Background
The emergence of multi-drug resistant (MDR) bacteria is a major health problem that has been likened in its global future impact on human health to that of terrorism [
1,
2]. Widespread inappropriate use of antimicrobials in food production (especially meat/seafood, some fruit) has been linked to environmental contamination with MDR pathogens and outbreaks of MDR infections in humans, but direct cause-and-effect has often been difficult to confirm, despite the strength of the observed associations [
3‐
8]. Most food testing programs for antimicrobial resistance (AMR) have focused on specific organisms (e.g.
E. coli, Salmonella
spp., Listeria
spp.), assuming direct food-to-human pathogen transfer, rather than considering resistant gene transfer between bacterial species [
6,
8,
9]. Furthermore, the optimum site of specimen collection (e.g. on-farm animal, manure, abattoir, point-of-sale supermarket products) has been debated [
9‐
13]. Although Australia has reasonably strict regulations regarding antimicrobial use in agriculture [
13,
14], use of some agents for prophylaxis and treatment (e.g. trimethoprim-sulfamethoxazole, some beta-lactams and macrolides) is common in some food sectors [
2,
13,
15,
16], such that this may have some implications for acquisition by consumers of multi-resistant pathogens via food consumption [
6,
7].
Hence, we aimed to assess the rates of contamination with potential extended-spectrum beta-lactamase (ESBL)-producing Gram-negative organisms (without restricting to specific species) in Australian-produced chicken and pork meat. To best identify any potential risk to the consumer and to be certain that the meat was produced in Australia, we purchased chicken drumsticks and pork ribs at local fresh food outlets, since national legislation requires that bone-containing meat products must be Australian-produced (by conventional or organic production), whereas de-boned meats (e.g. bacon) can be imported into Australia [
17].
Results
Of a total of 120 meat specimens (60 CD, mean ± SD weight: 155.4 ± 26.5 [range 78.5–223.9] grams; 60 PR, 160.5 ± 48.9 [range: 91.5–355.1] grams) that were assessed from 30 retailers (see locations in Additional file
1: Figure S1), 112 (56 CD, 93%; 56 PR; 93%) were contaminated with a total of 164 (86 CD; 78 PR) 3GCR (i.e. potential ESBL-producing) isolates (Table
1). Among these isolates, 59 (36%; 26 CD, 33 PR) displayed phenotypic evidence of 3GCR alone, 96 (59%; 54 CD, 42 PR) were 3GCR plus were also resistant to either anti-folates, aminoglycosides or carbapenems and 9 isolates (5.5%; 6 CD, 3 PR; 9 specimens; 5
Pseudomonas aeruginosa, 2
Pseudomonas spp., 1
Bordetella trematum, 1
Chryseobacterium gleum) were MDR with evidence of being 3GCR plus resistance to two other antibiotic classes. Resistance to anti-folates was most common (
n = 91 [55%] isolates, 49 CD, 42 PR, Table
1; 82 [68.3%] specimens). The four most common 3GCR species identified were
Acinetobacter baumannii complex (
n = 59),
Pseudomonas aeruginosa (
n = 22),
Serratia fonticola (
n = 19) and
Hafnia alvei (
n = 15). Only one
E. coli isolate was identified – this was in a CD specimen.
Table 1
Summary of isolates grown from fresh retail chicken and pork
Acinetobacter baumannii complex | 59 | 34 | – | 34 | – | – | | | 25 | – | 25 | – | – | – | | |
Acinetobacter ursingii
| 2 | – | | | | | | | 2 | – | 2 | – | – | – | | |
Aeromonas sobria
| 1f | – | | | | | | | 1 | – | – | 1 | – | – | | |
Bordetella trematum
| 1 | – | | | | | | | 1 | – | – | – | – | – | 1 | – |
Chryseobacterium gleum
| 1 | – | | | | | | | 1 | – | – | – | – | – | – | 1 |
Citrobacter braakii
| 7 | – | | | | | | | 7b | 7b | – | – | – | – | | |
Citrobacter freundii
| 6e | – | | | | | | | 6 | 6 | – | – | – | – | | |
Citrobacter youngae
| 1e | 1 | 1 | – | – | – | | | – | | | | | | | |
Enterobacter cloacae complex | 9g | 1 | 1 | – | – | – | | | 8 | 8 | – | – | – | – | | |
Escherichia coli
| 1a | 1a | 1a | – | – | – | | | – | | | | | | | |
Hafnia alvei
| 15c | 6 | 6 | – | – | – | | | 9 | 9 | – | – | – | – | | |
Pseudomonas aeruginosa
| 22 | 11 | – | 7 | – | 4 | – | – | 11 | – | 10 | – | – | 1 | – | – |
Pseudomonas alcaligenes
| 1 | 1 | – | – | 1 | – | | | – | | | | | | | |
Pseudomonas oleovorans
| 3 | 1 | 1 | – | – | – | | | 2 | – | 2 | – | – | – | | |
Pseudomonas putida
| 11 | 10 | 1 | 1 | 8 | – | | | 1 | – | 1 | – | – | – | | |
Pseudomonas spp. | 3 | 3 | – | – | 1 | 2 | – | – | – | | | | | | | |
Serratia fonticola
| 19d | 16 | 15 | 1 | – | – | | | 3 | 3 | – | – | – | – | | |
Stenotrophomonas maltophilia
| 1 | – | | | | | | | 1 | – | – | – | 1 | – | | |
Yokenella regensburgei
| 1 | 1 | – | – | 1 | – | | | – | – | – | – | – | – | | |
Among the 164 isolates, 158 had DNA available for PCR analysis. Beta-lactamase genes were identified in 23 (15%) isolates (7CD, 14PR [2 PR each had two isolates],
p = 0.15; 17.5% specimens). All were AmpC, with 22/23 considered to be inherently chromosomally located (ACC,
n = 12 [
H. alvei, 10;
S. fonticola, 2]; CMY-like
n = 7 [
C. freundii, 6;
C. youngae/freundii, 1]; FOX,
n = 1 [
A. sobria]; MIR-like/ACT-like,
n = 2; [
E. cloacae complex]), while the sole
E. coli isolate contained a CMY-like AmpC gene that was likely to be plasmid-mediated and was subsequently shown on whole genome sequencing to be a CMY-2 (see Table
1). All DNA samples were PCR-negative for other ESBL genes (including SHV, TEM, CTX-M) and all carbapenemase encoding gene families (including IMP, VIM, KPC, OXA-48-like and NDM).
Among the 30 food outlets, there were four supermarket chains (two large [
n = 10 and 11 stores sampled]; two smaller [
n = 2 and 3 stores] and 4 separate (unlinked) butcher shops. Overall, there were no differences in rates of contamination between supermarkets and unlinked butchers shops. All supermarkets and butcher shops had at least one CD or PR specimen that grew a potential ESBL-producing isolate, at some time. Only 8 specimens were culture-negative (4 CD, 4 PR; one supermarket site had both its PR specimens culture-negative). The numbers of 3GCR isolates per specimen were as follows: single isolate in 63 specimens; two isolates in 44 specimens; 3 isolates in 3 specimens, and one specimen contained 4 potential ESBL-producing isolates. Interestingly, it was this latter specimen (which was collected from a butcher’s shop) that grew the CMY-2-containing
E. coli, along with an
A. baumannii,
S. fonticola and an
E. cloacae complex isolate – although none of these latter 3 isolates contained any definable ESBL genes (Table
1).
Discussion
This study of Australian chicken and pork is notable for a number of reasons. Firstly, we assessed for a broad range of Gram-negative organisms, not simply the traditional species of
E. coli or Salmonella
spp. [
6,
8‐
10]. Taking this approach, we identified that 93% of specimens appeared to be contaminated with a wide variety of 3GCR species, including particularly
Acinetobacter baumannii complex,
Pseudomonas aeruginosa,
Serratia fonticola and
Hafnia alvei. We were surprised by the relatively high rates of these potential pathogens and initially speculated that perhaps they were due to a point-source within certain supermarkets or butcher shops, such as has been reported in one outbreak of multidrug-resistant
K. pneumoniae [
35]. However, they were identified from both CD and PR products purchased from a wide variety of food outlets which had no common supply chain. Notably, only one
E. coli isolate was identified – so testing programs which only assess for this species would have reported a much lower rate of potential contamination.
Secondly, our results highlight the importance of not relying solely on selective media such as ChromID ESBL agar in such programs, but instead confirming the presence of ESBL genes by PCR. Phenotypic detection methods alone may identify intrinsically 3GCR isolates or those that falsely suggest ESBL production [
24‐
26]. AmpC genes were identified in 15% isolates assessed (17.5% specimens), with most (22/23) being inherently chromosomal in location [
26]. Notably, however, the sole
E. coli isolate identified contained a plasmid-mediated CMY-2 which was potentially transferable.
The fact that resistance to anti-folate agents was the most common resistance phenotype identified among potential ESBL-producing strains and was noted in 68.3% of all CD/PR specimens is important, given that trimethoprim-sulfamethoxazole is widely used in pork production and some chicken farms [
15,
16]. Hence these results may be no surprise, but at least serve as a potential “wake up call” to farmers who are concerned about the consequences of frequent antibiotic use. Importantly, 9 isolates (7.5% CD/PR specimens) displayed an MDR phenotype, with only one strain (
Bordetella trematum) being resistant to fluoroquinolones – consistent with Australia’s strict controls on fluoroquinolone use in agriculture and similar to previous studies on this issue [
14].
Acinetobacter,
Serratia,
Hafnia and
Pseudomonas spp. are all known to be common in the environment and to be present on some fruit and vegetables [
36,
37], but their presence may be a potential source of resistance genes [
38].
Given the uncertainty about which testing regimen would be ideally suited for a large national food safety screening program for MDR contamination [
9‐
13], we believe our methodology was a practical approach that is potentially relevant and meaningful to retail consumers and which could be up-scaled without the need for major infrastructure or specialised training. In comparison, all previous published Australian studies have assessed non-meat items such as animal faeces or eggs [
39‐
42].
Our findings differ from those by other authors. Overdevest et al. [
8] reported that 79.8% of retail chicken meat samples in the Netherlands had organisms with ESBL genes present, while only 1.8% of pork samples grew an ESBL-producing organism. However, this study focused particularly on
E.
coli and
K.
pneumoniae without commenting on other organisms isolated. Stewardson et al. [
7] reported 86% contamination of chicken meat products delivered to a tertiary hospital in Switzerland with ESBL-producing
Enterobacteriaceae species. Similar to our results, MDR strains were uncommon.
This study has some limitations. Firstly, the sample size of 120 specimens, while consistent with similar studies, is relatively small in the context of overall Australian supply [
2,
7,
8,
43‐
47]. Secondly, we were not able to track the original farm source of the CD and PR products, although one might expect larger supermarket chains to have a limited number of defined contracted suppliers. Further research to investigate the rates of contamination at each step of the meat production process, including samples from animals in farms, carcasses and meat products in slaughterhouses and of meat products distributed to third party organisations for packaging and distribution, may be helpful to identify if there is a common source of contamination. Thirdly, our sample preparation (including initial 24 h culture in non-selective media), the subsequent selective culturing techniques provided enhanced sensitivity for 3GCR-GNBs but did not allow us to accurately quantify the burden of contamination in each CD/PR sample. Importantly, we did not assess for phenotypic colistin resistance since laboratory methods are evolving [
48,
49], nor did we assess for
mcr genes since this resistance mechanism was only first reported in 2016 [
50]. Notably, colistin resistance appears to be currently rare in Australia [
51,
52] and colistin is infrequently used in Australian agriculture [
2,
53]. Finally, Australia does not import fresh chicken meat, nor any fresh bone-containing pork products [
13,
17], which means that all of our specimens came from animals born and grown in Australia. As such we cannot comment on any possible difference in contamination between these Australian products and similar, but boned, imported chicken and pork processed meat products.
We believe our findings raise important questions regarding future food testing programs and potentially highlight the importance of routine public health measures related to safe food preparation such as appropriate hand hygiene before/after handling uncooked meat products, adequate washing of kitchen utensils and surfaces that have contact with uncooked meat and appropriate cooking methods to ensure destruction of any contaminating bacteria. These public health messages may be of particular importance to patient groups where immunosuppression is likely, such as those with haematological malignancy or transplant recipients. Further research into the potential source(s) of retail meat contamination is warranted.