A 14-year-old Lusitano horse with signs of abdominal pain was referred to a referral equine hospital in Lisbon in June 2016 and diagnosed with a small intestinal inguinal hernia. Laparotomy was performed, and the incarcerated portion of non-viable small intestine was resected under general anesthesia. The horse had never been subjected to antimicrobial therapy before and was treated with penicillin at a dose of 22 000 UI/kg for 7 days and gentamicin at a dose of 6,6 mg/kg for 5 days post-surgery. The patient developed a surgical site infection of the abdominal wall that was detected on day 4 post-surgery. A swab sample was collected from the site and further streaked in Blood and MacConkey agar. After a 24 h incubation at 37°C, pure bacterial cultures were obtained, posteriorly identified as K. pneumoniae using API 20 E test (Biomérieux, Marcy-Étoile, France).

The Laboratory of Microbiology of the Faculty of Pharmacy of the University of Porto provided the seven ST348 human isolates (Rodrigues et al., 2016, 2017 and unpublished data).

The isolates susceptibility profile was determined with the Vitek 2 system, using the AST N344 card. The minimal inhibitory concentration (MIC) for ciprofloxacin was determined using E-test, following EUCAST breakpoint guidelines (The European Committee on Antimicrobial Susceptibility Testing [EUCAST], 2018).

Whole genome sequencing of all isolates was performed. DNA extraction was performed with the DNeasy UltraClean Microbial Kit (Qiagen, Hilden, Germany) and DNA concentration was measured with a Qubit 2.0 fluorometer (Life Technologies, Thermo Fisher Scientific, Waltham, MA, United States).

The DNA was then diluted to 0.2 ng/μl and 1 ng was used for the library preparation, using the Nextera XT Library Preparation kit (Illumina, CA, United States), according to the manufacturer’s protocol. DNA was sequenced using the Miseq (Illumina) generating 2 times 250-bp reads and de novo assembly was done by CLC Genomics Workbench v11.0 (Qiagen) after quality (Phred scores >30) and adapter trimming. Annotation was performed using the RAST server¹. Resistome analysis was performed by uploading sequences to the CGE server². Plasmid replicons were identified through Abricate v0.8-dev³, while the ST, wzi, K and O types and a virulence score, based on the presence or absence of different siderophores genes and a genotoxin gene, were determined through Kleborate v0.2.0⁴ and Kaptive v0.5.1⁵.

For MinION long-reads, base calling and demultiplexing was performed using Albacore v1.2.2 (ONT) and data quality was analyzed through Poretools v0.6.0 (Loman and Quinlan, 2014). Adapter trimming was attained through Porechop v0.2.3⁶. Hybrid assemblies of short- and long-reads were performed using Unicycler v0.4.1 (Wick et al., 2017). Bandage v0.8.1 (Wick et al., 2015) was used to visualize the assembly graphics. Easyfig was used to compare plasmids pK1 and pDA33141-217 (Sullivan et al., 2011). BRIG was used to align plasmid pK1 with the assemblies of the human strains (Alikhan et al., 2011).

The comparative phylogenetic analysis of the eight isolates was performed using a core genome (cg) MLST approach through Ridom SeqShere+ software (Ridom, Munster, Germany) and BacWGSTdb was used to perform a comparison with similar isolates (Ruan and Feng, 2016)⁷.

Amino-acid substitutions were investigated in the quinolone resistance determining regions (QRDR) of the gyrA, gyrB, parC, and parE genes.

The K. pneumoniae isolate obtained from the surgical site infection was resistant to the majority of the antimicrobials tested, which included different cephalosporins, fluoroquinolones and aminoglycosides (Table 1). WGS allowed the analysis of its resistome, revealing the presence of bla_CTX–M–15, among several other resistance genes (Table 2). The complete sequence of a large plasmid was obtained through a combination of long- and short-read sequencing (∼190 kb), which carried several antimicrobial resistance genes (bla_CTX–M–15, bla_TEM–1B, bla_OXA–1, sul2, aph(6)-Id, aac(6′)-Ib-cr, catB4, tet(A), dfrA14, qnrB1) and genes associated with metal tolerance (ars operon). Replicons from incompatibility groups IncFII and IncFIB were identified. A comparison between our plasmid pK1 and pDA33141-217, the closest plasmid in the Genbank database (CP029588; homology <99%, sequence cover <83%) was performed, and differences are mainly due to rearrangements in the region surrounding the bla_CTX–M–15 gene. By using a BLAST approach, a similar plasmid was also found in the human strain 122.1, but not in the other human ST348 isolates.

A comparative analysis with other MDR K. pneumoniae ST348 isolated from humans in Portugal between 2012 and 2016 was performed to evaluate a possible genetic link. These human isolates were obtained as colonizers or causing infection in individuals in long-term health care facilities and hospitals, and produced CTX-M-15 and occasionally KPC-3 (Rodrigues et al., 2016, 2017; Table 3). As in the equine isolate, the same ST capsular type, KL62 (wzi94), and O1v1 antigen, were found. In addition, all isolates had similar content on ß-lactamases genes (bla_CTX–M–15, bla_SHV–11, bla_OXA–1, and/or bla_TEM–1), with differences for isolate Kp56 that had a carbapenemase gene, bla_KPC–3, isolate 102-1 that lacked the bla_OXA–1 gene, and isolate C1685 that lacked the bla_TEM–1B gene (Table 4). Additionally other antimicrobial resistance genes were also present in all isolates [aac(6′)-Ib-cr, fosA, catB4, oqxA, and oqxB (Table 4)]. All of them carried the FIB replicon, but variably the repFII or repFIA. The isolates were also analyzed concerning their virulence score, which, according to the score system provided by Kleborate, was 1 out of 5 for the isolates that contained the yersiniabactin gene, and 0 out of 5 for the only isolate in which this gene was not detected.

The clonal relatedness was evaluated using a gene-by-gene typing method (core genome Multi-Locus Sequence Typing [cgMLST]). Allele differences between the isolates ranged from 4 to 21 in a total of 2358 genes. There were 21 allele differences between the horse isolate (K) and the closest human colonizing isolate (isolate 122-1) obtained from a long term-care facility resident (Figure 1). Using the BacWGSTdb, 56 allele differences between the horse isolate (K) with the closest isolate deposited in the database were detected. Again, this isolate corresponded to a human strain from a respiratory infection in Portugal in 2013.

Our collection of K. pneumoniae ST348 showed no mutations in the QRDR regions, except for two isolates (C1685 and Kp56) that presented a mutation in codon 83 of the gyrA gene. MIC results to ciprofloxacin ranged from 0.75 mg/L to 8 mg/L. One isolate (K107) presented high resistance to ciprofloxacin, with an MIC of >32 mg/L.

According to the Centers for Disease Control and Prevention (CDC), complications or infections secondary to either device implantation or surgery are referred to as hospital-acquired infections (Boev and Kiss, 2017). Acquisition of infections within the hospital are silent threats, especially when the source remains unknown (Walther et al., 2018). This horse developed a surgical site infection due to a MDR K. pneumoniae during hospitalization, which was therefore considered a hospital-acquired infection. We do not know if the contamination source was exogenous (i.e., surgical personnel) or intrinsic (i.e., horse’s skin) and if it was developed during, or after surgery. However, a recent study showed that a positive intra-operative culture was not a predictor of SSI; and even when a SSI did occur, it may be due to a different bacterial isolate (Isgren et al., 2019). The occurrence of ESBL-producing bacteria in horses (Johns et al., 2012; Apostolakos et al., 2017), including isolates containing the bla_CTX–M–15 gene (Leigue et al., 2018), has been shown in previous studies. The challenge of treating such infections can be overcome by the use of antiseptics (Southwood, 2009) without the need for treatment with antimicrobials. This was the case in this patient, where the infected wound was successfully treated with lavages using a saline solution containing 0.05% of chlorhexidine, performed twice a day for 1 week (Wilkins and Unverdorben, 2013).

The isolate obtained from the wound was identified as a K. pneumoniae ST348, a ST not previously reported in animals. In addition, it has rarely been identified in international collections and databases of MDR K. pneumoniae isolates from humans (Baraniak et al., 2013; Breurec et al., 2013; Ruan and Feng, 2016) or in other hosts (Holt et al., 2015), and there is still scarce surveillance data on different non-human niches. Indeed, most reports of ST348 K. pneumoniae in humans are from the Northern region of Portugal, where this lineage seems to be circulating since at least 2012 as a CTX-M-15 producer that subsequently acquired KPC-3. The isolates were obtained from community-based patients or residents in long-term care facilities with previous contact to different hospitals, suggesting a more extended dissemination in the country (Rodrigues et al., 2014, 2016, 2017; Vubil et al., 2016).

Several studies have associated quinolone resistance to a higher fitness cost and the establishment of major and minor clones in different bacterial species (Marcusson et al., 2009; Fuzi, 2016). Tóth et al. (2014) have related a high number of amino-acid substitutions in QRDR with a better fitness in vitro and in vivo (e.g., major clones) in K. pneumoniae species. Results from this study are in accordance with this, as our isolates present no mutations, except for two isolates (C1685 and Kp56) that presented only one mutation. Although these ST348 isolates are CTX-M-15 producers, we hypothesized that this clone is also a minor clone that suffered large fitness cost with the introduction of fluoroquinolones but still acquired alternative mechanisms of resistance {both efflux mediated resistance [aac(6′)-Ib-cr, oqxA, and oqxB genes] and qnr determinants}, thus being a minor clone with a better fitness adaptation. The MIC results of ciprofloxacin also support this hypothesis as our isolates present MIC’s ranging from 0.75 mg/L to 4 mg/L that are lower than those associated with major clones (Tóth et al., 2014) with two exceptions for isolates C1741 (MIC 8 mg/L), and K107 (MIC > 32 mg/L), showing that this clone has also a better adaptation to these drugs.

A cgMLST analysis, comparing the horse and human isolates, was performed to detect a possible genetic link that could provide evidence for the source of the K. pneumoniae strain (Figure 1). Recent cut-offs advocate that clonal relatedness is accepted for the K. pneumoniae isolates when allele difference are less than 10 (Schürch et al., 2018); using this cutoff we could assume that there is no genetic link between the horse and human isolates. Despite that, other authors have interpreted these cutoffs from a different perspective as they advocate that 16 to 21 allele differences between isolates from carriers from different species suggest that transmission has occurred (Adler et al., 2017; Lifshitz et al., 2018). Moreover, most of the few studies that define cutoffs for cgMLST analysis are based on human isolates only (Schürch et al., 2018) and do not take into account interspecies transmission.

A direct clinical epidemiological link between the horse isolate and the human isolates could not be found except for the year of isolation: none of the horses’ tutors had contact with hospitals or long-term care facilities in the Northern part of Portugal, where the horse stable was located, and none of the veterinarians and caretakers had been previously infected or known to be colonized with an ESBL-K. pneumoniae strain (although active surveillance was not routinely performed). Furthermore, it was not possible to establish the previous colonization of the horse, as samples are usually not collected prior to emergency surgeries.

Taking into consideration the similarities between the horse and human isolates, one can consider the possibility of a silent dissemination of K. pneumoniae ST348 amongst humans and horses in Portugal. A direct human-to-animal transmission or a common source of infection cannot be ruled out, especially since active surveillance of MDR pathogens of people in contact with the horse was not performed. The low virulence presented by these isolates and therefore the improbability of the isolation of this particular ST of K. pneumoniae, also supports the hypothesis of a silent dissemination. This is also enhanced by the results revealed by the cgMLST analysis performed with BacWGSTdb, since all the K. pneumoniae ST348 strains deposited in this database were human strains, the majority from Portugal.

Horses have different uses in today’s society, which include sports, leisure, production and consumption. It is very likely that the lack of active surveillance and characterization of antimicrobial resistant bacteria in these animals, and the proximity to man is contributing to the spread and evolution of these MDR strains. This report suggests that some K. pneumoniae clones, as the ST348, can carry critical resistance genes and disseminate them in different populations, despite their low virulence.