Epidemiological Prevalence of Phenotypical Resistances and Mobilised Colistin Resistance in Avian Commensal and Pathogenic E. coli from Denmark, France, The Netherlands, and the UK

Colistin has been used for the treatment of non-invasive gastrointestinal infections caused by avian pathogenic E. coli (APEC). The discovery of mobilised colistin resistance (mcr) in E. coli has instigated a One Health approach to minimise colistin use and the spread of resistance. The aim of this study was to compare colistin susceptibility of APECs (collected from Denmark n = 25 and France n = 39) versus commensal E. coli (collected from the Netherlands n = 51 and the UK n = 60), alongside genetic (mcr-1–5) and phenotypic resistance against six other antimicrobial classes (aminoglycosides, cephalosporins, fluoroquinolones, penicillins, sulphonamides/trimethoprim, tetracyclines). Minimum inhibitory concentration (MIC) values were determined using a broth microdilution method (EUCAST guidelines), and phenotypic resistance was determined using disk diffusion. Colistin MIC values of APEC were significantly lower than those for commensals by 1 dilution (p < 0.0001, Anderson-Darling test), and differences in distributions were observed between countries. No isolate carried mcr-1–5. Three phenotypically resistant isolates were identified in 2/62 APEC and 1/111 commensal isolates. Gentamicin or gentamicin–ceftriaxone co-resistance was observed in two of these isolates. This study showed a low prevalence of phenotypic colistin resistance, with no apparent difference in colistin resistance between commensal E. coli strains and APEC strains.


Introduction
Escherichia coli (E. coli) is a commensal organism in the intestinal microbiota of many animals, including chickens [1]. Some strains may cause opportunistic infections, while others contain specific virulence factors that allow them to cause primary disease [2]. Isolates associated with clinical symptoms in birds are often termed avian pathogenic E. coli (APEC). Antimicrobial drugs (AMDs), including colistin, have been used to reduce the bacterial burden and risk of disease with colistin entering clinical use shortly after its discovery in 1947 [3]. Colistin has been used for the treatment of non-invasive gastrointestinal infections caused by E. coli and APEC, and across Europe, it is approved for administration
When analysed according to APEC (Denmark and France) vs. commensal (Netherlands and UK) status, a significant difference (two-sample Anderson-Darling test; p < 0.0001) was observed in the MIC distributions, with a higher mode at 0.25 mg/L for commensal than for the APEC mode at 0.125 mg/L ( Figure 3A). Of the three low-susceptible/resistant isolates, 2 (of 62) were APEC isolates (3.2%), and 1 (of 111) was commensal (0.9%), showing a non-significant difference ( Figure 3B).  No isolates tested positive for colistin resistance genes mcr-1, 2, 3, 4, or 5.

Phenotypic (Disk Diffusion) Susceptibility to Other AMDs in Conjunction with Colistin Resistance
For the 175 E. coli isolates received (APEC n = 64: Denmark n = 25 (yolk sac infection), France n = 39 (clinical infection); commensal n = 111: the Netherlands n = 51 (caecal sample), the UK n = 60 (faecal sample)) phenotypic susceptibility was determined for ampicillin, ceftriaxone, enrofloxacin, gentamicin, tetracycline and trimethoprim/sulfamethoxazole (1:19)  with a disk diffusion assay; inhibition diameter distributions are shown in Supplementary  Figures S1-S6. In total, 60 resistant isolates were identified for ampicillin ( Figure 4A), 5 resistant isolates for ceftriaxone ( Figure 4B), 3 resistant isolates for enrofloxacin ( Figure 4C), 27 resistant isolates for gentamicin ( Figure 4D), 57 resistant for tetracycline ( Figure 4E), and 29 resistant for trimethoprim/sulfamethoxazole ( Figure 4F). Significant differences in the relative frequency of resistant isolates were observed between countries for ampicillin, ceftriaxone, tetracycline, and trimethoprim/sulfamethoxazole. No significant difference was observed in the relative proportions of phenotypically resistant and commensal isolates between APEC and commensal E. coli strains, except for tetracycline where the proportion of resistance was higher in APECs (Fisher's exact test; p < 0.001), with 53% resistant in APEC and 22.5% in commensal ( Figure 5A-F). There was a higher proportion of tetracycline resistance in the avian pathogenic E. coli vs. commensals. Fisher's exact test: *** p ≤ 0.001, ns = not significant.

Phenotypic Resistance Patterns and Multi-Drug Resistance (MDR)
Overall, Denmark had the highest proportion of isolates showing no resistance (15 of 24; 62.5%), followed by the UK (33 of 60; 55%) and the Netherlands (25 of 51; 49%). France had the lowest proportion of non-resistant isolates at 25.6% (10 of 39). The remaining isolates showed resistance, to a single, double, or multiple (≥3) AMDs. French isolates showed a significant difference (Fisher's exact test, p < 0.001) from the other countries, with higher proportions of isolates showing co-resistance to two antimicrobials (20.5%) and MDR (35.9%), whilst Danish isolates showed the lowest proportion of MDR, at 0% ( Figure 6A). No difference was observed in the proportions of isolates in the different groups between APEC and commensal E. coli ( Figure 6B). Of the three previously identified colistin-resistant isolates, one isolate (from the UK) showed no phenotypic resistance to any of the other antimicrobials tested. Co-resistance was observed in the remaining two isolates, with the Danish isolate having co-resistance to gentamicin and the French isolate being MDR to colistin, gentamicin, and ceftriaxone. The various resistance patterns are shown in Table 1.

Discussion
The observed colistin MICs in this study may be considered in line with the EUCAST distribution (ECOFF 2mg/L) with the ECOFF in this study distribution (determined by ECOFFinder) at 2 mg/L, although having a different overall distribution and high variations between countries. Furthermore, the MIC distribution between APEC and commensal isolates was significantly different, with the APEC isolates having a lower mode. APEC is considered a highly diverse population, with studies showing a variety of virulence factors and resistant traits, including tetracyclines, sulphonamides, fluoroquinolones, and polymyxins [2,22]. As such, these disease-causing isolates may be assumed to have a higher phenotypic resistance to colistin than the wild-type commensal isolates, but the lower distribution of MICs for APEC isolates runs counter to this hypothesis.
In this study, a low proportion of phenotypic colistin-resistant E. coli isolates was observed in samples obtained from poultry farms in Denmark (4.2%), France (2.63%), the Netherlands (0%), and the UK (1.67%). This study is limited by the small number of isolates tested and the small number of resistant isolates observed, but no significant differences in proportions of colistin-resistant and susceptible isolates were found between countries or between APEC and commensal isolates. Nevertheless, it should be noted that the manner of isolation, e.g., collected from farms, at slaughter, from meat, from faecal material, or (in the case of APEC strains) from infected birds, may affect the number of resistant strains observed. For example, it has been indicated that meat samples have a lower prevalence of colistin resistance than faecal or animal samples [23]. The results of our study, although confounded by different ways of isolation with the APEC isolates from Denmark and France collected from diseased birds, the commensal isolates from the Netherlands obtained from caecal samples at slaughter and the commensal UK isolates from faecal (environmental) samples are consistent with the findings of other studies. Low prevalence of colistin resistance in European countries, taken from faecal or caecal samples, was also shown by Ceccarelli et al. [24] across nine European countries (Belgium, Bulgaria, Denmark, France, Germany, Italy, the Netherlands, Poland, and Spain), Perrin-Guyomard et al. [20] who investigated French isolates, and the surveillance reports for the countries in this study [10][11][12][13].
The low levels of prevalence in Europe, and specifically in the countries within this study, are likely a reflection of the consumption of colistin, in which the use of all antimicrobials for growth promotion purposes was banned in 2003 (European Parliament and Council Regulation 1831/2003/EC). In 2012, it was estimated that 545.2 tonnes of polymyxins (predominantly colistin, as other polymyxins are not approved since there are no minimum residue limits) were consumed by food-producing animals in Europe [25]. In 2015, the EMA reported 561.4 tonnes of polymyxins sold for food-producing animals in 29 European (EU and Switzerland) countries [8]. In 2016, advice from the antimicrobial advice expert group (AMEG) of the EMA prescribed all countries should reduce polymyxin use to prevent the spread of antimicrobial resistance (EMA/CVMP /CHMP/231573/2016). A declining trend in polymyxin usage has resulted in a drop to 210 tonnes by 2018 (31 EU countries median: 1.50 mg/PCU; range: 0 to 12.8 mg/PCU; PCU = population correction unit accounting for animal population and weight), although there was considerable variation between countries [8]. This continued decrease in colistin use has seen polymyxin sales between 2011 and 2018 drop by 69.8%. In 2018 colistin sales, as a percentage of total antimicrobial sales, accounted for <0.1%, 2.8%, 0.65%, and <0.1% for Denmark, France, the Netherlands, and the UK, respectively [8].
For comparison, reports have shown that China has a higher prevalence (14-18%) of colistin resistance in E. coli isolated from chicken samples [26,27]. This has been attributed to the fact that China is the largest colistin consumer globally, using between 2470 and 2875 metric tonnes annually (between 2011 and 2015) in animal farming [28]. Following a ban on colistin use as a growth promoter in 2017, reductions in both the production and sale of colistin in China have been recorded, and in line with this decline, a reduction from 18.1% (in 2015-2016) to 5% (in 2017-2018) in the prevalence of colistin-resistant E. coli amongst farms was observed [27]. This study also indicated that the co-harbouring of resistance genes for other antimicrobial classes may maintain colistin resistance, even in the absence of colistin use. Similarly, the withdrawal of colistin as a growth promoter in Japan in 2017 resulted in a decrease in plasmid-mediated colistin resistance in pigs by 52.5% in the succeeding 12 months [29]. This indicates a correlation between the use of colistin as a growth promoter and the selection and/or maintenance of colistin-resistant E. coli within chicken populations. Likewise, Ahmed et al. [30] demonstrated a positive correlation between the use of colistin in poultry and the frequency of MCR determination. The low consumption in European countries, in combination with strict animal husbandry procedures and regulations, may be a factor in the low prevalence of colistin resistance in poultry E. coli observed in this study.
The occurrence of MCR in commensal poultry E. coli populations differs significantly between countries in Europe, with 11.9% in Romania [19], 5.7% in Germany, 0.7% in the Netherlands, 0.3% in France, 0% in Denmark, 0% in Italy, and 0% in the UK [31]. In this study, three isolates were identified as being resistant to colistin, although no mcr genes were detected amongst all isolates. Luo et al. (2017) explored the prevalence of both chromosomal and plasmid-mediated resistance (47.5% and 52.5%, respectively), showing a similar prevalence between the two mechanisms in colistin-resistant E. coli. As no mcr genes were identified in this study it is assumed that a chromosomally mediated mechanism is responsible for resistance in these isolates. Previous studies have demonstrated that MCR typically confers a fitness cost on the harbouring bacteria, with mcr-1 positive E. coli having a lower relative fitness, compared with non-mcr-1-expressing E. coli, and that overexpression can lead to cell death [32]. Furthermore, a fitness cost is generally associated with resistance carriage and has been demonstrated for other mcr genes, coupled with a reduction in virulence [33,34], although this fitness cost may be mediated by other co-mutations [35]. Reduced fitness may reflect the absence of mcr genes in samples collected in European countries where the colistin use, and thus selective pressure, is low, and mcr-positive E. coli are less likely to be maintained in the environment.
In the absence of colistin use, other antimicrobial agents may be used to maintain animal health and welfare. The co-harbouring of resistance genes for other antimicrobials can lead to the continued selection/maintenance of MCR within the population (Wang et al. 2020b). To assess this aspect, our study was extended to investigate other classes of antimicrobials-namely, penicillins (ampicillin), cephalosporins (ceftriaxone), fluoroquinolones (enrofloxacin), aminoglycosides (gentamicin), tetracyclines (tetracycline), and sulphonamides (trimethoprim/sulfamethoxazole)-to identify the occurrence of multidrug resistance with colistin. A high proportion of isolates were resistant to ampicillin (34.5%), tetracycline (32.7%), and trimethoprim/sulfamethoxazole (17.1%), whilst the proportion of resistance was low for gentamicin (4.09%), ceftriaxone (2.34%), and enrofloxacin (1.75%) in this study. This finding was similarly reported by Roth et al. (2019), except for fluoroquinolones, which were reported with a high prevalence, in broilers in five European countries (Poland, United Kingdom, Germany, France, and Spain). The proportion of resistance correlated with the quantity of antimicrobial consumed, with the highest consumption being reported for penicillins, tetracyclines, sulphonamides, and trimethoprim (ESVAC, 2020). Aminoglycosides also had high consumption, but the proportion of gentamicin resistance was low; this may be attributed to higher use in cattle and pigs compared with poultry. Despite the low resistance to gentamicin in this study, two isolates (one from Denmark and one from France) showed co-resistance for colistin and gentamicin. Several studies have reported co-resistance to colistin and gentamicin in P. aeruginosa [36], K. pneumoniae [37], and E. coli [38] isolated from human samples; in poultry, co-resistance has been reported in E. coli [30,39,40]. The selection of colistin-gentamicin co-resistance is often related to the consecutive or concomitant administration of these antimicrobials, such as in cases of cystic fibrosis in man. This may suggest that the use of both colistin and gentamicin in animal populations selects the co-carriage of these resistance genes. In both of the aforementioned examples, E. coli isolates were isolated in Bangladesh, and although they were from separate studies, and there was no indication of animal to human spread, the isolation of co-resistance in animal reservoirs requires careful monitoring. Isolates were also identified with co-resistance to tetracycline and ceftriaxone, and as with gentamicin, the presence of isolates co-harbouring resistance may promote the maintenance of these populations in the absence of colistin use where the other antimicrobials continue to be used [41]. In fact, the most common plasmids (IncI2, IncH2, IncX4) carrying mcr-1 have also been shown to carry multiple resistance genes, be highly transferable, and confer phenotypic resistance to multiple AMDs including tetracycline, ceftriaxone, and gentamicin [42,43]. Furthermore, mcr has been associated with the transposable genetic element ISApl1 in these three plasmids, indicating the potential for transposition of mcr to plasmids carrying other resistance elements [44].
The differences in relative proportions of resistant isolates between APEC and commensal E. coli presented in this study showed no significant association, likely due to the low sample size. Additionally, the APEC strains originated from Denmark and France, whilst the commensal strains originated from the Netherlands and the UK. This geographical separation may indicate a country-specific effect rather than an APEC-commensal effect, and further study would be required to clarify the origin of this difference, which cannot be identified within the methods reported in this article. Moreover, the distinction between APEC and commensal E. coli is questionable. Indeed, a review by Collingwood et al. (2014) states that classifying any disease-causing E. coli as APEC is problematic, as a large portion of infection caused by E. coli in chickens is opportunistic, meaning a potentially large portion of E. coli classified as APEC could in fact be opportunistic commensal E. coli. A further definition may be obtained through serotyping isolates or exploring pathogenicity factors associated with APEC.

Sample Origin, Acquisition, and Storage
A total of 175 E. coli isolates were received, from which 111 were considered commensals (Netherlands n = 51, UK n = 60), and 64 were considered avian pathogenic E. coli Received isolates were recovered from glycerol storage by overnight incubation in Mueller-Hinton broth (MHB; Oxoid, Basingstoke, UK) at 37 • C, 130 rpm. Overnight cultures were streaked onto MacConkey agar (Oxoid, Basingstoke, UK). After the first MacConkey streak, plates were visually examined to confirm colony morphology representative of E. coli, and single colonies were sub-cultured onto Mueller-Hinton agar (MHA; OXOID, UK). All isolates were stored in 25% glycerol:MHB at −80 • C for long-term storage (secondary confirmation as E. coli by multiplex PCR of 16s RNA marker during mcr screening, see infra).
Minimum inhibitory concentration (MIC) measurements were performed for all isolates, using the broth microdilution method according to the European Committee for Antimicrobial Testing (EUCAST) guidelines and in accordance with ISO-20776, including two control isolates (mcr-1 negative; NCTC 12241 and mcr-1 positive; NCTC 13846) (EUCAST, 2020). Final plate concentrations ranged from 0.016 to 8 mg/L.
A two-fold dilution series (0.125-64 mg/L) was prepared in cation-adjusted Mueller-Hinton broth (CAMHB; Oxoid, Basingstoke, UK) using a semi-automated pipetting system (ViaFlo Assist, Integra Biosciences, Thatcham, UK). Bacterial suspensions were prepared from individual colonies suspended in PBS with comparison to 0.5 McFarland standard using Densicheck plus (BioMerieux, Basingstoke, UK). Dilution of this suspension was carried out with CAMHB, to achieve a final, in-plate, inoculum of 5 × 10 5 CFU/mL, with confirmation of inoculum size performed in parallel. The MIC was recorded following overnight static incubation at 37 • C. Two E. coli control isolates were included in each plate. MIC within one dilution of the expected range, to account for possible variation in MIC measurement, confirmed the validity of the results.

Multiplex PCR Screening for mcr
Isolates were screened for the presence of mcr genes (mcr-1, mcr-2, mcr-3, mcr-4, and mcr-5) using a multiplex screening method as described by Rebelo et al. [45]. E. coli-specific (16S rRNA) primers described by Le Devendec et al. [46] were used as a control, and as a secondary confirmation for E. coli identification, in each reaction. Briefly, the reaction parameters were as follows: The reaction mixture consisted of 12.5 µL DreamTaq green PCR master mix (Fisher Scientific, Basingstoke, UK), 5.5 µL nuclease-free water (Fisher Scientific, UK), 0.5 µL of each of the 12 primers (10 µM) and 2 µL of DNA template. Thermal lysis of 1 mL of overnight culture at 100 • C for 5 min, followed by centrifugation at 16,000× g, provided the DNA lysate. The thermal cycler (Techne, UK) conditions were as follows: 15 min denaturation at 94 • C, 25 cycles of 30 s at 94 • C, 90 s at 58 • C, 60 s at 72 • C, and final elongation at 72 • C for 10 min.

Disk Diffusion Testing for Antimicrobial Susceptibility
Disk diffusion antimicrobial susceptibility testing was performed to assess MDR, as described by EUCAST (EUCAST, 2021). Briefly, 90 mm circular MHA plates were prepared and stored at 4-8 • C for no more than 7 days prior to testing. The inoculum was prepared from a single colony selected from an MHA plate, the same colony used for MIC testing, and resuspended in saline to an optical density equivalent to 0.5 McFarland (Thermo Scientific, Basingstoke, UK). The inoculum was applied to the MHA agar plate and pre-dried within 15 min of preparation, to remove residual moisture, and even coverage over the entire surface area was then ensured. Antimicrobial discs were applied using an antimicrobial susceptibility disk dispenser (Oxoid, Basingstoke, UK) within 15 min of inoculation. Antibiotics tested, representing six different antimicrobial families, were ampicillin, tetracycline, enrofloxacin, ceftriaxone, gentamicin, and trimethoprim/sulfamethoxazole (Oxoid™, UK). Incubation was performed statically at 37 • C overnight (18 ± 2 h). Zone of inhibition was measured using digital callipers (Thermo Scientific, Basingstoke, UK), and classification of susceptibility was determined with EUCAST or CLSI (when data were not available using EUCAST) cut-offs (Appendix B).

Statistical Methods and Analysis
Antimicrobial susceptibility distributions, as determined by MIC for colistin and disk diffusion for ampicillin, tetracycline, enrofloxacin, ceftriaxone, gentamicin, and trimethoprim/sulfamethoxazole, were compared through a 2-sample Anderson-Darling test. Assessment of study distribution was performed using the ECOFFinder [47]. Proportions of susceptible and resistant isolates were compared between countries using Fisher's exact test.

Conclusions
In conclusion, although the levels of colistin resistance recorded in this study were low, and no mcr-positive isolates were identified, multiple, diverse resistant patterns were observed amongst isolates in this study, including the co-carriage of resistance with colistin. Whether or not the use of colistin may indirectly lead to the selection of other antimicrobialresistant genes, or vice versa, as well as whether or not the use of other antimicrobials may select or maintain colistin resistance in animal reservoirs in the absence of colistin use remains to be investigated. Although reductions in colistin usage appear to have a positive impact on the prevalence of colistin resistance, phenotypically resistant isolates persist, indicating that chromosomal resistance may provide stable reservoirs of resistance.
The associations between colistin resistance and the APEC vs. commensal status of E. coli are not directly apparent; the APEC MIC distributions were actually lower, but the data are confounded by low sample sizes, country-specific differences, and the definition of APEC in sampling.
Overall, colistin use and resistance is a global concern and requires continued surveillance and management from a consortium of countries to maintain low prevalence. The dose regimen for colistin use in food animals should be based on PK/PD criteria, as described in the literature [48], to avoid exposure of the intestinal flora to sub-inhibitory concentrations, likely to select for resistance factors.

Data Availability Statement:
The data presented in this study are available on request from the corresponding author.

Conflicts of Interest:
The funding for this study was provided through TransPharm on behalf of Dopharma, VMD Livestock pharma, and Virbac. P.R. (TransPharm) was involved in the study design and revision of the submitted article. The funders were not involved in the collection, analysis, and interpretation of data; the writing of this article; or the decision to submit it for publication. A.M.'s PhD stipend was covered by the Bloomsbury studentship (see Funding). The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

Appendix A
To prepare colistin stock, colistin sulphate was weighed (mg) and dissolved in 10 mL of cation-adjusted MHB (CAMHB). Colistin solution was filter-sterilised using a 0.22 µm syringe filter and concentration-adjusted by the proportion of colistin base (78.53%) according to Equation (A1). Colistin solution was then diluted with sterile CAMHB to achieve a final working stock at 16 mg/L colistin base. The stock solution was prepared immediately prior to use. Equation (A1): Concentration calculation to adjust for the potency of colistin sulphate (* Percentage colistin base calculated from the certificate of analysis for each batch).