Klebsiella pneumoniae as a key trafﬁcker of drug resistance genes from environmental to clinically important bacteria

known for its high frequency and diversity of antimicrobial resistance (AMR) genes. In addition to being a signiﬁcant clinical problem in its own right, K. pneumoniae is the species within which several new AMR genes were ﬁrst discovered before spreading to other pathogens (e


Antimicrobial resistance in Gram negative opportunistic pathogens
The antimicrobial resistance (AMR) crisis facing hospitals globally is driven by the ESKAPE pathogens (Gram positives Enterococcus faecium, Staphylococcus aureus; and Gram negatives Klebsiella pneumoniae, Acinetobacter baumannii, Pseudomonas aeruginosa, Enterobacter), which are responsible for the majority of infections in hospital patients that are difficult to manage with antimicrobial therapy [1].Notably the ESKAPE pathogens are environmental or commensal bacteria that cause opportunistic infections in hospitalised or immunocompromised patients, but are generally not pathogenic otherwise.Each of these species has intrinsic resistance to one or more antibiotics, and individual strains have accumulated resistance to many additional drugs [1].The Gram negative ESKAPE pathogens are considered the greatest threat, due to the emergence of strains that are resistant to all or most available antibiotics [2 ].Accumulation of AMR in these organisms is primarily due to horizontal gene transfer (HGT) aided by plasmids and mobile genetic elements [1].The catalogue of known mobile AMR genes subject to HGT amongst Gram negative pathogens numbers in the hundreds [3].The origins of the AMR genes themselves are environmental bacteria (particularly soil bacteria), assumed to be those which have co-evolved with the relevant antimicrobial producing organisms for millennia [4][5][6]; however there is typically a lag of several years between the clinical use of a drug and the arrival of relevant mobile AMR genes in human pathogen populations [7].Hundreds of mobile AMR genes have been found in K. pneumoniae [8,9], the species associated with the earliest reports of many AMR genes before their dispersal amongst other clinically relevant Gram negatives.Here we discuss this phenomenon in detail, and then explore what is currently known about K. pneumoniae ecology and its genome plasticity, arguing that these characteristics position the species as a key amplifier and spreader of AMR genes from environmental sources to human pathogen populations.
The canary in the coalmine K. pneumoniae are intrinsically resistant to ampicillin due to the presence of the SHV-1 penicillinase in their chromosome [8,10].Resistance to additional drugs occasionally arises through chromosomal mutations [11], however most AMR in K. pneumoniae results from acquisition of AMR genes via HGT, mainly via large conjugative plasmids [9,11,12].The accumulation of resistance determinants in a single strain can result in panresistant strains that are untreatable with all available antibiotics [2 ].The earliest mobile ampicillin resistance genes identified in Gram negative bacterial populations were TEM (present in the first described plasmids in the 1960s), and the K. pneumoniae chromosomal SHV-1 gene which was first detected in mobile, plasmidborne form in other Enterobacteriaceae in 1973 [13,14] (Figure 1).Following the introduction of third generation cephalosporins for clinical use in the early 1980s, extended spectrum beta-lactamase (ESBL) genes conferring resistance to these drugs began to be detected and characterised.The earliest forms include ESBL variants of mobile SHV (SHV-2; 1985) [15], TEM (1984) and CMY (1989) [16], which were first identified in K. pneumoniae (Figure 1) and are now widespread amongst Enterobacteriaceae [17], and in some cases have also spread to Acinetobacter [18] and Pseudomonas [19].The most widely dispersed ESBL gene is CTX-M, variants of which were detected in Escherichia coli and K. pneumoniae in the late 1980s and early 1990s, having been mobilised out of environmental Enterobacteriaceae (Kluyvera) [20,21].CTX-M is now intimately associated with the E. coli ST131 pandemic clone [22] and several K. pneumoniae clones [11], and is present in diverse plasmid backgrounds, resulting in broad dissemination amongst hospital, human commensal, and animal associated microbial populations [22,23].
The 1990s introduction of carbapenems and fluoroquinolones were met by rapid appearance of associated resistance genes, with K. pneumoniae often playing a key role (Figure 1).Mobile quinolone resistance genes qnrA and qnrB were first identified in K. pneumoniae [24,25], following mobilisation from marine bacterium Shewanella [26], and are now common amongst Enterobacteriaceae plasmids.The K. pneumoniae carbapenemase (KPC) appeared in the mid-1990s in the USA and drove the spread of pandemic hospital outbreak clone K. pneumoniae ST258, which is now globally disseminated [22].The KPC gene has been transferred to many different plasmids, is now widely dispersed amongst Enterobacteriaceae and has also found its way into Pseudomonas [27] and Acinetobacter [28].The OXA-48 carbapenemase originates from Shewanella [29] and was first detected in K. pneumoniae in Turkey in 2003 [30].It was initially associated with hospital outbreaks across Europe and is now reported worldwide [31], although not as widely dispersed as KPC.The NDM-1 metallo-beta-lactamase was first detected in 2008 in K. pneumoniae from a patient who had recently travelled to India [32].The gene was plasmid-borne and shortly thereafter was reported in different K. pneumoniae strains isolated from patients with and without recent travel [33]; by 2010 NDM-1 had spread to numerous plasmids and Enterobacteriaceae species, was detected within the chromosome of E. coli and Providencia stuartii [34], and was spreading amongst Acinetobacter and Pseudomonas [35,36].The first mobile colistin resistance gene MCR-1 was reported in China in 2015 in E. coli and K. pneumoniae [37 ]; by 2016 it had been detected across five continents, among Enterobacter and numerous other species, and in association with over a dozen distinct plasmids [38 ].
It is impossible to accurately reconstruct the precise flow of genes, plasmids and bacteria involved in the capture of AMR genes from environmental microbes and their dissemination among human-associated bacterial pathogen populations, although qnrA and OXA-48 provide compelling examples of AMR gene mobilisation from marine bacteria to K. pneumoniae, and onwards to other ESKAPE pathogens.Regardless of precise HGT flow, the dominance of K. pneumoniae amongst early clinical reports of new AMR genes is notable, and indicates K. pneumoniae to be a prime target for sentinel surveillance of new AMR genes entering Gram negative pathogen populations.

Means and opportunity: genome plasticity, plasmid diversity and ecology
Figure 2a shows the total number of distinct acquired AMR genes and load per strain for all genome sequences of K. pneumoniae and fellow Gram negative opportunists (A.baumannii, P. aeruginosa, E. cloaceae and E. coli) currently in the NCBI Pathogen Genomes portal (>1400 genomes for each species).Over 400 acquired AMR genes are present in the K. pneumoniae genomes, double the number in E. coli and 50% more than that of the other species (Figure 2a), suggesting that K. pneumoniae Timeline of mobile AMR genes first detected in Klebsiella pneumoniae.Shading indicates the period since which isolates of K. pneumoniae resistant to each drug class have been reported (regardless of mechanism).Selected mobile AMR genes that were first detected in K. pneumoniae are labelled on the timeline, within the row corresponding to the relevant class; all have since been reported in clinically important Enterobacteriaceae and other Gram negative bacteria.Note ampicillin resistance is intrinsic to K. pneumoniae due to the chromosomal betalactamase gene SHV-1, and this gene was shown to be mobilised by plasmids in E. coli and K. pneumoniae in the 1970s.The other genes shown did not originate in K. pneumoniae, but they were first detected in mobile form (i.e.within mobile genetic elements on plasmids) in K. pneumoniae isolates, as detailed in the 'The canary in the coalmine' section.
receives and/or amplifies a wider range of AMR genes from their ultimate source in environmental microbes [5,6].Below we consider the genomic and ecological characteristics of K. pneumoniae in comparison to the other ESKAPE pathogens and E. coli, highlighting factors that may enhance exposure to environmental AMR genes and the ability to pass these genes on to other clinically important pathogens.

Ecological range
Most HGT occurs between cells residing in the same habitat [39], hence bacteria adapted to survive in environmental and animal/human-associated microbial communities can be predicted to contribute most to the trafficking of AMR genes between these niches.Reports in the 1970s-1980s highlighted the ubiquitous distribution of K. pneumoniae among diverse fresh and salt water environments, plants, and soil [40].K. pneumoniae causes infections in cows, horses and other wild and domestic animals [8,41,42,43 ,44-48].However as an opportunistic pathogen, it is likely that K. pneumoniae is more often a component of the normal animal gut microbiota.In humans the rate of intestinal K. pneumoniae colonisation has been estimated at $6% [49 ,50], while in dairy cows the rate may be much higher ($44% among herds in New York, USA, [51]).K. pneumoniae has also been cultured from the faeces of other agricultural and domestic animals, from the cloacae of birds, and from fish, shellfish, insects and earthworms [42,48,[52][53][54][55][56].It is a common contaminant of animal and plant-based foods, which likely plays a key role in introducing environmental strains into the human gut [48].
Enterobacteriaceae are well known gut colonisers, and all the ESKAPE pathogens can be isolated from environmental sources [57], however systematic comparisons of isolation rates across environmental and animal sources are lacking.We used the IMNGS website [58 ] to query the growing body of publicly available 16S taxonomic profiling data for the presence of Gram negative opportunists across a wide range of sample types (Figure 3a).These data confirm the three Enterobacteriaceae species are frequent commensals of humans and animals, with E. coli the most commonly detected; however K. pneumoniae and E. cloaceae showed equivalent or even greater prevalence amongst plant and environmental samples (7-14%), whereas E. coli was significantly less common in these niches (3-4%, p = 0.01).Notably these data indicate that K. pneumoniae and E. cloaceae have similarly broad ecological distributions, but K. pneumoniae was more prevalent in human and other animal microbiomes (6.8% versus 3.6%), likely increasing its exposure to antimicrobial use and its contact with other clinically important pathogens, therefore enhancing amplification and onward dissemination of acquired AMR genes.
It is difficult to directly assess movement of individual K. pneumoniae strains between niches, however there is evidence that isolates from human, animal and environmental sources do not represent distinct subpopulations 'feline', 'dog', 'canine', 'cow', 'bovine', 'pig', 'porcine', 'water', 'aquatic' ('horse' and 'equine' were also searched but yielded no reports of globally distributed AMR STs); additional STs were compiled from two genomic studies which had each reported >10 isolates from non-human sources [8,64 ]. [8,42,48,[59][60][61].Clinically important K. pneumoniae lineages have been isolated from specific non-human sources (Figure 3b); for example, ESBL ST15 in cats and dogs [59][60][61], and hypervirulent ST23 or ST25 in horses, non-human primates and pigs [43 ,62,63].Genomic comparisons of K. pneumoniae from diverse sources are rare but show little evidence of segregation between niches: in a global diversity study, 59 bovine isolates were distributed around the species phylogeny comprising mostly human isolates [8]; and a comparison of ESBL isolates from Thai hospitals and a local canal system indicated phylogenetic intermingling [64 ].These data and others [48] confirm that at least some strains, including those recognised as globally distributed hospital pathogens [11], can move between and proliferate in multiple niches, providing opportunity for genetic exchange with a broad range of bacterial species (Figure 3b).

Genome composition and HGT
K. pneumoniae genomes are highly diverse [8], comprising hundreds of distinct phylogenetic lineages that differ from each other by $0.5% nucleotide divergence.Individual strains harbour $2000 'core' (shared) genes, plus a further $3500 accessory genes that differ between strains and are drawn from a large pool of >30 000 [8].The $2000 core genes likely facilitate K. pneumoniae's broad ecological range by providing metabolic and other capabilities enabling survival in a wide range of niches.A substantial proportion of the total pan-genome (core + accessory genes, [65]) is predicted to encode proteins with metabolic functions; 19% associated with carbohydrate metabolism, 18% with other metabolic pathways and 13% with membrane transport [8].This extensive diversity results in variable metabolic capacity [66 ], potentially supplementing individual strains with additional ecological range and providing even more opportunities for genetic exchange.Direct comparison of population structures are difficult due to different sampling and analysis strategies [65]; nonetheless it is clear that the other Gram negative opportunists also have many deep branching lineages and large pan-genomes [8,[67][68][69][70].
Coding capacity and genome size are easy to compare using public genome data (Figure 2c): K. pneumoniae has a significantly larger genome than the other Enterobacteriaceae species considered here (mean 5.7 Mbp, 5455 protein coding genes, versus 5.1 Mbp/4915 genes in E. coli and 5.0 Mbp/4680 genes in E. cloacae; p < 1 Â 10 À15 using two-sided t-test), which may help equip K. pneumoniae for survival in a broader range of niches.
DNA base composition varies widely between taxa, and can be used as a signature of bacterial species [71].The mean G + C content of K. pneumoniae core genes is 58%, whereas that of accessory genes ranges from 20% to >70%, suggesting they originate from a taxonomically diverse array of donors [8]. Figure 2d shows that genes annotated in complete genomes of K. pneumoniae display significantly more variability in their G + C content than those of the other species considered here, with 50% greater variance in G + C content than E. coli and E. cloaceae ( p < 1 Â 10 À15 using F-test or the non-parametric Fligner-Killeen test) and more than double the variance of P. aeruginosa and A. baumannii.

Plasmid load
The vast majority of AMR genes in K. pneumoniae are plasmid-borne [9,11], hence the ability to amplify and spread AMR genes across ecological niches is likely linked to plasmid-permissive traits.Highly diverse environmental microbial communities, especially soils, are considered hotspots for gene transfer [75], and Enterobacteriaceae have been identified as a component of the 'super-permissive community' that supports the spread of plasmids across diverse soil communities [76].The specific role of K. pneumoniae in such activities remains to be explored, however the species has been associated with hundreds of distinct plasmids spanning many plasmid replicon types [8,9,12,77], which suggests it acts as a recipient for plasmids originating from a wide array of HGT donors.The median number of plasmids per complete K. pneumoniae genome currently in NCBI GenBank is three (interquartile range, 2-5; range 0-10), significantly higher than that for the other species of interest ( p < 1 Â 10 À5 , see Figure 2b).This is consistent with numerous reports of K. pneumoniae strains carrying multiple AMR plasmids; for example, the ST11 reference genome, HS11286, harbours six plasmids (1.3-123 kbp in size), three of which carry AMR genes [78]; and there are many examples in other lineages [9,79,80 ].
Its elevated plasmid load (Figure 2b) suggests K. pneumoniae is particularly permissive for plasmids, meaning it may be more likely to capture plasmid-borne material from diverse donors in varied niches, and to hold on to this material long enough to transmit it to new recipients in human and animal-associated niches (Figure 4).This enhanced permissiveness may reflect a comparatively lower fitness burden of plasmid carriage in K. pneumoniae;  [8,9], suggesting that there may be significant variation in plasmid permissiveness between lineages.Related species and strains can vary substantially in their ability to act as plasmid donors [76,85], however variation in plasmid-donor potential between K. pneumoniae and other bacteria, or between strains of K. pneumoniae, remains to be investigated.

Conclusions
K. pneumoniae have the means and opportunity to capture plasmids from environmental microbial populations; to survive within and move between multiple environmental and animal-associated niches; to maintain AMR plasmids for prolonged periods; and to pass plasmids on to other clinically important Gram negative bacteria (Figure 4).Whilst the contribution of K. pneumoniae to the AMR crisis is impossible to quantify, the available evidence suggests it is unique amongst the Gram negative ESKAPE pathogens and E. coli in a few key respects including its high diversity of acquired AMR genes, high plasmid load, wide variability of G + C content reflecting diverse HGT partners and broad ecological rangewhilst systematic studies of comparative ecology are lacking, the available 16S data suggests K. pneumoniae is equally likely to be found living in human, animal and environmental niches.Combined these factors may position K. pneumoniae as a key amplifier and spreader of clinically important AMR genes.Better understanding and monitoring of this highway of AMR gene transfer could potentially help limit the spread of AMR and prolong the life of new antibiotics.Model for AMR gene and plasmid trafficking by K. pneumoniae.Individual K. pneumoniae strains can move between niches in the environment, human and/or animal hosts, carrying with them acquired AMR genes and/or plasmids.Strains can move from the environment to human/animal hosts via contact or consumption of contaminated water sources or plant matter; between human and animal hosts via contact or consumption; and from hosts back to the environment via effluent or sewerage.K. pneumoniae strains can receive or donate plasmids via HGT with a diverse array of donor species in any of these niches, providing a pathway for transfer of AMR genes from environmental microbes to human pathogens.

Figure 4 K
Figure 4