listeria monocytogenes strains isolated from clinical, environmental, and foods showed biofilm-forming abilities. Listeria monocytogenes serotype 4b isolates of ST328, a predominant and unique ST observed in India, was repeatedly isolated from different sources, times, and geographical locations. Here, we reviewed the occurrence of Listeria in different sources in India, its resistance to biocides, and provide epidemiological analysis on its genomic landscape.

Genetic Diversity of Listeria monocytogenes Isolates from Invasive Listeriosis in China.

Listeria monocytogenes is a deadly foodborne pathogen, and infections can result in meningoencephalitis and sepsis with mortality rates of up to 30%. In this study, we performed comparative whole-genome analysis of 30 clinical isolates sequenced together with 32 previously sequenced clinical and food isolates from China. The data indicate that L. monocytogenes isolates belonging to the clonal complexes (CC) -1, -8, -9, -87, -121, and -155 are present in human clinical cases. The majority of isolates are from CC-87, 9, and 8 and overlap with those CCs previously reported on the basis of multilocus sequence typing for isolates from Chinese food products. Detailed genome analysis of isolates, representative of CCs in clinical and food products, revealed strong similarities both in their core- and accessory genomes indicating that they are highly related. When compared to genome sequences of isolates of a given CC worldwide, clinical isolates of China were distinct and clustered in unified clades. Our data indicate that epidemic clones of L. monocytogenes (CC-87, 9, and 8) with unusually high occurrence of plasmids are unique to China and suggest that common populations of L. monocytogenes clones are present in both clinical and food products in China.

Genome-based analyses indicate that Serratia marcescens subsp. marcescens and Serratia marcescens subsp. sakuensis do not merit separation to subspecies status.

The species Serratia marcescens currently comprises two subspecies, viz. Serratia marcescenssubsp. marcescens and Serratia marcescenssubsp. sakuensis. The latter species was created by considering the differentiating endospore-formation capability of isolate KREDT. A member of the Subcommittee on the Taxonomy of Enterobacteriaceae for the International Committee on Systematics of Prokaryotes conducted a re-examination of S. marcescens KREDT and failed to detect its endospore-forming capability, thereby questioning the validity of S. marcescenssubsp. sakuensis as a separate subspecies. Nevertheless, the taxonomic allocation of S. marcescenssubsp. sakuensis has been followed till date. To obtain clarity for taxonomic allocation based on genome-based analysis, we sequenced the genome of type strain S. marcescenssubsp. sakuensis KREDT=DSM 17174T and carried out the analyses recommended to define the subspecies. As a result, DSM 17174T does not satisfy the genomic criteria for defining a subspecies: i.e. does not form a distinguishable clade by overall genomic related indexing and phylogenomic treeing. Also, there is currently no other rationale for separate recognition of this species. We propose the removal of the S. marcescenssubsp. sakuensis designation to avoid further confusion in the literature.

Genomic and physiological analyses of an indigenous strain, Enterococcus faecium 17OM39.

The human gut microbiome plays a crucial role in human health and efforts need to be done for cultivation and characterisation of bacteria with potential health benefits. Here, we isolated a bacterium from a healthy Indian adult faeces and investigated its potential as probiotic. The cultured bacterial strain 17OM39 was identified as Enterococcus faecium by 16S rRNA gene sequencing. The strain 17OM39 exhibited tolerance to acidic pH, showed antimicrobial activity and displayed strong cell surface traits such as hydrophobicity and autoaggregation capacity. The strain was able to tolerate bile salts and showed bile salt hydrolytic (BSH) activity, exopolysaccharide production and adherence to human HT-29 cell line. Importantly, partial haemolytic activity was detected and the strain was susceptible to the human serum. Genomics investigation of strain 17OM39 revealed the presence of diverse genes encoding for proteolytic enzymes, stress response systems and the ability to produce essential amino acids, vitamins and antimicrobial compound Bacteriocin-A. No virulence factors and plasmids were found in this genome of the strain 17OM39. Collectively, these physiological and genomic features of 17OM39 confirm the potential of this strain as a candidate probiotic.

Lelliottia aquatilis sp. nov., isolated from drinking water.

Five beige-pigmented, oxidase-negative bacterial isolates, 6331-17T, 6332-17, 6333-17, 6334-17 and 9827-07, isolated either from a drinking water storage reservoir or drinking water in 2006 and 2017 in Germany, were examined in detail applying by a polyphasic taxonomic approach. Cells of the isolates were rod-shaped and Gram-stain-negative. Comparison of the 16S rRNA gene sequences of these five isolates showed highest sequence similarities to Lelliottia amnigena (99.98 %) and Lelliottia nimipressuralis (99.99 %). Multilocus sequence analyses based on concatenated partial rpoB, gyrB, infB and atpD sequences confirmed the clustering of these isolates with Lelliottia species, but also revealed a clear distinction to the closest related type strains. Analysis of the genome sequences of these isolates indicated >70 % in silico DNA-DNA hybridization and high average nucleotide identities between strains. Nevertheless, they showed only <70 and <95 % similarity to the type strains of these two Lelliottia species. The fatty acid profiles of these isolates were very similar and consisted of the major fatty acids C16:0, C17 : 0cyclo, C15 : 0iso 2-OH/C16 : 1ω7c and C18 : 1ω7c. In addition, physiological/biochemical tests revealed high phenotypic similarity to each other. These cumulative data indicate that these isolates represent a novel Lelliottia species, for which the name Lelliottia aquatilis sp. nov. is proposed, with strain 6331-17T (=CCM 8846T=CIP 111609T=LMG 30560T) as the type strain.

Listeria goaensis sp. nov.

Two Listeria-like isolates obtained from mangrove swamps in Goa, India were characterized using polyphasic combinations of phenotypic, chemotaxonomic and whole-genome sequence (WGS)-based approaches. The isolates presented as short, non-spore-forming, Gram-positive rods, that were non-motile, oxidase-negative, catalase-positive and exhibited α-haemolysis on 5 % sheep- and horse-blood agar plates. The 16S rRNA gene sequences exhibited 93.7-99.7 % nucleotide identity to other Listeria species and had less than 92 % nucleotide identity to species of closely related genera, indicating that the isolates are de facto members of the genus Listeria. Their overall fatty acid composition resembled that of other Listeria species, with quantitative differences in iso C15 : 0, anteiso C15 : 0, iso C16 : 0, C16 : 0, iso C17 : 0 and anteiso C17 : 0 fatty acid profiles. Phylogeny based on 406 core coding DNA sequences grouped these two isolates in a monophyletic clade within the genus Listeria. WGS-based average nucleotide identity and in silico DNA-DNA hybridization values were lower than the recommended cut-off values of 95 and 70 %, respectively, to the other Listeria species, indicating that they are founding members of a novel Listeria species. We suggest the name Listeriagoaensis sp. nov. be created and the type strain is ILCC801T (=KCTC 33909;=DSM 29886;=MCC 3285).

blaCTX-M-27-Encoding Escherichia coli Sequence Type 131 Lineage C1-M27 Clone in Clinical Isolates, Germany.

We examined extended-spectrum ß-lactamase-producing isolates from livestock, humans, companion animals, food, and the environment during 2009-2016 in Germany for the presence of CTX-M-27 allele within Escherichia coli sequence type (ST) 131. E. coli ST131 C1-M27 was exclusively present in humans; its incidence increased from 0% in 2009 to 45% in 2016.

The status of the species Enterobacter siamensisKhunthongpan et al. 2014. Request for an Opinion.

In the course of a taxonomic study describing novel species of the genus Enterobacter it was found that the 16S rRNA gene sequence of the type strain of Enterobacter siamensis, obtained both directly from the authors of the publication on Enterobacter siamensis and from the Korean Collection for Type Cultures (C2361T and KCTC 23282T, respectively), was not congruent with the 16S rRNA gene sequence deposited in the GenBank database under the accession number HQ888848, which was applied for phylogenetic analysis in the species proposal. The remaining deposit in the Japanese type culture collection, NBRC 107138T, showed an identical 16S rRNA gene sequence to the other two cultures and overall, this sequence differed at 35 positions in comparison with the 1429 bp sequence published under the accession number HQ888848.Therefore, the type strain of this species cannot be included in any further scientific comparative study. It is proposed that the Judicial Commission of the International Committee on Systematics of Prokaryotes place the name Enterobacter siamensis on the list of rejected names, if a suitable replacement for the type strain is not found or a neotype strain is not proposed within two years following the publication of this Request for an Opinion.

Taxonomic reassessment of the genus Elizabethkingia using whole-genome sequencing: Elizabethkingia endophytica Kämpfer et al. 2015 is a later subjective synonym of Elizabethkingia anophelis Kämpfer et al. 2011.

The taxonomic status of all species of the genus Elizabethkingia was re-evaluated by comparative genomics based on whole-genome sequencing. From these results it is clear that Elizabethkingia endophytica is a later subjective synonym of Elizabethkingia anophelis. In addition, genome-based analysis revealed the misidentification of isolates previously identified by traditional approaches and indicates the presence of two more species. We propose a more rapid identification scheme on the basis of an in silico PCR assay derived from comparative genomics of whole-genome sequences (WGS) from 29 well-curated strains.

Enterobacter bugandensis sp. nov., isolated from neonatal blood.

A total of 17 Enterobacter-like isolates were obtained from blood during a septicaemia outbreak in a neonatal unit, Tanzania, that could not be assigned based on phenotypic test to any existing Enterobacter species. Eight representative outbreak isolates were investigated in detail. Fermentation characteristics, biochemical assays and fatty acid profiles for taxonomic analysis were determined and supplemented with information derived from whole genome sequences. Phenotypic and morphological tests revealed that these isolates were Gram-stain-negative, rod-shaped, highly motile and facultatively anaerobic. The fatty acid profile was similar to those of the type strains for all recognized Enterobacter species, with quantitative differences in C17 : 0, C18 : 1ω7c and C17 : 0 cyclo fatty acids. Whole genome sequencing was used to identify taxonomically relevant characteristics, i.e. for 16S rRNA gene sequence analysis, multi-locus sequence analysis (MLSA), in silico DNA-DNA hybridization (isDDH) and average nucleotide identity (ANI). Draft genomes were approximately 4.9 Mb in size with a G+C content of 56.0 mol%. The 16S rRNA gene sequence of these eight isolates showed >97 % similarity to all Enterobacter species, while MLSA clustered them closely with the type strains of Enterobacter xiangfangensis and Enterobacter hormaechei. These eight strains showed less than 70 % isDDH identity with the type strains of Enterobacter species. In addition, less than 95 % ANI to the type strains of Enterobacter species was observed. From these results, it is concluded that these isolates possess sufficient characteristics to differentiate them from all recognized Enterobacter species, and should therefore be considered as representing a novel species. The name Enterobacter bugandensis sp. nov. is proposed with EB-247T ( = DSM 29888T = NCCB 100573T) as the type strain.

Mangrove Ecosystems: An Adopted Habitat for Pathogenic Salmonella spp.

Mangroves are affected by industrial and anthropogenic factors. Although mangroves have been widely studied, investigations of pathogens that may affect public health significance are largely lacking even while incidences of diseases linked with the consumption of mangrove-associated food have increased. A total of 150 samples of water, sediment, and biota were collected from ten mangrove ecosystems in Goa, India. Total viable counts of pathogens such as E. coli, Listeria, Salmonella, and Vibrio spp. ranged from 1.25 to 3.9 × 10(3) cfu/ mL, which were above the relevant standards. Salmonella counts were the highest at 3.1 to 3.9 × 10(3)cfu/mL, with a prevalence of 40%. Considering its high prevalence, the virulence of Salmonella spp. was studied. The invA gene was detected in 35% of the Salmonella isolates by polymerase chain reaction (PCR). The findings suggested that pathogens adapt to this habitat, resulting in contamination of the indigenous fauna.

Phylogenomic grouping of Listeria monocytogenes.

The precise delineation of lineages and clonal groups are a prerequisite to examine within-species genetic variations, particularly with respect to pathogenic potential. A whole-genome-based approach was used to subtype and subgroup isolates of Listeria monocytogenes. Core-genome typing was performed, employing 3 different approaches: total core genes (CG), high-scoring segment pairs (HSPs), and average nucleotide identity (ANI). Examination of 113 L. monocytogenes genomes available in-house and in public domains revealed 33 phylogenomic groups (PGs). Each PG could be differentiated into a number of genomic types (GTs), depending on the approach used: HSPs (n = 57 GTs), CG (n = 71 GTs), and ANI (n = 83 GTs). Demarcation of the PGs was concordant with the 4 known lineages and led to the identification of sublineages in the lineage groups I, II, and III. In addition, PG assignments had discriminatory power similar to multi-virulence-locus sequence typing types and clonal complexes of multilocus sequence typing. Clustering of genomically highly similar isolates from different countries, sources, and isolation dates using whole-genome-based PG suggested that dispersion of phylogenomic clones of L. monocytogenes preceded their subsequent evolution. Classification according to PG may act as a guideline for future epidemiological studies.

Multi-virulence-locus sequence typing of 4b Listeria monocytogenes isolates obtained from different sources in India over a 10-year period.

Listeria monocytogenes is an emerging foodborne pathogen responsible for listeriosis. The incidence of listeriosis has increased during the last 2 decades due to the increase in consumption of ready-to-eat foods and change in food consumption habits. Outbreaks and sporadic cases of listeriosis have been reported in developed countries. These reports have helped determine the safety practices needed to control listeriosis. Although L. monocytogenes has been reported from humans, animals, and a variety of foods in India, limited data exist with respect to prevalence and distribution of L. monocytogenes in the Indian subcontinent. The Indian Listeria Culture Collection Centre in Goa maintains all of the isolates received for subtyping and molecular characterization. Of the listerial isolate collection maintained by this center, three fourths of the isolates are of 4b serotype, while the number of other serotypes is very low. Therefore, we screened L. monocytogenes serotype 4b isolates to determine their relevance to previously defined epidemics and/or outbreaks using multi-virulence-locus sequence typing (MVLST). A total of 25 isolates in serogroup 4b of L. monocytogenes were randomly selected from a repository of 156 L. monocytogenes 4b isolates obtained from different sources in India over a period of 10 years. MVLST sequence types (virulence types, VTs) were compared to known epidemic clones and other known isolates in the L. monocytogenes MVLST database. The 25 isolates were grouped into three clusters. Cluster I comprised 21 isolates including animal (n=9), human (n=4), and food (n=8), which matched Epidemic Clone I (ECI, VT20). Three isolates-two from animal and one from food-formed a cluster while a single animal isolate was placed into two novel VTs (VT98 and VT99), respectively. Based on these findings, it can be inferred that ECI has been isolated from a variety of sources and places and has persisted in India for at least 10 years.

Genome Analysis of Pseudomonas aeruginosa Strains from Chronically Infected Patients with High Levels of Persister Formation.

The appearance of persister cells with low metabolic rates are key factors leading to antibiotic treatment failure. Such persisters are multidrug tolerant and play a key role in the recalcitrance of biofilm-based chronic infections. Here, we present the genomic analyses of three distinct Pseudomonas aeruginosa Egyptian persister-isolates recovered from chronic human infections. To calculate the persister frequencies, viable counts were determined before and after treatment with levofloxacin. The susceptibilities of isolates to different antibiotics were determined using the agar-dilution method. To determine their recalcitrance, the levofloxacin persisters were further challenged with lethal concentrations of meropenem, tobramycin, or colistin. Furthermore, the biofilm formation of the persister strains was estimated phenotypically, and they were reported to be strong biofilm-forming strains. The genotypic characterization of the persisters was performed using whole genome sequencing (WGS) followed by phylogenetic analysis and resistome profiling. Interestingly, out of the thirty-eight clinical isolates, three isolates (8%) demonstrated a persister phenotype. The three levofloxacin-persister isolates were tested for their susceptibility to selected antibiotics; all of the tested isolates were multidrug resistant (MDR). Additionally, the P. aeruginosa persisters were capable of surviving over 24 h and were not eradicated after exposure to 100X-MIC of levofloxacin. WGS for the three persisters revealed a smaller genome size compared to PAO1-genome. Resistome profiling indicated the presence of a broad collection of antibiotic-resistance genes, including genes encoding for antibiotic-modifying enzymes and efflux pump. Phylogenetic analysis indicated that the persister isolates belong to a distinct clade rather than the deposited P. aeruginosa strains in the GenBank. Conclusively, the persister isolates in our study are MDR and form a highly strong biofilm. WGS revealed a smaller genome that belongs to a distinct clade.

A Single Residue within the MCR-1 Protein Confers Anticipatory Resilience.

The envelope stress response (ESR) of Gram-negative enteric bacteria senses fluctuations in nutrient availability and environmental changes to avert damage and promote survival. It has a protective role toward antimicrobials, but direct interactions between ESR components and antibiotic resistance genes have not been demonstrated. Here, we report interactions between a central regulator of ESR viz., the two-component signal transduction system CpxRA (conjugative pilus expression), and the recently described mobile colistin resistance protein (MCR-1). Purified MCR-1 is specifically cleaved within its highly conserved periplasmic bridge element, which links its N-terminal transmembrane domain with the C-terminal active-site periplasmic domain, by the CpxRA-regulated serine endoprotease DegP. Recombinant strains harboring cleavage site mutations in MCR-1 are either protease resistant or degradation susceptible, with widely differing consequences for colistin resistance. Transfer of the gene encoding a degradation-susceptible mutant to strains that lack either DegP or its regulator CpxRA restores expression and colistin resistance. MCR-1 production in Escherichia coli imposes growth restriction in strains lacking either DegP or CpxRA, effects that are reversed by transactive expression of DegP. Excipient allosteric activation of the DegP protease specifically inhibits growth of isolates carrying mcr-1 plasmids. As CpxRA directly senses acidification, growth of strains at moderately low pH dramatically increases both MCR-1-dependent phosphoethanolamine (PEA) modification of lipid A and colistin resistance levels. Strains expressing MCR-1 are also more resistant to antimicrobial peptides and bile acids. Thus, a single residue external to its active site induces ESR activity to confer resilience in MCR-1-expressing strains to commonly encountered environmental stimuli, such as changes in acidity and antimicrobial peptides. Targeted activation of the nonessential protease DegP can lead to the elimination of transferable colistin resistance in Gram-negative bacteria. IMPORTANCE The global presence of transferable mcr genes in a wide range of Gram-negative bacteria from clinical, veterinary, food, and aquaculture environments is disconcerting. Its success as a transmissible resistance factor remains enigmatic, because its expression imposes fitness costs and imparts only moderate levels of colistin resistance. Here, we show that MCR-1 triggers regulatory components of the envelope stress response, a system that senses fluctuations in nutrient availability and environmental changes, to promote bacterial survival in low pH environments. We identify a single residue within a highly conserved structural element of mcr-1 distal to its catalytic site that modulates resistance activity and triggers the ESR. Using mutational analysis, quantitative lipid A profiling and biochemical assays, we determined that growth in low pH environments dramatically increases colistin resistance levels and promotes resistance to bile acids and antimicrobial peptides. We exploited these findings to develop a targeted approach that eliminates mcr-1 and its plasmid carriers.

Genetic diversity and risk factor analysis of drug-resistant Escherichia coli recovered from broiler chicken farms.

A total of 38 Escherichia coli isolates were recovered from 120 samples collected from various sources of broiler chicken farms (n = 10 each) in Andhra Pradesh and Telangana states. Though the recovered E. coli isolates were found variably resistant to the tested antibiotics, all the tested isolates were susceptible to meropenem. Alarming multi-drug resistance (MDR) was observed (34/38) among the recovered isolates, wherein antibiotic-resistant genes (blaTEM, blaSHV, and tetA) were detected, except for blaCTX-M-9. The heatmap with cluster analysis exhibited that majority of the E. coli isolates recovered from different sources and regions clustered together based on their phenotypic resistance suggesting co-sharing of resistance. However, the pulsed-field gel electrophoresis (PFGE) typing revealed an extremely diverse genotypic profile. Further, a significant statistical association was not observed between hypothesized risk factors and recovered MDR- E. coli isolates from various sources, although a significant statistical association between antibiotic resistance with large flock size, poor biosecurity practices, poor workers' hygiene, and poor disinfection practices was noticed. Since the study highlighted an alarming level of drug resistance among the recovered E. coli isolates, further in-depth research in similar veins is required to ensure the prudent use of antimicrobials in the poultry sector and the implementation of an antimicrobial surveillance system.

Resolving colistin resistance and heteroresistance in Enterobacter species.

Species within the Enterobacter cloacae complex (ECC) include globally important nosocomial pathogens. A three-year study of ECC in Germany identified Enterobacter xiangfangensis as the most common species (65.5%) detected, a result replicated by examining a global pool of 3246 isolates. Antibiotic resistance profiling revealed widespread resistance and heteroresistance to the antibiotic colistin and detected the mobile colistin resistance (mcr)-9 gene in 19.2% of all isolates. We show that resistance and heteroresistance properties depend on the chromosomal arnBCADTEF gene cassette whose products catalyze transfer of L-Ara4N to lipid A. Using comparative genomics, mutational analysis, and quantitative lipid A profiling we demonstrate that intrinsic lipid A modification levels are genospecies-dependent and governed by allelic variations in phoPQ and mgrB, that encode a two-component sensor-activator system and specific inhibitor peptide. By generating phoPQ chimeras and combining them with mgrB alleles, we show that interactions at the pH-sensing interface of the sensory histidine kinase phoQ dictate arnBCADTEF expression levels. To minimize therapeutic failures, we developed an assay that accurately detects colistin resistance levels for any ECC isolate.

Multi-label classification for multi-drug resistance prediction of Escherichia coli.

Antimicrobial resistance (AMR) is a global health and development threat. In particular, multi-drug resistance (MDR) is increasingly common in pathogenic bacteria. It has become a serious problem to public health, as MDR can lead to the failure of treatment of patients. MDR is typically the result of mutations and the accumulation of multiple resistance genes within a single cell. Machine learning methods have a wide range of applications for AMR prediction. However, these approaches typically focus on single drug resistance prediction and do not incorporate information on accumulating antimicrobial resistance traits over time. Thus, identifying multi-drug resistance simultaneously and rapidly remains an open challenge. In our study, we could demonstrate that multi-label classification (MLC) methods can be used to model multi-drug resistance in pathogens. Importantly, we found the ensemble of classifier chains (ECC) model achieves accurate MDR prediction and outperforms other MLC methods. Thus, our study extends the available tools for MDR prediction and paves the way for improving diagnostics of infections in patients. Furthermore, the MLC methods we introduced here would contribute to reducing the threat of antimicrobial resistance and related deaths in the future by improving the speed and accuracy of the identification of pathogens and resistance.