General enquiries on this form should be made to:Defra, Science Directorate, Management Support and Finance Team,Telephone No. 020 7238 1612E-mail: research.competitions@defra.gsi.gov.uk

SID 5 Research Project Final Report

SID 5 (2/05) Page 1 of 25

NoteIn line with the Freedom of Information Act 2000, Defra aims to place the results of its completed research projects in the public domain wherever possible. The SID 5 (Research Project Final Report) is designed to capture the information on the results and outputs of Defra-funded research in a format that is easily publishable through the Defra website. A SID 5 must be completed for all projects.

A SID 5A form must be completed where a project is paid on a monthly basis or against quarterly invoices. No SID 5A is required where payments are made at milestone points. When a SID 5A is required, no SID 5 form will be accepted without the accompanying SID 5A.

This form is in Word format and the boxes may be expanded or reduced, as appropriate.

ACCESS TO INFORMATIONThe information collected on this form will be stored electronically and may be sent to any part of Defra, or to individual researchers or organisations outside Defra for the purposes of reviewing the project. Defra may also disclose the information to any outside organisation acting as an agent authorised by Defra to process final research reports on its behalf. Defra intends to publish this form on its website, unless there are strong reasons not to, which fully comply with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.Defra may be required to release information, including personal data and commercial information, on request under the Environmental Information Regulations or the Freedom of Information Act 2000. However, Defra will not permit any unwarranted breach of confidentiality or act in contravention of its obligations under the Data Protection Act 1998. Defra or its appointed agents may use the name, address or other details on your form to contact you in connection with occasional customer research aimed at improving the processes through which Defra works with its contractors.

Project identification

1. Defra Project code OD2010

2. Project title

Use and abuse of non-antibiotic antimicrobials as major contributors toward the development of antimicrobial resistance

3. Contractororganisation(s)

Veterinary Laboratories AgencyWoodham LaneNew HawAddlestoneSurreyKT15 3NB

54. Total Defra project costs £ 433,926

5. Project: start date................ 01 April 2004

end date................. 30 March 2007

SID 5 (2/05) Page 2 of 25

6. It is Defra’s intention to publish this form. Please confirm your agreement to do so...................................................................................YES NO (a) When preparing SID 5s contractors should bear in mind that Defra intends that they be made public. They

should be written in a clear and concise manner and represent a full account of the research project which someone not closely associated with the project can follow.Defra recognises that in a small minority of cases there may be information, such as intellectual property or commercially confidential data, used in or generated by the research project, which should not be disclosed. In these cases, such information should be detailed in a separate annex (not to be published) so that the SID 5 can be placed in the public domain. Where it is impossible to complete the Final Report without including references to any sensitive or confidential data, the information should be included and section (b) completed. NB: only in exceptional circumstances will Defra expect contractors to give a "No" answer.In all cases, reasons for withholding information must be fully in line with exemptions under the Environmental Information Regulations or the Freedom of Information Act 2000.

(b) If you have answered NO, please explain why the Final report should not be released into public domainThe journals in which we wish to publish will view the final report as prior publication and we will be unable to publish our data. As soon as the manuscripts are accepted for publication, which we anticipate to be in the near future, we will inform Defra so that the final report can be released in full to the public domain.

Executive Summary7. The executive summary must not exceed 2 sides in total of A4 and should be understandable to the

intelligent non-scientist. It should cover the main objectives, methods and findings of the research, together with any other significant events and options for new work.

Defra is responsible for reducing the risks to human health of zoonotic infection by control of pathogens in food animals and through the food chain. Zoonotic bacteria are becoming increasingly antibiotic resistant and increasingly prevalent, in both animals and humans. There are several mechanisms by which bacteria can become antibiotic-resistant and of concern in the context of this study is multiple antibiotic resistance (MAR) mediated by chromosomal genes (loci) giving rise to reduced susceptibility (~ 4 to 8) to several unrelated antibiotic classes (fluoroquinolones, ß-lactams, tetracycline’s, chloramphenicol, trimethoprim). Recent studies at VLA and Birmingham have shown that exposure of pathogens to sub-lethal concentrations of a variety of antimicrobials including disinfectants (e.g. Triclosan and phenolics) can readily select Multiple Antibiotic Resistance (MAR) and gyrA (quinolone and fluoroquinolone-resistant) mutants.

Changes in EU legislation regarding the removal from use of various antibiotics as growth promoters is associated with increasing reliance upon cleansing and disinfection (C&D) regimes and evidence is available that shows many farms use C&D regimes inappropriately such that pathogens persist in ‘cleansed’ premises. Exposure to, and survival against, sub-lethal disinfection will lead to the selection of Certain MAR phenotypes are associated with changes, often point mutations, within chromosomal genes that modulate expression of effector genes such as marRAB of E. coli and Salmonella Typhimurium that alter expression of efflux pump proteins and outer membrane proteins (OMPs). Such mutations may cause decreased permeability toward unrelated antibiotics. We have shown the incidence of MAR phenotypes amongst Salmonella enterica serotypes, especially those strains with reduced susceptibility to FQ’s (fluoroquinolones), as assessed by resistance to cyclohexane is approaching ~20% of strains. One cause of emergence of MAR mutants may be inappropriate use of disinfectants.

We argue that it is essential to understand the possible role of disinfection in the selection of MAR isolates in the farm environment and to assess relative fitness of emergent MAR mutants. In this report we describe two main avenues of research. First, genetic analyses of the basis of resistance selected by disinfectant treatment. Second, an evaluation of the biology of sensitive and resistant pathogens in response to the environmental pressures they might face on farm, in particular their survival in the environment and their colonisation potential and persistence in the animal host.

A single passage of eight Salmonella enterica serovar Typhimurium strains in the presence of sub-lethal concentrations of farm disinfectants gave rise to mutants with decreased susceptibility to disinfectants and/or antibiotics. The disinfectants used in the studies included a tar oil phenol (TOP or FFS) disinfectant, an oxidising compound disinfectant (OXC or VS), a compound disinfectant containing a mix of Quaternary ammonium compounds, formaldehyde and glutaraldehyde (QACFG, SK or ABD) and a dairy

SID 5 (2/05) Page 3 of 25

sterilizer disinfectant (DSD, based on a quaternary ammonium biocide) in agar. All disinfectants gave rise to mutants and the majority of mutants recovered after disinfectant exposure showed reduced susceptibility to the selective agent compared with the parent strain and required statistically significantly longer exposure times to disinfectants to generate a 5 log kill (a standard measure of disinfectant efficacy). Small decreases in antibiotic susceptibility were observed also. Notably, however, exposure to QACFG decreased susceptibility to ciprofloxacin in some strains. None of these mutants derived by single step mutation, however, demonstrated the typical MAR phenotype.

To understand the molecular basis of decreased susceptibility to disinfectants in strains grown with disinfectants, the protein content of the bacterial cell wall and cytoplasm (the so-called proteome) was analysed using two-dimensional liquid chromatography mass spectrometry. After exposure to TOP and OXC disinfectants, significant increases in expression of the AcrAB-TolC efflux pump system were revealed. To verify these findings, test strains were engineered to have either acrB or tolC inactivated and these strains, as anticipated, were more susceptible to disinfectants than other strains proving conclusively the role of efflux pumps in protecting bacteria from disinfectants (to be reported elsewhere; see publication listing).

Another very important question regarding these single step mutants with increased disinfectant tolerance was the impact, if any, on the biology of the bacterium. Studies were performed which showed conclusively that single step mutants made in this study were able to survive and persist in chicks as well as parent strains. Thus, the acquisition of reduced disinfectant susceptibility did not attenuate virulence (to be reported elsewhere; see publication listing).

Data from these single step studies demonstrated that efflux pumps are required for intrinsic resistance to some disinfectants and that exposure to disinfectants can induce expression of the AcrAB-TolC efflux system. Such strains remain as pathogenic as their parent strains

The studies described above focused on a single exposure to a disinfectant. However, it was important to assess the impact of multiple or extended exposure, as might happen on farm with multiple rounds of cleaning, because such exposure may give rise to mutants, possibly with different behaviours than those described above. To investigate this, Salmonella enterica serotype Typhimurium strain SL1344 was exposed to sub-lethal concentrations of three widely-used farm disinfectants in serial daily passage for seven days. Stable isolates OXCR1, QACFGR2 and TOPR2 were obtained following treatment with an oxidising compound blend (OXC), a quaternary ammonium disinfectant containing formaldehyde and glutaraldehyde (QACFG) and a tar acids-based disinfectant TOP respectively. All isolates demonstrated reduced susceptibility to multiple antibiotics, including chloramphenicol, tetracycline and ampicillin which is indicative of a MAR phenotype which were not recovered by single step exposure.

To understand the molecular basis of resistance in these multiply exposed strains, the proteomes of the cell wall and cytoplasm of a number of derivatives were analysed as described above. All derivatives tested possessed reduced levels of outer membrane proteins and for two derivatives designated OXCR1 and QACFGR2 high levels of efflux pump AcrAB-TolC proteins were demonstrated. Further analysis of cytoplasmic protein profiles showed all derivatives over-expressed a range of proteins associated with a protective responses to oxidants, nitro-aromatics, disulphides and peroxides. Furthermore, it was deduced that there were likely to be metabolic changes, suggesting increased protein synthesis and a shift from oxidative phosphorylation (reduced levels of ATP synthase) to substrate level phosphorylation. Other changes in protein expression were detected in the isolates and these correlated with their altered phenotypes.

It was important to understand whether these multiply exposed mutants retained a similar phenotype to their parents. A range of studies showed that all derivatives were impaired for growth in both simple and complex media and were less motile. Furthermore, their colony morphology was smooth and possibly associated with the loss of modal distribution of O-antigen chain length of the lipopolysaccharides, a major component of the cell wall. In addition, when tested in the chick model, the mutants were less invasive and less persistent indicating that these mutants posed less threat to both animals and humans.

Triclosan is becoming a widely used antibacterial agent. Upon exposure to triclosan, resistant mutants were readily selected from a panel of Salmonella enterica serovar Typhimurium strains. Three distinct resistance phenotypes were observed: low (LoT), medium (MeT), and high-level (HiT) requiring 4-8 g/ml, 16-32 g/ml and >32 g/ml of triclosan respectively for inhibition. The bacterial target of triclosan is fatty acid biosynthesis and specifically the enzyme FabI, encoded by the gene called fabI. Genetic analysis of fabI and proteomic studies were performed on LoT, MeT and HiT mutants. Some mutants, as anticipated possessed substitutions within FabI thereby creating resistance although these did not correlate with the triclosan inhibitory concentrations. Artificial over-expression of fabI and re-creation of a glycine to valine substitution at codon 93 of FabI in SL1344 resulted in low level triclosan resistance suggesting that high-level triclosan resistance is not mediated by FabI alone.

Relating these studies to the work on efflux, we showed that inactivation of acrB and tolC in triclosan resistant mutants led to large decreases in triclosan resistance. Proteomic analysis of triclosan resistant

SID 5 (2/05) Page 4 of 25

mutants identified a set of proteins with differential expression in all mutants but also demonstrated specific patterns of expression for each phenotype including efflux pumps. Exemplars of each phenotype were evaluated for fitness in-vitro and in day old chicks were found to colonise and persist in chickens throughout a 28 day competitive index experiment (i.e. birds were dosed with the mutant and parent together to assess if either out-competed the other but in these studies the mutants were as virulent as the parent). These data show that triclosan resistance can occur via a number of distinct pathways in salmonella. In addition we demonstrated the expression of a novel generic set of proteins for which a triclosan resistance network may be hypothesised were modulated. We also showed efflux is required for intrinsic and high-level triclosan resistance and that triclosan resistant mutants of all classes are fit in-vivo.

This study clearly points to the problem that ineffective cleansing and disinfection will lead to the emergence of bacterial pathogens with a range of altered phenotypes. Single step mutants show reduced tolerance to antibiotics and disinfectants but these strains remained fully pathogenic. These may be progenitor MAR strains which is a cause for concern and further study is required to assess onward mutation. Triclosan that is becoming ever more popular will select similar efflux based mutants again with a range of phenotypes. These too are fully pathogenic. Again there is a need to assess onward mutation especially toward MAR. Continued exposure to inappropriate levels of disinfectants will eventually attenuate pathogens whereby they are less resilient under non selective conditions but will still persist in disinfectants. However, seven cycles of sub-lethal disinfectant treatment is an artificial scenario that demonstrates a point. Key then, is appropriate use of disinfectants and avoidance in C&D regimes of permitting niches populated with bacteria that are only partially treated as his will lead to the derivation of resistant mutants.

Advice for cleansing and disinfection on farms and in food production: Prevent the use of single mode of action biocides (such as triclosan). Promote both rotation of different disinfectants and of those with multiple mechanisms of action. Ensure that disinfectants are applied at full strength to all areas which have been previously

cleaned of organic material which may reduce their activity. Determine the fitness of mutants isolated from the field.

Project Report to Defra8. As a guide this report should be no longer than 20 sides of A4. This report is to provide Defra with

details of the outputs of the research project for internal purposes; to meet the terms of the contract; and to allow Defra to publish details of the outputs to meet Environmental Information Regulation or Freedom of Information obligations. This short report to Defra does not preclude contractors from also seeking to publish a full, formal scientific report/paper in an appropriate scientific or other journal/publication. Indeed, Defra actively encourages such publications as part of the contract terms. The report to Defra should include: the scientific objectives as set out in the contract; the extent to which the objectives set out in the contract have been met; details of methods used and the results obtained, including statistical analysis (if appropriate); a discussion of the results and their reliability; the main implications of the findings; possible future work; and any action resulting from the research (e.g. IP, Knowledge Transfer).

INTRODUCTION Defra is responsible for reducing the risks to human health of zoonotic infection by control and reduction of such pathogens in food animals and through the food chain. Zoonoses are becoming increasingly antibiotic resistant and may be increasing in prevalence, both in animals and humans. There are several mechanisms by which bacteria can become antibiotic-resistant including acquisition of transferable resistance genes. Of increasing concern is multiple antibiotic resistance mediated by chromosomal genes (loci) that can give rise to reduced susceptibility (~4 to 8x) to several unrelated antibiotic classes (fluoroquinolones, ß-lactams, tetracycline’s, chloramphenicol, trimethoprim). Recent studies at VLA and Birmingham have shown that exposure of pathogens to sub-lethal concentrations of a variety of antimicrobials including disinfectants (e.g. Triclosan and phenolics) can readily select Multiple Antibiotic Resistance (MAR) and gyrA (quinolone and fluoroquinolone-resistant) mutants. Changes in EU legislation regarding the removal from use of various antibiotics as growth promoters is associated with increasing reliance upon cleansing and disinfection regimes and evidence is available that shows

SID 5 (2/05) Page 5 of 25

many farms use C&D regimes inappropriately and that pathogens persist in ‘cleansed’ premises. Exposure to and survival against sub-lethal disinfection will lead to the selection of resistant pathogens that will pose a serious threat to human health.Certain MAR phenotypes are associated with changes, often point mutations, within chromosomal genes that modulate expression of effector genes such as marRAB of E. coli and Salmonella Typhimurium that alter expression of efflux pump proteins and outer membrane proteins (OMPs), causing decreased permeability toward unrelated antibiotics. We have shown the incidence of MAR phenotypes amongst Salmonella enterica serotypes as assessed by resistance to cyclohexane is ~20% for a panel of c. 450 strains examined. One cause of emergence of MAR mutants may be inappropriate use of disinfectants. We argue that it is essential to understand the possible role of disinfection in the selection of MAR isolates in the farm environment and to assess relative fitness of emergent MAR mutants. In this report we describe two main avenues of research. First, genetic analyses of the basis of resistance selected by disinfectant treatment. Second, an evaluation of the biology of sensitive and resistant pathogens in response to the environmental pressures they might face on farm, in particular their survival in the environment and their colonisation potential and persistence in the animal host.

Because this work was carried out at three different institutions, different abbreviations (mainly in the appendices) have been used for some of the different disinfectants and for the bacterial mutants derived from growth with the disinfectants.

The five disinfectants/biocides which are widely used on farms and in food production were:-

1. A compound disinfectant containing a mix of Quaternary ammonium compounds, formaldehyde and glutaraldehyde. The abbreviations used for this disinfectant are QACFG, SK or ABD (aldehyde based disinfectant).

2. An oxidising compound based disinfectant with the abbreviations OXC or VS. 3. A tar oil phenol disinfectant with the abbreviations TOP or FFS.4. A dairy sterilizing disinfectant which comprised a quaternary ammonium biocide, non-ionic surfactant and

excipients. The abbreviations used this disinfectant are DSD.5. Triclosan has the abbreviation T or TRIC.

OBJECTIVES

Objective 01: Bacterial isolates will be made from farm premises pre and post cleansing and disinfection. Current farm practice with regard to the use of non-antibiotic biocides for cleansing and disinfection will established. Correlates between the methods used and deduced efficacy by examination of sentinel E. coli strains from enlisted farm premises (poultry and pigs) from current Defra-funded projects will be made.Objective 02: To determine the rate of acquisition of biocide resistance under various C&D regimes and to assess the environmental fitness of mutants that arise from these regimes.Objectives 03: To determine the in vivo fitness (using the poultry model) of biocide resistant mutants derived from objective 02 above. Objective 04: To undertake a detailed analysis at the phenotypic and genotypic level of selected biocide resistant mutants to assess the basis of resistance and the impact that has upon environmental survival. Objective 05: To write the final report that draws together the conclusion of the study and relate these to our starting hypothesis that improper of C&D regimes results in the emergence and survival of resistant organism that are biologically fit. The report will aim to provide advice to Defra on possible strategies to improve use of C&D regimes. These outputs will be available as a foundation data for fuller risk analyses.

ALL OBJECTIVES WERE MET ON TIME AND WITHIN BUDGET

This report will focus on three broad areas of study outlined in objectives 02-04. We have reported previously on our findings relating to objective 01 (Randall et al., 2005; Webber, 2005) and these are already in the public domain (see appendices) for which brief details of these studies are provided below.

Tables are given in appendix (A.1)Figures are given in appendix (A.2)References from this section are given in appendix (A.3)

Objective 01/

SID 5 (2/05) Page 6 of 25

Objective 01/01 Collect E. coli form poultry farms and assess current farm practice on use of cleansing and disinfection regimes.E. coli strains of porcine (n=128) and poultry (n=261) origin were obtained from healthy animals on a total of 43 farms in the UK. Strains were identified as presumptive E. coli on the basis of typical colony colour and morphology on Chromagar (Chromagar Microbiology) and their identity was confirmed for a representative number of isolates using API 20E strips (Bio-Merieux) according to the manufacturers instructions. Representative isolates of E. coli were also sero-grouped by serum agglutination (Sojka, 1965).Disinfectant resistance was reviewed and a chapter was prepared and published in ‘Antimicrobial Resistance in Bacteria of Animal Origin’ (Webber et al., 2005) that included sections on disinfectant use in the veterinary industry, mechanisms of action and resistance and potential for selection of biocide resistance. The article concluded that whilst there is evidence that use of disinfection in the home (Barker et al, 2003; Cozad et al, 2003) prevents infections, it is likely that many of the products available for domestic use contain triclosan, QACs or other antimicrobial substances that are not contributing to improved hygiene and frivolous use of antimicrobial compounds should be discontinued – there should be a clear benefit associated with the use of such products.

Objective 01/02 Analysis of MAR phenotypes amongst sentinel E. coli isolates from existing pig and poultry farm studies.The E. coli panel was evaluated for those with a MAR phenotype using cyclohexane tolerance and for reduced susceptibility to certain antibiotics. Organic solvent (cyclohexane and hexane) tolerance was determined as previously described (Randall et al. 2001). Briefly, strains were grown overnight at 37C in Luria Bertani (LB) broth and inoculated (1 l) onto LB agar in glass plates. Inoculum spots were allowed to dry before plates were flooded with 6 ml of cyclohexane or hexane. Flooded plates were wrapped with cling film and incubated overnight at 30C. Colony spots that grew in the presence of organic solvents were deemed tolerant. MICs of ampicillin, ciprofloxacin, chloramphenicol, tetracycline were determined using the BSAC agar doubling dilution method (Andrews 2001).Cyclohexane tolerance was observed in 10 / 389 strains although for 9 of these strains only limited tolerance (very limited growth in the presence of cyclohexane, e.g. only a few cyclohexane tolerant colonies from an inoclum spot of c. 106 cfu ml-1) was observed. This one cyclohexane tolerant strain also had reduced susceptibility to ampicillin, ciprofloxacin, chloramphenicol and tetracycline consistent with a MAR phenotype. Hexane tolerance was observed in 352 / 389 farm strains although for 63 of these strains only limited tolerance (very limited growth in the presence of hexane) was observed. None were MAR.

Objective 01/03 E. coli isolates analysed by MIC and genetic methods to determine their phenotype and related genotype.Due to the very small number of E. coli isolated with a MAR phenotype (n=1) the ability of three common farm disinfectants to select for ciprofloxacin resistant or cyclohexane tolerant E. coli was investigated. Details of these studies are provided below in the abstract produced under objective 01/04. The disinfectants were not able to select for ciprofloxacin resistant E. coli in poultry slurry but higher numbers of cyclohexane tolerant strains were isolated that had a MAR phenotype. In view of the fact that only one cyclohexane tolerant E. coli strain was isolated from farms, it would appear that the risk from farm disinfectants selecting for E. coli MAR strains is outweighed by usefulness of disinfectants. Of interest is the higher proportion of Salmonella MAR isolates reported previously (Randall et al, 2004) suggesting different mechanisms for MAR between these species.

Objective 01/04 Draft report and paper on association between phenotypes and genotypes.An abstract of a paper produced under this objective is provide below (Randall et al., 2005)

Aims: The aim of this study was to determine if three classes of farm disinfectants were able to select for ciprofloxacin resistant or cyclohexane tolerant (indicative of a multiple antibiotic resistance (MAR) phenotype) E. coli and if cyclohexane tolerant E. coli could be isolated from farms.Methods and Results: Chicken slurry containing c. 1:99 ratio ciprofloxacin resistant: susceptible E. coli (10 different resistant strains examined) was treated for 24 h with each of the disinfectants and examined for survival of resistant: susceptible strains. Ciprofloxacin sensitive (n=5) and resistant (n=5) E. coli were grown with sub-lethal concentrations of the disinfectants and then plated to agar containing ciprofloxacin or overlaid with cyclohexane. E. coli (n=389) isolated from farms were tested for cyclohexane tolerance. MICs were determined against representative isolates and mutants. The disinfectants did not select for the ciprofloxacin resistant E. coli in poultry slurry but following growth with each of the three disinfectants, higher numbers (P 0.023) of cyclohexane tolerant E. coli were isolated and these had a MAR phenotype. Of the 389 farm E. coli tested, only one was cyclohexane tolerant. Conclusions: It is possible that in a farm environment, E. coli could be exposed to similar concentrations of the disinfectants that selected for MAR type organisms under these laboratory conditions.

Objectives 02-04

SID 5 (2/05) Page 7 of 25

Single Step mutation to resistance to disinfectants

INTRODUCTIONIn two recent studies we demonstrated that growth of Salmonella enterica and Escherichia coli with sub-inhibitory concentrations of farm disinfectants led to a small but statistically significant increase in isolation of multiply antibiotic resistant (MAR) strains which were detected when subsequently exposed to cyclohexane or antibiotics (1,2). In E. coli and S. enterica serovar Typhimurium (hereafter referred to as S. Typhimurium) MAR is associated with reduced susceptibility to antibiotics such as -lactams, chloramphenicol, fluoroquinolones and tetracyclines, increased tolerance to organic solvents and decreased susceptibility to disinfectants such as pine oil (3-5). In E. coli over-expression of acrAB, marRAB and soxRS genes can all lead to MAR (2). Although the mechanisms of MAR in S. enterica are not so clearly defined as for E. coli, the same resistance phenotype is seen in isolates from farms and farm animals (4).In recent years there have been increasing concerns that disinfectant exposure may help drive selection of antibiotic resistant bacteria (6,7). However, disinfectants are often a mixture of active compounds and as a result usually have multiple, intracellular targets, therefore it has been proposed that disinfectant resistance is unlikely to arise from a single mutational event as can occur with antibiotic resistance (8). Previous studies have demonstrated that over-expression of multi drug xenobiotic efflux transporters (e.g. AcrAB-TolC) with broad substrate specificity can confer decreased resistance to antibiotics, dyes, disinfectants and detergents (9). The selection of MAR is a concern as such strains may be able to act as a stepping stone to high-level resistance in E. coli and S. enterica (1,4-5).In our previous studies however (1,2), there were essentially two selection stages, the first being growth in sub-inhibitory concentrations of disinfectants and the second being plating the strains on either media with 4x MIC of antibiotics or media overlaid with cyclohexane. We were not able to isolate strains with reduced susceptibility to the farm disinfectants themselves, nor were we able to isolate MAR mutants after exposure to farm disinfectants, without the additional exposure to either antibiotics or cyclohexane. Whilst in a farm environment it is likely that bacteria will be exposed to both antibiotics and disinfectants and that this exposure may be persistent, so we were interested in whether limited exposure to farm disinfectants alone could select antibiotic resistant strains.In section of the report, we describe the selection strains of S. Typhimurium from agar containing four different farm disinfectants and a comparison of the phenotype of mutant and parent strains with respect to efflux, antibiotic and disinfectant sensitivity, growth in the presence of sub-inhibitory concentrations of disinfectants, protein expression and the ability to survive and persist in chicks.

MATERIALS AND METHODSStrains, media and chemicals. All strains used in this study are listed in Table 1 and were routinely cultured on Luria-Bertani agar and in Luria-Bertani broth unless stated otherwise. All chemicals were obtained from Sigma-Aldrich (Poole, Dorset U.K) except for ciprofloxacin, which was kindly donated by Bayer (Newbury, Berkshire, UK). The farm disinfectants used were a tar oil phenol (TOP or FFS) which was a blend of high boiling point tar acids and organic acids, an oxidising compound based disinfectant (OXC or VS), a compound disinfectants containing quaternary ammonium compounds, formaldehyde and glutaraldehyde (QACFG, SK or ABD) and a dairy sterilizing disinfectant (DSD) which comprised a quaternary ammonium biocide, non-ionic surfactant and excipients. Farm disinfectants were obtained from appropriate suppliers. The triclosan (Irgasan DP300) was kindly donated by Ciba Consumer care (Macclesfield, Cheshire).Selection of mutants after single exposure to disinfectants. Mutants were selected as previously described (10). Parent strains were grown overnight in antibiotic-free broth, concentrated by centrifugation, and

resuspended in sterile broth to give a range of inocula (106 to 1010 CFU/ml). Agar plates containing disinfectants at 2x the MIC were inoculated with 100μl (105 to 109 CFU) of each cell suspension and incubated at 37°C in air for up to 7 days. Ten colonies with the typical size and morphology of the original strain were chosen randomly from each selecting plate and subcultured onto disinfectant-free media and retained for further study of the mechanism of resistance.Calculation of disinfectant exposure times required for a 5-log kill. Disinfectant activity was quantified by determining the ability of a disinfectant to reduce the viable numbers of a suspension of bacteria, the decimal reduction time (D-value) is the time required for a specified concentration of disinfectant to lead to a one log (90%) reduction in viable numbers (11). In this study the ability of each disinfectant used to kill mutants with decreased susceptibility was determined using an assay based on the European suspension test which requires a 5-log (99.999%) kill (EN1040; Donald Morrison, personal communication). Initial experiments defined the concentrations of each disinfectant, to give a 5-log reduction in viable numbers of bacteria within 30 min. These conditions were then used to compare the time required for a 5-log kill for mutant and parent strains. For each strain an overnight culture in LB broth was diluted to approximately 5 x 108 cfu/ml, 1 ml of this suspension was then added to 8 ml of test disinfectant solution and 1 ml of sterile water. At appropriate time intervals 0.5 ml aliquots were removed and added to universal tubes containing 0.5 ml of sterile water and 4.5 ml of neutralising buffer. Each suspension created in this way was serially diluted in sterile water and aliquots plated onto LB agar plates, which were incubated overnight at 37˚C before colony numbers were recorded. Viable counts were performed in parallel from the original cultures used in order to allow enumeration of the numbers of cells killed

SID 5 (2/05) Page 8 of 25

after disinfectant treatment. Viable count data was also used to correct for differences in sizes of the original inocula within each experiment in order to allow comparison of data from parent and mutant strains. Data were analysed using Excel (Microsoft, USA) to calculate means and standard deviations for each strain. Differences between strains were analysed for statistical significance using the Student’s t test.MICs and cyclohexane resistance. Organic solvent (cyclohexane and hexane) tolerance was determined as previously described (4). MIC testing of antibiotics and farm disinfectants used the BSAC agar doubling dilution method (12).Cell envelope analysis. The expression of major outer membrane proteins and lipopolysaccharide was investigated as described previously (13) and patterns compared between parent and mutant strains. All strains were also visualised microscopically in order to detect any filamentation or other gross morphological changes.Sequencing of quinolone resistance determining regions of gyrA, gyrB, parC and parE. The quinolone resistance determining regions of the QACFG mutants with cross resistance to ciprofloxacin were amplified and sequenced as previously described (14).In-vitro analysis of growth kinetics. The ability of parent and disinfectant tolerant strains to grow in the presence of concentrations of disinfectants at MIC and sub MIC concentrations was monitored in LB broth using a FLUOstar OPTIMA plate reader (BMG LABTECH, UK). 100μl of sterile Luria-Bertani broth was dispensed into sterile microtitre trays and inoculated with overnight culture of each strain to give a final inoculum of 4% v/v. Readings were taken every ten minutes of the absorbance of each well (scanned at 600nm) in the microtitre trays over the 24h time period and results recorded automatically. Each strain was analysed in triplicate wells on at least three separate occasions to give nine data sets for analysis. Additionally, each strain was challenged with either 0.5x, or the MIC of the selective biocide for the parent after two hours growth (mid logarithmic growth phase) in order to determine whether the inhibitory ability of each biocide was reduced in mutants, respective to parent strains. Data were analysed using Excel (Microsoft, USA) to calculate means and standard deviations for each strain. Differences between strains were analysed for statistical significance using the Students t-test.Accumulation of norfloxacin. The accumulation of norfloxacin by each strain in the presence and absence of, CCCP (which dissipates the proton motive force and hence efflux) was directly measured fluorometrically as previously described (15).Effect of disinfectant exposure on the proteome of SL1344. Protein expression was assessed in order to determine bacterial stress response after disinfectant treatment. Proteomes were prepared in triplicate from SL1344 following exposure to TOP (0.04% v/v) and OXC (0.15% w/v) disinfectants for 90 minutes. Protein extracts were prepared and analysed by 2D-LC-MSn as described previously (16,17). The relative abundance of the proteins was compared using the spectrum count method (18) following published guidelines (19) and denotes the number of peptide counts (‘hits’) detected for each protein. Protein expression was given as a ration of the spectrum count. Expression analysis was limited to only those proteins common to all three replicates from control or disinfectant treated cultures. Proteomes were compared using Microsoft Access and Excel. The statistical significance of percentage changes in protein expression was determined using a two-tailed Student’s t-test. Survival of mutant strains in chicks. Chick studies were performed in a similar manner to previous studies (20). In brief, at three weeks of age each group of birds were infected by intra-gastric gavage with 106 cfu per bird with either strain DSD-1 or TOP-2 or OXC-2 (Table 2). For each of these groups of birds, there was a control group of birds infected with the respective parent strain (Table 2). Cloacal swabs were taken from each bird twice weekly for up to 27 days after inoculation using pre-weighed swabs which were weighed after swabbing so that results gave cfu/gram of caecal contents. Swabs were plated onto Rambach agar (Merck, Darmstadt, Germany), which was incubated overnight at 37C. As chicks were kept together in groups, for statistical analysis individual birds were regarded as independent units that did not interact with any other birds in the same group as far as the infection is concerned. With this assumption, the non-parametric Mann-Whitney was used to compare the mean log counts of the parent and disinfectant tolerant strains for each day post-infection.

RESULTSMutants with increased disinfectant tolerance were selected. Disinfectant tolerant mutants were selected from a range of strains (Tables 1 and 2) at a frequency between 1 x 10-9 to 1 x 10-10. It was only possible to select disinfectant tolerant strains from agar containing 2x MIC of disinfectants, not from agar with 4x or higher multiples of disinfectant MICs. Disinfectant, antibiotic and organic solvent susceptibility. L108 (SL1344, tolC::aph) and L643 (SL1344, acrB::aph) were hyper-susceptible to the phenolic (TOP), quaternary ammonium dairy steriliser (DSD) and mix of Quaternary ammonium compounds, formaldehyde and glutaraldehyde (QACFG) based disinfectants with >8-fold increases in susceptibility to these agents with respect to SL1344 (Table 2), but were not more susceptible to the oxidative disinfectant (OXC). There was little variation in the susceptibility of the other strains to disinfectants; the aldehyde-based disinfectant was significantly (c. 10-fold) more active than the other agents tested.Using the European suspension test (EN1040), five of the nine mutants required statistically significantly longer times to generate a five-log reduction in viability than their respective parents (Table 2). The MICs of ampicillin, chloramphenicol, ciprofloxacin, nalidixic acid, tetracycline, triclosan, ethidium bromide and the disinfectants for the parent and mutant strains isolated from agar containing 2x the MIC of disinfectant were generally identical, or within one doubling dilution of each other. Mutants obtained after exposure to TOP showed no increased

SID 5 (2/05) Page 9 of 25

tolerance to disinfectants or antibiotics. Mutants obtained after exposure to DSD were more resistant to exposure to DSD as seen by MIC (DSD-1) or increased time required to obtain a five log kill (DSD-2). Interestingly, DSD-2 was obtained from L108 (tolC::aph) after exposure to DSD and was significantly more resistant than its parent (Table 2). Mutants isolated after exposure to QACFG displayed a four-fold increase in MIC of ciprofloxacin as well as requiring longer exposure to QACFG for a 5-log kill (Table 2). OXC selected mutants required longer exposure to OXC for a 5-log kill (Figure 1) and the MIC of OXC was one-dilution higher in these mutants (Table 2). No increased organic solvent tolerance was detected for any of the disinfectant selected mutants compared to their respective parent strains (data not shown).Growth kinetics and accumulation assays.Accumulation of norfloxacin was unchanged for all mutants apart from QACFG-1, QACFG-2 and QACFG-3 (derived from L358 after exposure to QACFG), which accumulated significantly less norfloxacin than their parent (Figure 2). The addition of CCCP increased norfloxacin accumulation in each QACFG mutant indicating that the reduced accumulation seen in these strains is likely to result from an increase in active efflux. None of the disinfectant selected mutants were compromised in their ability to grow in Luria-Bertani broth relative to their parents in the absence of disinfectant. The majority of disinfectant selected mutants were more resistant to the addition of disinfectants to the media than their respective parent strains, including those strains for which the MIC of the selective disinfectant had remained unchanged when compared to the parent (Table 2). DSD selected mutant DSD-2 was able to grow significantly (p>0.01) better upon both the addition of 0.5 x and the MIC of DSD to the media (Figure 3). QACFG selected mutants QACFG-1, QACFG-2 and QACFG-3 all of which grew significantly (p>0.01) faster than their parent, L358, in biocide free broth and all grew significantly (p>0.01) better than L358 in the presence of both 0.012 and 0.025% QACFG. OXC selected mutants OXC-1 and OXC-2 both grew significantly (p>0.01) better than SL1344 when challenged with 0.5 x or the MIC of OXC. No significant differences were observed between TOP selected mutants TOP-1 and TOP-2 and their parent L357 in disinfectant free broth or when exposed to 0.5 x the MIC of TOP. However both mutants grew significantly (p>0.05) better when challenged with the MIC of TOP (0.2%).Mechanism of increased ciprofloxacin resistance in QACFG mutants.The frequency of mutant selection of QACFG-1, QACFG-2 and QACFG-3 was consistent with a one-step point mutation. The parent of QACFG-1, QACFG-2 and QACFG-3, L358 carries a Ser83-Phe substitution within GyrA; despite being cross resistant to ciprofloxacin no additional changes were found in the QRDRs of any of the three QACFG mutants. Investigation of the LPS and OMP profiles of all three QACFG mutants indicated no changes compared to L358 demonstrating that loss of porins was not responsible for increased ciprofloxacin resistance in these strains.Different disinfectants provoke different changes in the proteome of SL1344The effect of the TOP (0.04% v/v) and the OXC (0.15% w/v) disinfectants on the proteome of SL1344 following analysis by 2D-HPLC-MSn was determined. Treatment with both disinfectants had no significant effect on the total number of proteins detected. For the TOP experiment 360 proteins were detected in both the treatment and control (no disinfectant) and 424 proteins were detected in the OXC experiment (Table 3). The expression of 12 and 32 proteins were significantly (P >0.05) altered following treatment with the TOP and OXC disinfectants, respectively (Tables 3 and 4). The pattern of protein expression was very different after exposure to the two disinfectants. The TOP disinfectant significantly (P<0.05) increased expression of the AcrAB/TolC efflux pump, EmrA another multidrug efflux pump protein and others associated with detoxification of hydroperoxides (AhpC and AhpE) and pyruvate dehydrogenase (AceE and AceF) (Table 4). The OXC disinfectant increased expression of a range of proteins including RpsC a 30S ribosomal subunit, HtrA a periplasmic serine protease and FhuA a transporter for ferrichrome (Table 5). Both disinfectants increased expression of PqiB the paraquat inducible protein.Single step disinfectant selected mutants are fit in the chick modelThere were no statistically significant differences between the ability of the parent and single step disinfectant mutant strains in their ability to colonize and persist in chicks (data not shown). All groups of birds were still shedding c. 106 cfu of Salmonella / gram of caecal contents at the end of the experiments, 27 days after infection indicating that mutants colonised and persisted as well as parent strains.

DISCUSSION Mutants with a stable increased tolerance to disinfectants were selected from a range of strains at a frequency which suggests a single mutational event and that a primary target of each disinfectant has been altered in each mutant. As the disinfectants used all have multiple modes of antibacterial action it is likely that selection of mutants highly resistant to disinfectants will require multiple exposures to disinfectant and successive selection events.

The MIC data indicated that an intact AcrAB-TolC system contributes to intrinsic resistance to TOP, DSD and QACFG, but not OXC. The proteomic data supports this conclusion with over-expression of AcrAB-TolC and EmrA detected after exposure to TOP but not after exposure to OXC, which would appear to be insensitive to active efflux. The selection of mutant DSD-2 (selected from L109, tolC::aph) with a level of tolerance to DSD similar to most strains although it lacks a functional tolC gene, suggests that another efflux system may be over-expressed that can compensate for the loss of TolC.

SID 5 (2/05) Page 10 of 25

None of the single step disinfectant selected mutants were MAR, indicating that short, single exposure to these agents was insufficient to select for such mutants. However, the pattern of proteins whose expression were increased by exposure to the TOP was consistent with those previously associated with MAR, including increased expression of AcrAB-TolC. These data suggest that there is a link between disinfectant exposure and the major effectors of MAR. These first step mutants with a low level of disinfectant resistance could act as a stepping-stone to MAR strains.

The pattern of increased protein expression after treatment with OXC and TOP was distinct reflecting the different active constituents of each disinfectant. Any overlap was limited to increased expression of PqiB the paraquat inducible protein. Whereas TOP induced expression of several efflux proteins, OXC increased expression of different proteins including NfnB (dihydropteridine reductase/oxygen-insensitive NAD(P)H nitroreductase), AhpF (alkyl hydroperoxide reductase, F52a subunit; detoxification of hydroperoxides), which have been associated with resistance to hydrogen peroxide.

The increase in the MIC of ciprofloxacin seen for mutants QACFG-1, QACFG-2 and QACFG-3, after exposure to QACFG is likely to be a result of decreased accumulation of ciprofloxacin in these mutants, as no topoisomerase substitutions in addition to the Ser83-Phe substitution present in the parent strain were found. The accumulation of norfloxacin was reduced in these strains compared to their parent and the sensitivity of these strains to CCCP suggests that active efflux is responsible for the reduced norfloxacin accumulation observed. No decreased susceptibility to ciprofloxacin was detected in any of the other mutants selected from other strains after exposure to QACFG (data not shown) suggesting that the GyrA substitution in L358 may predispose selection of higher level ciprofloxacin resistance as seen in QACFG1, QACFG-2 and QACFG-3. The potential for disinfectants to drive resistance to ciprofloxacin is a real concern and may provide a selective pressure for the selection or maintenance of ciprofloxacin resistant strains in the farm environment in the absence of ciprofloxacin itself. None of the disinfectant selected mutants were compromised in their ability to grow in-vitro and were better able to tolerate challenge with disinfectants. The fitness of single step disinfectant tolerant strains in chicks was not compromised. A key consideration for assessing the risk that any mutants may present is their ability to survive in different food production environments. These may include the general farm environment, and associated specific niches, and the animals themselves where different selective pressures may apply. The present observation that the fitness of the disinfectant tolerant mutants was not compromised in chickens is cause for concern as survival in birds would enable both persistence on the farm and transit through the food chain.

Multi-step mutation to resistance to disinfectants

Tables are given in appendix (B.1)Figures are given in appendix (B.2)References from this section are given in appendix (B.3)

INTRODUCTIONNon typhoidal salmonella caused over 13,000 reported cases of enteritis in England and Wales in 2005 (http://www.hpa.org.uk/infections/topics_az/salmonella/ data_human.htm) over 65,000 across Europe (http://www.hpa.org.uk/hpa/inter/enter-net_reports.htm), resulting in significant morbidity and mortality. Whilst most cases require only simple bed rest and re-hydration, a proportion, particularly amongst the more vulnerable, may require antibiotic therapy. Clinically significant resistance to multiple antibiotics in salmonella has been increasing and over 30% of isolates of S. Typhimurium reported to the Enter-net surveillance network in 2004 showed resistance to three or more antibiotics http://www.hpa.org.uk/hpa/inter/enter-net_reports.htm) resulting in reduced therapeutic options. Recent regulations preventing the use of antibiotics as growth promoters in animals have been introduced in an attempt to reduce the emergence of antibiotic resistance in the food chain (http://europa.eu.int/comm/health/ph/others/ antimicrob_resist/am_02_en.pdf). This has necessitated an increase in the use of biosecurity measures including the application of disinfectants to reduce microbial contamination of animal houses in order to maintain standards of animal health. Antiseptics and disinfectants are used extensively in food and agricultural premises and they are an essential part of infection control practices. A wide variety of active chemical agents (or “biocides”) are found in these products (Mcdonnell and Russell 1999). Resistance to a disinfectant is thought unlikely to occur because most inactivate multiple cellular targets, and a single mutation would be unlikely to confer resistance (Mcdonnell and Russell 1999, Russell 2003). However, in cases where a disinfectant such as triclosan inactivates a specific cellular target, resistance can occur as a result of specific mutations in the genes encoding DNA gyrase (McMurry et al. 1998 a; Randall et al. 2004). Although disinfectant resistance is rare, reduced susceptibility to specific agents (quaternary ammonium compounds, chlorhexidine, diamidines and acridines) is known to occur (Mcdonnell and Russell 1999; Poole 2002). This phenomenon has been implicated as a possible cause for the selection and persistence of bacterial strains with reduced susceptibility to a range of agents, including antibiotics (Mcdonnell and Russell 1999) and it has been speculated that these bacteria could pass from the farm to the food chain (Piddock 2002). It is well known that Staphylococcus aureus can become resistant to certain antiseptics and disinfectants usually after receipt of plasmids carrying efflux-related transporters, such as qacA or smr-qacC, but

SID 5 (2/05) Page 11 of 25

also antibiotic resistance determinants (Moken et al. 1997). Similarly, in Gram-negative bacteria integrons carried by plasmids are also known to distribute genes including qacE together with several antibiotic resistance genes (Moken et al. 1997). Efflux, together with membrane impermeability, is a highly important non-specific defence mechanism that can confer multiple antibiotic resistance (MAR) in Salmonella enterica and Escherichia coli and a better understanding of the genes mediating this phenomenon and the pressures favouring selection of the MAR phenotype is urgently required (Giraud et al. 2000; van Bambekel et al. 2003). Efflux pumps are membrane proteins that actively export a wide range of toxic substrates, including antibiotics, dyes and disinfectants thereby preventing accumulation of toxic agents and mediating resistance (Levy 1992; Levy 2002). Various studies in E. coli and Salmonella have demonstrated that efflux pumps play an important role in intrinsic resistance to unrelated antibiotics such as fluoroquinolones and tetracycline and disinfectants including quaternary ammonium compounds and triclosan (Okusu et al. 1996; McMurry et al. 1998 b; Giraud et al. 2000; Levy 2002; Poole 2002; van Bambekel et al. 2003). They also seem to play a role in virulence (Buckley et al. 2006; Nishino et al. 2006) as well as survival in the presence of bile (Thanassi et al. 1997). The wide spectrum of substrates recognized by efflux systems has prompted concern that exposure of a bacterium to one substrate could confer resistance to others (Russell 2000 ; Fraise 2002). Thus exposure to disinfectants could result in selection of efflux mutants with reduced susceptibility to unrelated antibiotics and biocides, not previously encountered by the bacterium.

In the previous section of this report we described the derivation of single step mutants. Here we investigated the hypothesis that disinfectants can select MAR S. Typhimurium by multiple exposure.

MATERIALS AND METHODSBacterial strains and growth conditions. Wild-type Salmonella enterica serovar Typhimurium SL1344 (Wray and Sojka 1978) and a mutant carrying the substitution Asp87Gly within GyrA (L696), were used throughout (Table 1). Strains were stored in microbank tubes (Pro-lab Diagnostics, Neston, Wirral, UK) at -80oC and were routinely grown in Luria Bertani (LB) broth and on LB agar (Oxoid, Basingstoke, Hamphshire, UK) with or without disinfectants at 37oC with shaking at 160 rpm for the broth.Chemicals. All chemicals were obtained from Sigma-Aldrich (Poole, Dorset U.K) except for ciprofloxacin, which was kindly donated by Bayer (Newbury, Berkshire, UK). The farm disinfectants used were a tar oil phenol (TOP or FFS) which was a blend of high boiling point tar acids and organic acids, an oxidising compound based disinfectant (OXC or VS), a compound disinfectants containing quaternary ammonium compounds, formaldehyde and glutaraldehyde (QACFG, SK or ABD) and a dairy sterilizing disinfectant (DSD) which comprised a quaternary ammonium biocide, non-ionic surfactant and excipients. Farm disinfectants were obtained from appropriate suppliers. The triclosan (Irgasan DP300) was kindly donated by Ciba Consumer care (Macclesfield, Cheshire).Disinfectant adaptation experiments. Sequential growth experiments in the presence of disinfectants were performed as follows; strains from storage were inoculated to LB broth, for one overnight culture, as described above. Ten-fold serial dilutions of these cultures were prepared in sterile saline solution (Oxoid, Basingstoke, Hampshire, UK) and transferred in triplicate onto LB agar plates (Oxoid, Basingstoke, Hamphshire, UK), containing 0.05 μg ml-1 triclosan (TRIC), 400 μg ml-1 acriflavine (ACR), 2 μg ml-1 ampicillin (AMP), 6 μg ml-1

chloramphenicol (CHL), 0.06 μg ml-1 ciprofloxacin (CIP), 12 μg ml-1 kanamycin (KAN), 1.2 μg ml-1 tetracycline (TET) or without antibiotics. Agar plates were incubated at 37oC for 24 hours and numbers of colony forming units determined. In addition, overnight cultures were inoculated into fresh LB broth containing 0.2% OXC, 0.025% TOP, 0.006% QACFG, 0.06 μg ml-1 TRIC or no disinfectants and subcultured for six consecutive days using 1% (vol/vol) inocula in fresh LB with or without disinfectants, as appropriate. In parallel, cultures were transferred daily for up to 16 days to gradually increasing concentrations of disinfectants in order to select for populations resistant to high concentrations of disinfectants. On day six, for sequential growth in constant disinfectant concentrations and day 16, for sequential growth in gradually increasing concentrations, cultures were stored in microbank tubes at -80oC for subsequent antibiotic susceptibility testing. Before exposure to antibiotics, cultures were removed from storage and grown overnight on fresh LB with or without the corresponding concentration of the disinfectant in which they were selected. Subsequently, overnight cultures were sub-cultured on agar plates containing antibiotics and antimicrobials, as described above.Determination of minimal inhibitory concentrations. The minimal inhibitory concentrations of a range of antibiotics, dyes and disinfectants were determined using the agar doubling dilution method according to the recommendations of the British Society for Antimicrobial Chemotherapy (Andrews 2001). Isolation of antibiotic resistant variants from populations. In order to investigate the composition and the possible mechanisms conferring the multiple antibiotic resistance of the populations we obtained individual isolates TOP R1 and TOP R2, QACFG R1 and QACFG R2, OXC R1 and OXC R2, TRIC R1 and TRIC R2 from each of the corresponding wild type-derived populations TOP 7d, QACFG 7d, OXC 7d and 16 TRIC 16d, respectively. Individual isolates from populations QACFG 7d and 16 TRIC 16d were obtained by streaking a sample from each of the parent cultures on LB agar plates containing no antimicrobials as these populations were homogenous and contained only variants with increased antibiotic tolerance. However, in order to obtain individual variants with increased antibiotic tolerance from populations OXC 7d and TOP 7d a selective pressure had to be applied as these populations were a heterologous mix of wild-type cells and cells with increased antimicrobial tolerance (Fig. 1). A sample from the OXC 7d population was streaked on an agar plate containing

SID 5 (2/05) Page 12 of 25

400 μg ml-1 acriflavine and 10 colonies randomly selected colonies were screened for increased antibiotic tolerance. Similarly, individual isolates were obtained from population TOP 7d by streaking a sample on agar plates containing 6 μg ml-1 chloramphenicol. All isolates were identified as Salmonella sp. after growth on XLD agar plates and Gram staining. Stability of phenotypes. All individual isolates were assessed for phenotypic stability. Isolates were subcultured for 10 consecutive days using 0.3 % v/v inocula in fresh LB medium in the absence of any disinfectant at 37oC. On day 10, cultures were tested for their antibiotic and disinfectant resistance. acrB expression of representative isolates from the disinfectant-adapted populations. The expression of acrB in one multiply antibiotic resistant variant derived from each one of the wild type populations grown in the presence of each disinfectant was investigated. Three separate cultures were used to obtain RNA from each isolate. Logarithmic phase cultures grown in LB broth were harvested and pellets re-suspended in 1/5 volume of ice cold 95% ethanol and 5% phenol. Cultures were harvested during mid-logarithmic growth. Suspensions were kept on ice for 30 min before cells were harvested again by centrifugation at 3660 rpm for 10 min at 4 C. Pellets were re-suspended in 100 l of TE buffer containing 50 mg ml-1 lysozyme and incubated at room temperature for 5 min. RNA was purified using the Promega SV total RNA purification kit (Promega, U.K) according to the manufacturers recommendations. The quality and quantity of RNA extracted was determined by agarose gel electrophoresis and by analysis using a NanoDrop (NanoDrop Technologies). cDNA was synthesised from 2 g of total RNA for each sample in triplicate, using the SuperScript III system (Invitrogen, U.K.) with random primers according to the manufacturers instructions. RT-PCR reactions and DHPLC analysis for 16S rRNA and acrB (Transgenomic Ltd, U.K.) were performed as previously described (Eaves et al. 2004). Data were normalised to the expression of 16S rRNA to minimise cell density-dependent errors, by calculating the mean 16S expression from all RNA preparations. Subsequently, area under the curve values (AUC) for acrB from the same biological repeat were normalised as described previously (Eaves et al. 2004). The mean and standard deviation of the nine repeats for each data set was calculated, followed by two-tailed, two-sample, equal variance Student’s t-tests to compare gene expression relative to SL1344.

RESULTSDisinfectant adaptation experiments.In general, extended growth of SL1344 and L696 for seven days in the presence of disinfectants increased the frequency of antibiotic-tolerant variants in the resulting population. Results for SL1344 and L696 were similar throughout, indicating that the presence of the gyrA mutation does not pre-dispose SL1344 to disinfectant tolerance (data not shown). Extended growth in 0.06 μg ml-1 triclosan did not affect the frequency of antibiotic-tolerant variants. Except for triclosan, tolerance to increased concentrations of disinfectants was not achieved. During the period of the experiments (16 days), cultures derived from both SL1344 and L696 were obtained, that were able to grow in LB broth containing 16 μg ml -1 tricosan (Fig. 1). Exposure to disinfectants had no effect on kanamycin or ciprofloxacin tolerance in the final populations (data not shown). Extended growth of cells in OXC increased the frequency of triclosan (Fig. 1A) and acriflavine-tolerant variants (Fig. 1B) in the final population reaching 0.03% and 0.02%, respectively. Similar results were obtained with TOP, reaching 5.9% for triclosan (Fig. 1A) and 0.2% for acriflavine (Fig. 1B). Extended growth in QACFG and in gradually increasing concentrations of triclosan (max 16 μg ml-1) significantly increased antibiotic-tolerant variant frequency, reaching 100% for triclosan (Fig. 1A), acriflavine (Fig. 1B), tetracycline (Fig. 1C), chloramphenicol (Fig. 1D) and ampicillin (Fig. 1E).In parallel, the MICs for the populations exposed to disinfectants over time in broth were determined (Table 1). Growth of cultures for seven days in LB did not have any effect on the MICs with one exception, that of triclosan which increased 4-fold. The MIC of all disinfectants, except triclosan, remained stable for all populations tested, as growth with or without disinfectants did not have any effect. MICs of OXC, TOP, QACFG and DQACS were 0.2%, 0.4%, <0.05% and between 0.2 and 0.4%, respectively. The MIC of ciprofloxacin remained unaffected by sequential growth in any of the disinfectants. Repeated growth in OXC did not increase the MIC of chloramphenicol, tetracycline, kanamycin and ampicillin. Sequential growth in TOP did not affect MICs except a minor increase of 2-fold for kanamycin for the wild type. Sequential growth in QACFG increased the MIC of chloramphenicol and ampicillin by 2 and 4-8-fold respectively. In addition the MICs of tetracycline and kanamycin were raised up to 2-fold. The population grown in gradually increasing concentrations of triclosan had an increased MIC of chloramphenicol and ampicillin by 2 and 2-8-fold, respectively. The MIC of tetracycline and kanamycin was also raised by 2-fold.

Isolation of antibiotic resistant variants from the populationsIn order to determine the mechanism of resistance present in the MAR sub-populations generated in each condition single isolates were required. From two randomly selected isolates obtained from the QACFG 7d population, only one, QACFG R2, was multiply antibiotic resistant. For this isolate the MIC of chloramphenicol was increased 2-fold, while those of tetracycline and ampicillin were increased 8 and 2-4-fold, respectively, compared to the wild type. In addition, the MIC of TOP and DQACS were increased respectively 8 and 4-fold for this isolate. QACFG R1 had a similar phenotype to the wild type, with the exception of a 2-4-fold increase in the MIC of ampicillin and 4-fold increase in the MIC of triclosan. It was also 2-fold less susceptible to kanamycin.

SID 5 (2/05) Page 13 of 25

TRIC R1 and TRIC R2 had the same susceptibilities as 16TRIC 16d population from which they were isolated but, in addition, showed a 2 fold decrease in the MIC of kanamycin.Ten isolates were obtained from population TOP 7d by growth on LB agar plates containing 6 μg ml-1

chloramphenicol. Nine isolates, one of which was TOP R1, was not antibiotic resistant, even against chloramphenicol, but the MIC of triclosan was increased 4-fold (Table 2). Only one isolate, TOP R2 was multiply antibiotic resistant with an increase of 2 and 8-fold in the MIC of chloramphenicol and tetracycline, respectively, compared to the wild type. This strain also had a 4-fold reduction in susceptibility to TOP and DQACS. Ten isolates were obtained from the OXC 7d population by growth on LB containing 400 μg ml-1 acriflavine. Nine, including isolate OXC R2, did not show any antibiotic resistance. However, the MIC of triclosan was unchanged or increased by 2-fold. The remaining isolate (OXC R1) demonstrated a 2-fold decrease in susceptibility to chloramphenicol which was the selective pressure. However, the MIC of kanamycin reduced 2-fold. All isolates from these populations showed a stable phenotype following sequential growth in LB for 10 days.

acrB expression of representative isolates from the disinfectant-adapted populations mRNA levels for the transcript of acrB were determined for strains OXC R1, QACFG R2, TOP R2, TRIC R1 and TRIC R2 and compared to those of SL1344. TOP R2 demonstrated levels of expression for acrB similar to those of the wild type (P>0.05), while variants QACFG R2, OXC R1, TRIC R1 and TRIC R2 demonstrated a 3 to 4-fold higher expression (P<0.05) as shown in Fig 2.

DISCUSSIONAntiseptics and disinfectants (biocides) contribute to the production of safe and healthy food by maintaining the required levels of hygiene on farms and food processing facilities. They are freely available without prescription, unlike antibiotics, and as such are used on a daily basis in homes, schools, hospitals, restaurants, farms, abattoirs, other work places even in health care products (Randall et al. 2004). However, their widespread distribution and unlimited usage, has potentially been implicated in the development of antibiotic resistance, a concept that has been a subject of discussion (Russell 2000; Fraise 2002; Russell 2003). In the present study, the ability of several widely used disinfectants to select for antibiotic- or biocide-tolerant populations and isolates was assessed. Experiments were performed on pure cultures focusing on tolerance driven by changes in the endogenous resistance mechanisms within the bacterial cell. Work on this is rare, although the role of mobile elements and several related genes has been well studied (Moken et al. 1997; Randall et al. 2001). All the increased tolerance demonstrated by the disinfectant-adapted populations was due to the presence of variants with stable tolerant phenotypes and not to transient stress responses of wild type cells. It has been reported previously that mutations in gyrA might affect other unrelated characteristics including fitness (Randall et al. 2005). For this reason and to identify if any of the disinfectants interact with DNA gyrase, we included a gyrA mutant in our experiments. This mutation had no additional effect on behaviour in the presence of disinfectants or antibiotics, apart from resistance to quinolones. Triclosan was the only disinfectant where it was possible to increase its MIC by sequential growth of salmonella in gradually increasing concentrations of the compound in the growth medium, resulting in a population with a high-level resistance (1000-2000-fold increase of the MIC compared to that of the wild type). This is probably due to the fact that, unlike most disinfectants, triclosan inactivates a specific cellular target, FabI (McMurry et al. 1998 a; McMurry et al. 1998 b; Heath et al. 1999; Heath et al. 2000). The effects were not due to prolonged sequential growth since high level resistance was also achieved within seven days (data not shown), but on the 16 th day the population was able to survive the maximum concentration of triclosan in aqueous solutions (16 μg ml -1). It is remarkable the ease by which triclosan-resistant variants can occur within a pure culture, even without any selective pressure as demonstrated by the 4-fold increase of the MIC of triclosan following sequential growth in LB alone for 7 days. Population 16TRIC 16d comprised entirely tolerant variants that were resistant to the threshold concentrations of most antibiotics tested, but also to acriflavine as well as triclosan. These phenomena were associated with over-expression of acrB, as found in two randomly selected isolates (TRIC R1 and TRIC R2). It is well-known that over-expression of efflux pumps like AcrAB can confer resistance to triclosan and acriflavine in E. coli, as well as low level resistance to several antibiotics like β-lactams (ampicillin), tetracycline and chloramphenicol (McMurry et al. 1998 b). It is likely that additional mechanisms contribute to the high level of triclosan resistance, we are currently investigating these mutants for additional resistance mechanisms, including mutation within fabI. QACFG, like triclosan, selected for populations that almost entirely comprised variants resistant to the threshold concentrations of all the antibiotics and antimicrobials tested (Fig 1). These agents gave a wild type-derived population with increased MICs of chloramphenicol and ampicillin by 2 and 4-fold, respectively. Two representative variants were randomly isolated without any selective pressure on LB agar plates, from the QACFG 7d population. The MIC of ampicillin for QACFG R1 was increased 2-4-fold increase and that of kanamycin was reduced. The MIC of ampicilling for QACFG R2 was similarly increased, but there was also a 4 and 8-fold increase in the MICs of chloramphenicol and tetracycline. QACFG R2 expressed more acrB mRNA compared to the wild type. This could explain the reduced susceptibility to chloramphenicol, tetracycline and ampicillin, as has been shown previously (Giraud et al., 2000; Piddock et al., 2000). AcrAB of E. coli has previously been shown to be involved in tolerance to quaternary ammonium compounds (Moken et al. 1997). This variant was highly sensitive to TOP and the MIC of DQACS, which was the other quaternary ammonium disinfectant, was reduced by 2-fold although it was isolated after exposure to a quaternary ammonium-based disinfectant. It is possible that the surfactants that are present in both TOP and DQACS, but not in QACFG, are

SID 5 (2/05) Page 14 of 25

highly active against this variant. It is clear that continuous treatment with this disinfectant is able to select for antibiotic resistant strains with stable phenotype. These strains were able to overgrow the wild type population in only a week in the selective environment containing the disinfectant. Despite the above results with QACFG, it is important to note that the concentration used (MIC) was around 166 times lower than the recommended in-use concentration of 1%. OXC and TOP showed a reduced potential for selection of antibiotic-resistant variants. We did not find any increase in the MICs of antibiotics as an effect of exposure to these two disinfectants, but there was a slight increase in the frequency of acriflavine- and triclosan- tolerant cells. Sequential growth in TOP resulted in a population within which 100% and 7% of cells were tolerant to acriflavine and triclosan respectively, while growth in OXC resulted in values below 0.1%. Although the TOP 7d population comprised entirely of variants resistant to the threshold concentration of acriflavine, the MICs of all antimicrobials and antibiotics were identical to those of the wild type. Since cells in TOP 7d were similar to the wild type we decided to investigate the sub-fraction that was resistant to the threshold concentration of chloramphenicol. Of ten isolates, one (TOP R2) was less susceptible to tetracycline and chloramphenicol, which could also explain the increased frequency of tolerant variants to these antibiotics in the TOP 7d population. However, the expression of acrB was similar to that of the wild type and can not explain the 8-fold increase in the MIC of tetracycline. Since resistance to tetracycline is commonly related to efflux, other efflux pumps might be involved and we are investigating this further. The TOP R2 isolate was more susceptible to DQACS, but also to TOP, which was the disinfectant used to select its parent population. It is possible that the surfactants contained in both disinfectants are responsible for this. However, this variant, despite being four times more susceptible to this disinfectant, compared to the wild type, was able to grow at a concentration close to the wild type MIC, which was used to select for the TOP 7d population. A possible explanation lies in discrepancies between MICs performed in Muller Hinton agar and those in LB broth for TOP. The MIC of TOP in Muller Hinton agar for the wild type was 0.2 – 0.4% while in LB broth it was 0.04% (data not shown), leading us to select the concentration of 0.025% TOP to obtain the population TOP 7d. In further tests, TOP R2 was able to grow at 0.025% TOP in LB broth. In addition, the multiple antibiotic resistance of TOP R2 was not due to over-expression of acrB. OXC, like TOP showed a low ability to select for antibiotic and biocide resistance. All variant tolerance frequencies were below 0.1%. We investigated ten isolates derived from this population, which were resistant to the threshold concentration of acriflavine (400 μg ml-1). One of these, OXC R1 was less susceptible to chloramphenicol, but was more susceptible to kanamycin, compared to the wild type. However, this was a phenotype of the individual isolate and did not correlate with a similar phenotype in the corresponding OXC 7d population. Further investigation revealed that OXC R1 contained more than double the amount of mRNA levels for the acrB gene compared to the wild type. This could explain the increased MIC of chloramphenicol as described previously (Giraud et al., 2000; Piddock et al., 2000). With the exception of triclosan it was not possible to select mutants resistant to the disinfectants used in this study. However, poor disinfection and cleaning procedures might lead to low concentrations of disinfectants and biocides being in contact with the bacteria. Our work demonstrates that such concentrations could lead to selection of antibiotic-resistant strains that could infect farm animals and subsequently be passed to the consumer via the food chain. The oxidising compound blend (OXC) and the tar acids based disinfectant (TOP) selected the fewest antibiotic resistant strains. QACFG and triclosan, despite their very good antimicrobial activity, showed the highest selectivity for antibiotic and antimicrobial resistant strains. In most cases the resultant resistance was against tetracycline and chloramphenicol, which was linked with over-expression of acrB. This confirms the proposal that active efflux could be a connective link between disinfectant and antibiotic resistance, as it plays an important role in both phenomena (Randall et al. 2001; Levy 2002). However, there are numerous parameters that might influence the survival of the bacteria following disinfection in the environment and this could also explain the different points of view in this debate. More research is needed in this direction to help in the development of better disinfection and cleaning procedures as there are clear differences in the potential of different disinfectants to select antibiotic resistance, a consideration that could be useful when licensing these products.

Triclosan resistance in Salmonella enterica serovar Typhimurium

Tables are given in appendix (C.1)Figures are given in appendix (C.2)References from this section are given in appendix (C.3)

INTRODUCTIONSalmonella enterica serovar Typhimurium are a major cause of gastrointestinal illness in humans causing significant morbidity and mortality (Roberts et al., 2003). S. enterica are predominantly zoonotic pathogens, as a result the occurrence of S. enterica in the food chain can lead to human disease (Weill et al., 2004). Salmonellosis is often associated with the consumption of poultry products, which have been undercooked or have been contaminated with S. enterica (Carraiminana et al., 2004) and act as the major vehicle of delivery of

SID 5 (2/05) Page 15 of 25

salmonella to humans. Increased antibiotic resistance has been observed amongst many bacteria including S. enterica in recent years (Michael et al., 2006), a phenomenon which can compromise the efficacy of antibiotic therapy (Fluit, 2005). Recent studies have demonstrated that antibiotic resistant salmonella are associated with much higher mortality than antibiotic susceptible strains with quinolone resistance a particular risk factor (Helms et al., 2002). Efflux pumps are proteins, which are capable of transporting a wide range of toxic compounds out of the cell including antibiotics, dyes and biocides (disinfectants) and can confer a low-level multiple antibiotic resistance (MAR) phenotype (Piddock, 2006). The presence of the major efflux system AcrAB-TolC has recently been demonstrated to be a requirement for selection of high-level resistance to certain antibiotics and a pre-requisite for pathogenicity in S. enterica (Giraud et al., 2000, Baucheron et al., 2002, 2005, Buckley et al., 2006, Ricci et al., 2006). Compounds which inhibit bacterial growth are commonly used to reduce bacterial loads in the home and during food processing. Triclosan is a chlorophenol with broad-spectrum antimicrobial activity (Russell, 2004), which is commonly found in a very large array of domestic products marketed as antimicrobial including toothpaste, hand washes and cosmetics. Triclosan can also be used to impregnate surfaces and has been added to chopping boards, refrigerators, plastic lunchboxes, mattresses as well as being used in industrial settings, including food processing plants where walls, floors and exposed machinery have all been treated with triclosan in order to reduce microbial load (Medlin, 1997). The use of triclosan in industrial food production when incorporated to floors has been questioned and there is evidence that achievable antimicrobial concentrations are insufficient to significantly reduce viable numbers of many bacterial species (Cutter, 1999, Moretro et al., 2006). Recent data has indicated that high-levels of triclosan are present in groundwater in the USA and Europe (Kolpin et al., 2002; Singer et al., 2002) and a Swedish study has demonstrated significant accumulation of triclosan in the bloodstream and breast milk of a cohort of mothers (Allmyr et al., 2006). Triclosan acts as an inhibitor of fatty acid biosynthesis and binds to the enoyl-acyl carrier protein, FabI preventing elongation of the nascent fatty acid chain (McMurry et al., 1998). Mutation of the fabI gene, which lead to alterations of the affinity of triclosan to the active site of FabI and consequent resistance to triclosan, have been identified previously in E. coli, B. subtilis and S. aureus (Heath et al., 2000; Sivaraman et al., 2003). The FabL protein is a homologue of FabI and acts as the primary target for triclosan in B. subtilis (Heath et al., 2000) but is not present in Enterobacteriaceae. The MexAB-OprM efflux system of P. aeruginosa has been shown to confer a high level of intrinsic triclosan resistance (Chuanchuen et al., 2003) and E. coli which over-express the AcrAB-TolC efflux system or the global regulators marA and soxS have decreased susceptibility to triclosan (McMurry et al., 1998) demonstrating that triclosan is a substrate for efflux pumps. These observations have raised specific concerns regarding the proliferation of triclosan use due to the potential for selection of efflux pump over-expressing strains in pathogenic food-borne bacteria with concomitant multidrug resistance.The aim of this study was to select and characterise triclosan resistant mutants of S. Typhimurium and to determine whether the presence or absence of acrB or tolC influenced frequency of mutant selection and finally to analyse the fitness of mutants in the day old chick model.

MATERIALS AND METHODSBacterial strains. All bacteria used in this study are listed in Table 1. Construction of mutants derived from SL1344 with the acrB and tolC genes disrupted have been described previously (Eaves et al., 2004; Buckley et al., 2006). L696 is a ciprofloxacin resistant mutant selected from SL1344 (GyrA Asp87Gly, Ricci et al., 2006), L699 is a cyclohexane tolerant mutant selected from SL1344 after exposure to cyclohexane for 24h on Luria-Bertani agar (Webber et al., 2006). L357, L358 and L359 are Veterinary isolates of S. Typhimurium serotype DT104, L378 is a ciprofloxacin resistant strain of S. Typhimurium isolated from a chicken (GyrA Ser83Phe). All bacteria were stored on Protect beads at -80C until required.Media and Chemicals. Bacteria were routinely grown on Luria-Bertani agar plates (Oxoid, U.K) and in Luria-Bertani broth (Oxoid, U.K) unless stated otherwise. All chemicals were from Sigma (Poole, U.K), apart from Triclosan that was a kind gift of Ciba-Geigy (Macclesfield, U.K). Selection of triclosan resistant mutants. Mutants resistant to triclosan were selected on agar as described previously (Ricci et al., 2006) by exposing 105 to 109 cfu/ml of each strain to twice its MIC of triclosan and incubating at 37C overnight. Plates were scored for growth the next day, ten random colonies retained and frequencies of mutation calculated using viable count data from serial dilutions of culture plated onto triclosan free LB plates. Determination of antimicrobial susceptibilities. The MICs of a range of antibiotics, dyes and triclosan were determined using the agar dilution method according to the guidelines of the British Society for Antimicrobial Chemotherapy (Andrews 2001). All MIC determinations were repeated at least three times in independent experiments. Growth kinetics. The growth kinetics of parent and triclosan resistant mutants were determined by monitoring optical density (read at 600nm) using a FLUOSTAR OPTIMA (BMG laboratory technologies) every ten minutes at 37C for 24 hours. Triclosan at 8 g/ml was added to cultures at mid logarithmic growth phase (2h). Samples from broths were removed and visualised microscopically in order to detect any gross changes to cell morphology i.e. filamentation. Data were interpreted using Microsoft Excel. A non-paired student’s t-test was used to compare growth kinetics.Analysis of motility. The ability of each mutant to migrate through semi-solid motility test agar was determined. Agar plates consisting of Yeast extract 3g/L, peptone 10g/L, sodium chloride 5g/L and agar 3g/L were inoculated

SID 5 (2/05) Page 16 of 25

with triclosan resistant mutants and parent strains and incubated at 30ºC for 5 days. The diameter of the zone of migration of each strain was then measured and compared. Genotyping of fabI. The genotype of fabI was determined from all triclosan mutants, the entire fabI gene was amplified by PCR using primers: FabIF 5’AACGTCACCTGCCGGAGATA 3’, and FabIR 5’ TGGATTATCCTGGCGTATGC 3’. The resulting PCR amplimers were purified using a PCR clean up kit (QIAGEN, UK) before being sequenced at the University of Birmingham Functional Genomics Laboratory. DNA sequences were analysed using the Chromas software package (Technelysium, U.K).Over-expression and complementation of fabI. As mutations within fabI and increased expression of fabI were detected in various mutants two experiments were performed in order to define the role of fabI in triclosan resistance. In the first experiment the full-length fabI gene was amplified by PCR using primers: pBADFabIF, 5’ ACCATGGGTTTTCTTTCCGGTAA 3’ and pBADFabIR, 5’ TGCGCTTACTTCAGTTCCAG 3’ before being cloned into the pBAD vector expression system according to the manufacturer’s instructions (Invitrogen, U.K). Constructs were transformed into TOPO 10 cells (Invitrogen, U.K) by electroporation and plated onto Luria Bertani agar containing 100 g/ml ampicillin before overnight incubation at 37°C. Ten colonies were used to inoculate Luria Bertani broths (5ml) containing 100 g/ml ampicillin and incubated overnight again before plasmid DNA was extracted using the QIAGEN spin cell mini kit according to the manufacturers instructions (QIAGEN, U.K). PCR and sequencing was used to verify the correct nature and orientation of the fabI gene within pBAD. Once the construct was verified it was electroporated into SL1344, L696, L700, L701 and L702 after each had been made electrocompetent by harvesting 10ml of cells from logarithmic phase growth and three centrifugation/wash steps in ice-cold 15% glycerol. Candidate colonies were identified and verified as before. The expression of fabI was modulated by the addition of arabinose to logarithmic phase cultures grown in Luria-Bertani broth to determine the concentration of arabinose yielding maximum expression of fabI. Serial ten fold dilutions of arabinose were added to cultures of SL1344 with and without the pBAD-fabI construct to give a range of final concentrations from 0.2% to 0.00002%. At each concentration of arabinose, expression of fabI was determined from SL1344 and SL1344 (pBAD-fabI) by extracting RNA as in the gene expression experiments above, before RT-PCR was used to determine expression of fabI and 16SrRNA. The effect of differential expression of fabI on susceptibility to triclosan in each mutant was also determined by inoculating a series of agar plates containing a range of triclosan concentrations and arabinose concentrations. Luria-Bertani agar plates were prepared and inoculated with 104

cfu/ml-1 of SL1344, L696, L700, L701 and L702 and each strain carrying pBAD-fabI before being incubated overnight at 37C. After incubation the minimum concentration of triclosan required to inhibit each strain at each different concentration (0.00002, 0.0002, 0.002, 0.02 and 0.2%) of arabinose was determined.In the second experiment wild-type fabI and 250bp of flanking sequence was amplified with primers FabISDMF 5’ GCCAGATCTGACTTCGTTACCGTGTGGTT 3’ and FabISDMR 5’ GCCAGATCTACGAGATGAGTGGTGAGTGA 3’ before being cloned into pWSK30 (Wang and Kushner., 1991) after digestion with BglII. The resulting construct, pWSK30-fabI was electroporated into SL1344, L696, L700, L701 and L702, ampicillin resistant colonies were recovered and the presence of the correct construct verified by plasmid restriction analysis and PCR. Triclosan sensitivity of each strain +/- pWSK30 was determined as previously.Site directed mutagenesis of fabI. The G93V substitution in FabI found within various strains was re-created by site directed mutagenesis in SL1344 and L696 following the method of Turner et al., (2006). Mutant fabI alleles were amplified from L690 (table 2 Appendix C1) using primers FabISDMF and FabISDMR which incorporated BglII restriction sites. This amplimer was ligated into pJCB12 and propagated in SM10pir. Plasmids were transformed by electroporation into SL1344 and chloramphenicol resistant (25 g/ml) colonies obtained after growth on selective LB agar plates. In the absence of the pir gene pJCB12 cannot replicate so chloramphenicol resistance can only occur after integration into the host chromosome. Removal of the pJCB12 vector by excision after homologous recombination between the host cells fabI, and the mutant copy on the plasmid was selected for by growth of cells in the presence of 5% sucrose which counter selects against the sacB gene present on pJCB12. Colonies able to grow on LB agar containing 5% sucrose were replica plated onto chloramphenicol and sucrose and chloramphenicol resistant colonies retained. Ten such colonies were selected and plated onto antimicrobial free agar and agar containing 0.5 g/ml triclosan, 4 colonies were able to grow on this concentration of triclosan and the corresponding cultures on antimicrobial free agar were retained and the fabI locus amplified and sequenced to confirm the presence of the desired mutation. The same process was repeated to introduce the same mutation into L696.Accumulation of norfloxacin and Hoechst 33342 by triclosan resistant mutants. Two assays were employed to determine the level of efflux pump activity in SL1344, L696. L700 (LoT), L701 (MeT) and L702 (HiT). Firstly the efflux activity of strains exhibiting reduced susceptibility to disinfectants was compared to that of their parent strains by monitoring the uptake of the fluorescent dye bis-benzimide (Hoechst 33342 used at 2.5M) at excitation and emission wavelengths of 350 and 460 nm respectively over 30 minutes using a FLUOstar OPTIMA (BMG labtech, Aylesbury U.K). In a second assay, the accumulation of norfloxacin by each strain in the presence and absence of 100µM CCCP (which dissipates the proton motive force and consequently acts as an inhibitor of active efflux) was directly measured fluorometrically as previously described (Mortimer and Piddock, 1991). Differences in accumulation between strains were analysed for statistical significance using the Student’s t-test.Expression of fabI, acrB, marA, soxS and ramA by triclosan resistant mutants. Expression of genes implicated in triclosan resistance was measured to determine whether any of the known effectors of triclosan resistance were over-expressed in each mutant. Over-expression of fabI has been shown to confer increased

SID 5 (2/05) Page 17 of 25

resistance to triclosan in S. aureus, over-expression of efflux systems have also been implicated in triclosan resistance in E. coli and Pseudomonas aeruginosa. AcrB is the major efflux pump in salmonella; in E. coli this pump can be activated by transcriptional regulators marA and soxS. Homologues of both genes exist in salmonella as well as an additional gene; ramA which can also determine multiple antibiotic resistance in salmonella.Comparative RT-PCR (Eaves et al., 2004) was used to compare expression of fabI, acrB, marA, soxS and ramA from triclosan resistant strains L700, L701, L702 as well as SL1344 and L696. Expression of 16SrRNA was also measured and used as a control to normalise variation expression data for test genes before analysis. Expression of 16SrRNA was invariant between each strain. RNA was harvested from cultures grown in LB broth to mid logarithmic growth phase using a Promega SV total RNA kit. RNA was converted into cDNA using Superscript III (Invitrogen, U.K) and used as a template for PCR reactions after serial dilution. The resulting PCR amplimers were quantified using a WAVE denaturing HPLC machine (Transgenomic, U.K) and data analysed using Excel (Microsoft, U.K).Inactivation of acrB, tolC, ramA and marA in triclosan resistant mutants. P22 phage transduction was used to transfer mutant alleles of acrB, tolC, marA and ramA, each disrupted by insertion of a kanamycin resistance cassette into L696, L700 (LoT), L701 (MeT) and L702 (HiT) as previously described (Buckley et al., 2006; Eaves et al., 2004, Ricci et al., 2006). After inactivation of the chromosomal loci each mutant was complemented in-trans using the low copy number vector pWSK30, again as previously described (Buckley et al., 2006)Determination of protein expression. The proteomes of SL1344, L696, L700 (LoT), L701 (MeT) and L702 (HiT) were prepared in triplicate from each strain and protein extracts were analysed by 2D-LC-MS n as described previously (Coldham and Woodward, 2004; Coldham et al., 2006). The relative abundance of the proteins was compared using the spectrum count method (Liu et al., 2004) following published guidelines (Gao et al., 2004) and denotes the number of peptide counts (‘hits’) detected for each protein. Expression analysis was limited to only those proteins common to all three replicates from control or test cultures. Proteomes were compared using Microsoft Access and Excel. The statistical significance of percentage changes in protein expression were determined using a two-tailed Student’s t-test.Fitness experiments. Competitive index experiments were performed using the day-old chick model to assess the relative fitness of each triclosan resistant mutant relative to L696 (previous studies have established that L696 colonises and persists in the avian gut with equal efficiency to SL1344, its isogenic parent; unpublished data). Day old leghorn chicks were inoculated with 104 cfu per bird via oral gavage. Three groups of 12 birds were challenged with parent and mutant strains in a 1:1 ratio. Infections were monitored by cloacal swabbing at days 1,3,6,9,13,16,20,23 and 27, swabs were weighed pre and post sampling, vortexed in 1ml of sterile saline and diluted and plated (0.1ml) onto both BGA agar containing 4mg/L nalidixic acid to obtain the total salmonella count and onto BGA containing 1mg/L triclosan to obtain the number of triclosan resistant mutants present. Colonies were counted after overnight incubation at 37°C. At 28 days birds were sacrificed and post mortem examinations performed which included direct enumeration of cfu/g of caecal content.

RESULTSTriclosan resistant mutants are readily obtained from various strains.Nine strains of S. Typhimurium (Table 1) were exposed to triclosan in agar; triclosan resistant mutants were obtained from all strains apart from L108 (tolC::aph) and L643 (acrB::aph), both lacking major components of the AcrAB-TolC efflux complex. Where obtained, the frequency of selection of triclosan resistant mutants varied between 10-8 and 10-9 (Table 2). It was more difficult to select triclosan resistant mutants from SL1344 than other strains, and a high inoculum of ~1010 cfu/ml was required to obtain triclosan resistant mutants from this strain. Three distinct triclosan resistance phenotypes were identified amongst resistant mutants (Table 2); these were classified as low level (MIC of triclosan <8 g/ml), medium level (MIC of triclosan 16-32 g/ml) and high-level mutants (MIC of triclosan >32 g/ml). An exemplar of each phenotype was selected for further study; L700 (MIC of triclosan 4 g/ml; termed LoT), L701 (MIC of triclosan 32 g/ml; termed MeT) and L702 (MIC of triclosan 128 g/ml; termed HiT). Each of these mutants was derived from a common parent strain, L696 (Table 2), which is a spontaneous GyrA (Asp87) mutant of SL1344 (Ricci et al., 2006).Mutation within fabI does not correlate with triclosan MIC.Sequencing of fabI revealed a substitution of glycine with valine at position 93 of FabI in various mutants, including L701 (MeT) and L702 (HiT) (Table 2). However, when all the mutants selected were considered this substitution did not correlate with triclosan MIC, i.e. mutants classed as LoT, MeT or HiT could carry this mutation. Importantly, mutants with low, medium and high level triclosan resistance were also selected that did not possess this substitution (Table 2). One other substitution was detected in FabI, where serine rather than valine had replaced the wild-type glycine residue in L709 (LoT), selected from SL1344. No other mutations within fabI were detected in any strain. In order to determine precisely the contribution of mutation within FabI to triclosan resistance the glycine to valine mutation at codon 93 of FabI was recreated by site directed mutagenesis in SL1344 and L696. The resulting mutants were inhibited by 4, or 8 g/ml of triclosan respectively, 66 or 32-fold increases when compared to parent strains, respectively. This level of resistance is still significantly lower than that of MeT and HiT mutants demonstrating that this mutation alone cannot account for the higher level of triclosan resistance seen in these mutants.Complementation of fabI mutants partially restores triclosan susceptibility.

SID 5 (2/05) Page 18 of 25

To further determine the contribution of fabI to triclosan resistance in salmonella, wild-type fabI was introduced into mutant strains in trans for complementation studies. Introduction of wild-type fabI on plasmid pWSK30-fabI resulted in a fall in triclosan resistance in L702 (HiT; FabI G93V) from an MIC of 128 g/ml to 16 g/ml, representing an 8-fold reduction in susceptibility. No increase in triclosan susceptibility was seen when pWSK30-fabI was introduced into L696, L700 or L701.Over-expression of fabI leads to an increase in triclosan resistance.As previous work has indicated that over-expression of fabI can lead to increased triclosan resistance fabI was cloned into the pBAD vector system under the control of an arabinose inducible promoter and introduced into various mutants. Expression of fabI was measured by RT-PCR, which showed an increase in expression of fabI that correlated with incubation with increasing concentrations of arabinose; a five fold increase in fabI expression was achieved at an arabinose concentration of 0.002% (data not shown).For SL1344, L696 and L700 (LoT), all of which possess a wild-type fabI triclosan, MICs increased 2-4 fold when carrying pBAD-fabI induced with 0.002% arabinose (Table 3). The introduction of pBAD-fabI into L701 (MeT, fabI::G93V) had no effect on triclosan MIC when exposed to 0.002% arabinose. Over-expression of wild-type fabI in L702 (HiT) led to a decrease in triclosan resistance from 128 g/ml to 16 g/ml.Efflux contributes to intrinsic and high-level triclosan resistance in S. Typhimurium.In order to define the contribution of the AcrAB-TolC system to triclosan resistance in salmonella, mutant acrB::aph and tolC::aph alleles (from L643 and L108, respectively) were transduced into L696, L700 (LoT), L701 (MeT) and L702 (HiT) using phage P22. Disruption of both genes increased sensitivity to triclosan between 4 and 256-fold (Table 3). Loss of tolC gave larger increases in triclosan sensitivity than loss of acrB, but no transductants regained full wild-type (SL1344) triclosan susceptibility. Interestingly inactivation of tolC in L701 (MeT) resulted in a smaller fall in triclosan resistance than seen in L700 (LoT) and L702 (HiT). L701 (MeT) tolC::aph transductants required 8 g/ml of triclosan for inhibition of growth, a four fold increase in sensitivity compared to the 33 and 256 fold changes seen in the LoT and HiT mutants, respectively. Complementation of the triclosan mutants which had had either acrB or tolC inactivated with pWSK30-acrB or pWSK30-tolC, respectively led to full restoration of the triclosan resistance seen in the original mutant (data not shown).As inactivation of AcrAB-TolC increased triclosan sensitivity the activity of efflux systems was investigated in triclosan resistant mutants. The accumulation of norfloxacin and Hoechst 33342, substrates of the AcrAB-TolC system by all strains were determined. No significant differences in accumulation of either agent between SL1344 and L696 were observed (Figure 1A and 1B). L700 (LoT) and L701 (Met) accumulated significantly less (P = <0.05) norfloxacin (Figure 1A) and Hoechst 33342 (Figure 1B) than L696 or SL1344. The addition of the proton motive force inhibitor, CCCP led to an increase in the concentration of norfloxacin accumulated by all strains although the amount of norfloxacin accumulated by L700 was still significantly lower than that of L696 ( P = <0.05). The addition of CCCP also increased accumulation of Hoechst 33342 by L700 (LoT), and L701 (MeT) but did not increase Hoechst accumulation by L702 (HiT, Figure 1B).Expression of fabI, acrB, marA, and ramA varies between triclosan resistant mutants.The expression of genes previously associated with resistance to triclosan was determined by RT-PCR in order to identify any correlation between gene expression and phenotype. The 16SrRNA gene was used as a control and expression of this gene was invariant between all strains. Expression of fabI was increased in all three triclosan resistant mutants but only statistically significantly in L702 (HiT, 1.3 fold increase when compared to L696). Expression of acrB was similar for all strains except for L700 (LoT), RT-PCR indicated that L700 (LoT) expressed significantly more (1.32 fold increase, P >0.01) acrB than the other triclosan resistant mutants and L696. No statistically significant changes in AcrB production were detected by proteomics. Expression of the transcriptional regulator marA was highest in L701 (MeT; 1.4 fold increase), compared to the other strains; however this was not statistically significant. Both L701 (MeT) and L702 (HiT) significantly over-expressed the marA homologue ramA (1.31 and 1.27 fold increases compared to L696, respectively). Proteomic analysis did not detect peptides specific for either MarA or RamA so the production of these proteins could not be quantified.As marA and ramA were found to be over-expressed in either L701 (MeT), L702 (HiT) or both and these genes can induce expression of acrAB both genes were also inactivated in the triclosan resistant mutants by transfer of mutant marA::aph and ramA::aph alleles from L101 and L103 (Ricci et al., 2006), respectively using P22. The inactivation of marA or ramA gave similar results to those recorded when acrB was disrupted (Table 3). Inactivation of marA and ramA in L696 had no effect on triclosan susceptibility, whereas inactivation of either allele reduced the triclosan resistance of L700 (LoT) two fold (Table 3). Loss of a functional ramA gene in L701 (MeT) increased sensitivity to triclosan 16 fold but inactivation of marA in this strain had a less dramatic effect, resulting in a 2-fold increase in triclosan susceptibility, a similar result to that seen when tolC was disrupted in this strain. Disruption of marA in L702 (HiT) resulted in a 64 fold increase in triclosan susceptibility compared to a 32 fold increase observed when ramA was inactivated.Analysis of the proteomes of triclosan resistant mutants.For all three strains (L700, L701 and L702) investigated over 500 proteins were detected which were common to the test and reference strain (L696). The majority of these did not display significantly altered abundances, only proteins with significantly altered (P < 0.05) expression were included in further analyses. L700 (LoT) had the lowest number of significantly altered proteins (65), L701 (MeT) had over twice as many proteins with altered expression (138) and L702 (HiT) had the highest number of proteins with significantly altered expression (169). For all three strains more proteins with increased expression were detected than with decreased expression. Twenty-five proteins had significantly altered expression in all three triclosan resistant mutants (Table 4),

SID 5 (2/05) Page 19 of 25

indicating that these proteins may represent a common triclosan resistance network. Amongst these 25 proteins were 9 which are all involved in the generation of pyruvate which feeds fatty acid biosynthesis or were part of pathways which can generate fatty acid by alternative metabolic routes (Figure 2). Amongst these proteins was a putative arginine-deiminase (STM4467/ ArcA) involved in the conversion of arginine residues in cellular proteins to citrulline, which was highly (7-fold) over-expressed in L700 (LoT) and over-expressed to a lower extent in L701 (MeT) and L702 (HiT) (Table 4). Also increased in all three mutants is the glycine decarboxylase complex P protein, GcvP involved in the breakdown of free glycine, MdH (malate dehydrogenase), MaeB (malate transferase), GapA (glyceraldehyde-3-phosphate dehydrogenase), PpS (phosphoenolpyruvate synthase), FadB (3-hydroxyacyl-coA dehydrogenase) and GltA (citrate synthase). Other proteins not associated with fatty acid synthesis were also up-regulated in all three mutants, both HemL and HemX involved in porphyrrin biosynthesis were over-expressed in all triclosan resistant mutants. Three proteins involved in stress response or global gene regulation demonstrated increased expression in all three triclosan resistant mutants (Table 4) including HNS, HtrA (involved in response to heat shock) and CspC (involved in response to cold shock). As well as those proteins with differential expression in all triclosan resistant mutants there were distinct individual patterns of protein expression observed in each mutant studied. L700 (LoT) had increased expression of FabI (3.8 fold), and FabB (2.2 fold), both involved in fatty acid biosynthesis (Table 5). FadB was also increased (3 fold) which metabolises medium chain length fatty acids. A range of proteins involved in motility and chemotaxis were repressed in L700 (LoT) relative to L696 (Table 5) indicating a likely impairment of motility in this mutant. This loss of motility in L700 (LoT) was confirmed by growth on semi-solid motility agar, L701 (MeT) and L702 (HiT) were not defective in their motility (data not shown). Whilst there was significant divergence (38% of the 139 significantly altered proteins detected in L701 were not significantly altered in L702) between the proteomes of L701 (MeT) and L702 (HiT), a group of 86 proteins displayed similar patterns of expression in both strains (Table 6). These included Gnd which is a dehydrogenase specific for gluconate, and UspA and IbpA, both stress response proteins (Table 6). Analysis of proteomic data for FabI showed a good correlation with the RT-PCR data with increases in production of the FabI protein in all strains although this was only statistically significant for L700 (LoT). Similar increases in the amount of FabI protein detected in L701 (MeT, 1.5 fold, P = 0.057) and L702 (HiT, 1.4 fold, P = 0.055), compared to the fold increases in fabI detected by RT-PCR (1.2 fold increase detected in both L700 and L701) were seen. A large increase in FabI production (3.8 fold, P = <0.01) was seen in L700 (LoT) compared with only a 1.3 fold increase in transcription of fabI from this strain. This indicates that the increase in FabI produced in this strain is a result of protection from degradation or increased stability in some way rather than a large increase in mRNA expression. The presence of the G93V substitution within FabI was confirmed by analysis of tryptic peptide sequences from both L701 (MeT) and L702 (HiT). YadG, which encodes the ATP-binding domain of a putative multidrug efflux system, with YadH was over-expressed (2-fold) in L701 (MeT) relative to L696 but not in L702 (HiT), YadH was not detected in any strain and hence could not be enumerated. No statistically significant over-expression of AcrAB-TolC was detected in the triclosan resistant mutants.Triclosan resistant mutants are fit.To determine whether the development of triclosan resistance resulted in a fitness burden the growth of L700 (LoT), L701 (MeT) and L702 (HiT) in Luria-Bertani broth was compared to that of L696 and SL1344 over 24 hours and strains were challenged with 8mg/L of triclosan during mid-logarithmic growth (Figure 3). Upon addition of triclosan all strains suffered an initial decrease in growth from which SL1344 did not recover. However, three hours after exposure to triclosan the three mutants recovered and resumed growth (Figure 3). The growth kinetics of L700 (LoT) and L701 (MeT) were similar with no statistically significant difference. L702 (HiT) was able to grow significantly better when exposed to triclosan than the other mutants (P=<0.05) and attained a higher peak absorbance value (Figure 3). The ability of L700 (LoT), L701 (MeT), and L702 (HiT) to persist in vivo in the avian gut in competition with their parent L696 was determined. All three strains were able to colonise chicks (Figure 4). The pattern of colonisation and persistence of each of the three mutants was very similar; each was able to persist with similar numbers of colonies being isolated at the end of the 28 day experiment as at the beginning (i.e. between 104 to 105 cfu/g of faeces). The level of colonisation of each triclosan mutant was similar throughout the experiment indicating successful establishment of a stable population (Figure 4). Although all triclosan mutants were able to colonise and persist in the avian gut the numbers of bacteria were lower than for L696, which showed an amplification of bacteria with average numbers of ~106-107 cfu/g isolated at the end of the experiments. DISCUSSIONSelection of triclosan resistant S. Typhimurium from several strains proved relatively easy at a frequency suggestive of single point mutations. The frequency with which triclosan resistant mutants were obtained was 10-fold less from SL1344 than from DT104 isolates or strains carrying gyrA mutations. Three distinct triclosan resistance phenotypes (LoT, MeT and HiT) were obtained from a variety of strains and all three phenotypes could be isolated from the same strain indicating that there are at least three distinct mechanisms of resistance underpinning each different phenotype. Previous work with S. aureus and E. coli (Sivaraman et al, 2003, McMurry et al., 1998) has suggested that substitutions within FabI are the primary mechanism mediating triclosan resistance, particularly high-level triclosan resistance. However, in the present study although mutations within fabI were detected in a variety of mutants, these did not correlate with the level of triclosan resistance observed. One predominant mutation of Glycine 93 to Valine within FabI was discovered amongst various isolates and was recreated in SL1344 and L696

SID 5 (2/05) Page 20 of 25

in order to determine the contribution of this substitution alone to triclosan resistance. This experiment demonstrated that this substitution alone conferred a level of triclosan resistance of 4-8 g/ml indicating that other resistance mechanisms determine higher levels of triclosan resistance. Complementation of a highly triclosan resistant (HiT) mutant with wild-type fabI carried on vector pWSK30 resulted in an 8 fold decrease in triclosan resistance but wild-type susceptibility was not restored, providing further evidence that high level triclosan resistance in salmonella is not determined by mutation in fabI alone. In the same experiment complementation of L701 (MeT) with wild-type fabI resulted in no decrease in triclosan resistance, again suggesting that mutation within fabI alone cannot account for the phenotype of MeT and HiT mutants. Over-expression of fabI has been implicated as a mechanism of triclosan resistance in S. aureus (Fan et al., 2002), and one mutant, L700 (LoT) over-produced FabI (and FabB) indicating that this mechanism may be relevant in salmonella. However, artificial over-expression of fabI in L696 only resulted in a level of triclosan resistance four fold lower than that seen in L700. The huge (greater than 500 fold) decreases in triclosan susceptibility in HiT mutants can not be accounted for by fabI expression increases. In mutants carrying a mutant fabI gene no increased triclosan resistance was seen when fabI was over-expressed, possibly as a result of complementation of the mutant allele by the wild-type gene.It was not possible to select resistant mutants from strains lacking the acrB or tolC genes demonstrating that an intact AcrAB-TolC system is required for the development of triclosan resistance in salmonella. These data indicate that an intrinsic level of efflux activity is required to prevent accumulation of toxic triclosan concentrations within the cell which are higher than the MIC of any mutant arising from a single mutational event. Further evidence that efflux contributes to acquired triclosan resistance in salmonella was demonstrated when inactivation of the tolC or acrB genes in triclosan resistant strains led to significant decreases in triclosan resistance although wild-type susceptibility was not restored. These observations are similar to analogous findings with fluoroquinolone resistance in E. coli where target site topoisomerase mutations can give high level resistance to drugs such as ciprofloxacin but are ineffective in the absence of the AcrAB-TolC system (Oethinger et al., 2000). No increased production of AcrAB-TolC was detected in the triclosan resistant mutants by proteomics. The accumulation of norfloxacin and Hoechst 33342 by L700 (LoT) and L701 (MeT) was significantly lower than in L696 suggesting reduced permeability of these mutants as a result of porin down-regulation or active efflux; this observation indicates that one or more of the numerous other efflux systems of S. Typhimurium may be relevant to high-level triclosan resistance. Inactivation of the transcriptional regulators ramA and marA led to increased triclosan susceptibility, in a manner similar to that seen when acrB was inactivated indicating that these genes may be required for maintenance of acrB expression in triclosan resistant mutants. L701 (MeT) showed a different pattern of triclosan resistance when tolC and marA were inactivated compared to the LoT and HiT mutants. Small (2-4) fold increases in triclosan susceptibility were seen rather than the larger changes seen when the same genes were disrupted in L700 (LoT) or L702 (HiT). This indicates that tolC and marA are not as important in maintaining the level of triclosan resistance seen in L701 (MeT) as in L700 (LoT) and L702 (HiT). L701 (MeT) was shown by proteomics to over-express the putative multidrug efflux system YadGH, it is possible that this system contributes to triclosan resistance in L701 (MeT) and that AcrAB-TolC is less important than in other strains. Proteomics identified a set of proteins commonly up-regulated in all triclosan resistant mutants; this ‘triclosan resistance network’ included 9 proteins involved in production of pyruvate or fatty acid. This may represent a mechanism by which the triclosan resistant mutants have increased throughput of fatty acid biosynthesis by increased pyruvate production or have altered metabolic pathways in order to produce fatty acid via a different pathway (conversion of glycerol to hexadecanoate or increased citrate production to feed acetyl-CoA production). These data extend current understanding of the effects of triclosan on cells and suggest alternative mechanisms of triclosan resistance are relevant. Triclosan resistant mutants were not compromised for growth in LB broth in vitro, and all the triclosan mutants analysed in the day old competitive index chick model were able to colonise and persist within chicks with similar efficiency throughout the experiment, albeit in lower numbers than their parent strain, L696. This indicates that development of triclosan resistance in salmonella does not carry a prohibitive fitness burden demonstrating that such mutants will not be out-competed readily by other strains in the environment and are likely to survive in the food chain. Data presented here demonstrate that there are three distinct phenotypes and mechanisms of triclosan resistance in salmonella and that a variety of genes are involved in triclosan resistance including fabI, acrB, tolC and ramA. Proteomic data revealed specific patterns of protein expression in each mutant but also identified 25 proteins which constitute a common triclosan resistance network in all mutants studied. A functional AcrAB-TolC system was shown to be essential for intrinsic and high-level triclosan resistance in S. Typhimurium. Triclosan exposure can select for diverse resistance phenotypes, the increasing use of triclosan in an expanding range of applications including those in food processing should be considered carefully. The new observation of a triclosan resistance network may have implications for other bacteria which are subject to triclosan exposure.

Overall conclusions and recommendations

One step biocide selection work: It is clear that a single exposure to a disinfectant can generate mutants with an altered phenotype, most noticeably decreased sensitivity to the selecting agent. The molecular studies into the disinfectant tolerance showed that is was mediated, in part at least, by efflux systems. The studies showed that

SID 5 (2/05) Page 21 of 25

these single step mutants did not possess the classic multiple antibiotic resistance (MAR) phenotype which readily mutate to higher levels of resistance especially to quinolones (as shown in our previous work OD2006). We must be cautious in our interpretation however. We did not perform forward mutation studies with these single step mutants neither were the studies exhaustive. Only a limited number of mutants of the many that arose under the selection pressure were studied in detail and it is possible that those not selected for detailed study may well have been MAR. That said, it is evident that MAR would be in the minority of disinfectant tolerant mutants if indeed they did arise. It would be pertinent to undertake studies to answer the two questions raised here.The next question is whether or not these studies performed under laboratory conditions reflect the likely scenario on farm, in food processing, in hospitals or in the home where disinfectants are used extensively. Under what circumstance would bacteria be exposed to sub-lethal concentrations of disinfectants? A review of disinfection regimes shows there are many different practices applied under a wide range of circumstances and most end with a washing stage to remove residual disinfects which can be toxic to man and animals. This will generate dilute disinfectants. Furthermore, bacteria are known to occupy many niches, some of which are difficult to penetrate with disinfectant. The elegant studies of Dr R. H. Davies has more than adequately shown high numbers of bacteria, including pathogenic strains, do survive full C&D procedures. We must conclude many bacteria will be exposed to sub-lethal disinfectants and , therefore, the risk of mutants arising is high.

Mutilpe exposure selection work: The reason for undertaking studies with prolonged exposure to disinfectants was in part to take some of the questions discussed above to the next obvious stage. The question arising relating to whether or not true MAR mutants could arise from exposure to disinfectants which, with the caveats discussed, did not arise by single exposure needed to be tested. In our study, we showed that for many disinfectants the entire population of organisms exposed to seven daily cycles of sub-lethal disinfectant became tolerant. Furthermore, isolates made from these populations had stable phenotypes and it is not unreasonable to deduce stable genotypes also. Many of the isolates from this study could be defined as MAR mutants. Thus, we have answered an important question, disinfectants can give rise to MAR. The next question is whether or not these studies performed under laboratory conditions reflect the likely scenario on farm, in food processing, in hospitals or in the home were disinfectants are used extensively. Under what circumstance would bacteria be exposed to sub-lethal concentrations of disinfectants for such an extended period? Multiple C&D rounds are used in many environments. So, whilst it is not that likely current C&D procedures would generate the exposure we created in the laboratory it is likely, that overtime, bacteria may well be exposed to multiple exposures to sub-lethal concentrations of the disinfectant. The model we used in the study was based on continuous cycling in sub-lethal disinfectant whereas a more likely scenario may be multiple exposures over time. The mutants arising from the longer term exposure were MAR but in every case they were also attenuated with a range of other phenotypes which, even on brief inspection, indicate these mutants were ‘unfit’. What we must focus on is the possibility, and a very likely possibility, that bacteria in farm and other environments may well be exposed one, two or three times and therefore selected for additional traits at each exposure. It seems likely that progression from one to two exposures may select a variety mutants which could be MAR and which retain fitness. By a seventh exposure as performed here in the laboratory it is likely that the attrition rate is so high that detrimental mutations arise resulting in a lack of fitness. Future work should focus on fewer rounds of disinfectant exposure to establish the transition from tolerance to MAR. We propose that selective pressure in the field, as opposed to the laboratory, may drive mutation down pathways to MAR which do not compromise fitness.A subsidiary question arises regarding which disinfectants are less likely to generate MAR phenotypes. It seems reasonable to suggest to those that use disinfectants that they might consider rotating different classes of disinfectants. We obtained evidence that this might be a good practice as certain isolates that were obtained by growth in sub-inhibitory concentrations of disinfectants had reduced MICs of other disinfectants than those they were selected from. Whilst this is a sensible strategy, we should first select those disinfectants that do not induce MAR as readily as others. In addition understanding why, if this is shown to be the case, becomes a significant question relating to the proposed mechanisms of MAR. We have shown quite conclusively there are (at least) two generic routes for bacteria becoming tolerant. First relates to the ability to pump the offending agent out by efflux pumps. Up-regulation of these pumps will give MAR. The second relates to increased metabolic activity aimed at reducing cellular damage. There are many established survival responses but what we do not understand is how these survival networks and efflux work co-operatively. If we were to gain insights into this we would have totally new targets for control measures. Additionally, we have also found reduced permeability to play a role without any additional efflux mechanism indicating other than the two generic routes described above to reduced susceptibility.

Triclosan work: Triclosan is used widely and manufacturers are quick to state that resistant mutants arise either very rarely or not at all. We and many others have shown conclusively that they do. Triclosan resistant mutants were selected easily from various strains and three discrete resistance phenotypes were observed (Low, Medium and High). Previous work and that reported here indicates FabI is a major target of triclosan and tolerant mutants with altered FabI amino acid structure arise. This is now established fact. What we have shown in addition to this is that FabI does not account for all triclosan resistance observed. This is a highly novel finding. We showed conclusively that an intact AcrAB-TolC efflux pump system is required for the development of high-level triclosan

SID 5 (2/05) Page 22 of 25

resistance. The link between efflux and MAR should be considered in the context of Triclosan also. This warrants further investigation especially as Triclosan is becoming ever more widely exploited.We have a common theme cutting across our studies, efflux is crucial to the survival of the bacteria and mutants with up-regulated efflux are tolerant to a wide range of antimicrobials. We hypothesise that depending upon the degree of up-regulation, the type of efflux pump, the linkage with other outer membrane protein changes (i.e. lower total OMP, down-regulated porins) and the antimicrobial, a range of phenotypes are readily selected of which MAR is one. The step from Mar to higher level resistance is potentially more rapid than for non-MAR parent strains. Therefore, any process that generates MAR is inappropriate. Finally, we focused on the fitness of the mutants arising from these studies. Both single step, single step triclosan and long term selection Triclosan mutants were as fit as the progenitor parent strain as assessed by a number of assays [with the exception of HiT-like mutants isolated following growth at increasing concentrations (max conc. 16μg/ml) of this biocide for 16 days which had reduced invasiveness in Caco-2 although having normal growth rates]. The majority of these mutants were fully virulent in the chick model. This model was selected as it known for broiler production with alI-in-all-out systems, single disinfections between batches are applied. We hypothesised that single step mutants would persist in that environment and we have demonstrated conclusively that these mutants have the potential to colonise and persist in that host. The risk to human health is obvious. Is a single C&D enough between batches? Dr R. H. Davies showed that up to 40% houses positive for salmonella prior to cleansing remained positive after C&D. We need to establish best practice and consider multiple C&D rounds whereby we would predict that surviving bacteria are sufficiently attenuated so as to not cause a risk. A question arises as to how many rounds of C&D, as discussed above.

Future workThe present laboratory investigations have clearly identified a causal link between disinfectant exposure and the development of resistance to multiple antibiotics mediated by changes in the expression of proteins which may reduce or negate the effect of such agents. Further studies are required to substantiate these adaptive effects in the field, the likely risk that they present and the control measures required. This would include the attempted recovery of Salmonella with reduced susceptibility to multiple antibiotics from a farm type environment, and crucially, assessment of their fitness. These studies are necessary due to the reliance on cleansing and disinfection procedures and likely divergence between laboratory and farm environments. Such differences include the modifying effects of temperature, organic waste, and specific environmental niches which may promote the emergence of such mutants. Clearly, growth of bacteria on agar media is likely to be divergent from the situation in the field and this may have different modifying effects from those in the field.We have focused on the impact of efflux mechanisms on reduced sensitivity and shown conclusively the significance of these common bacterial systems upon development of MAR. Essential work is now required to assess the impact of efflux and MAR on the total biology of Salmonella. We have already shown one-step mutants to be as virulent as the parent strains in the chick model but from companion studies (OZ0324) we are using the OMNILOG system to analyse 2000 phenotypes per strain and future research should now address how MAR and efflux influences the overall phenotype. This author contends that whilst we can readily identify MAR by reduced susceptibility to a range of unrelated antimicrobials there will be other metabolic impacts that alter the phenotype and therefore behaviour of these mutants. We need to progress from descriptive to predictive biology. By analysis of metabolic shifts and fluxes due to changes in efflux, we can predict changes in the metabolism and the impact those changes have on responses to the environment. For example, it would be important to know whether such mutants are better adapted to survival through the many and various processes in the food chain. Thus, by analysis of the ‘metabolome’, we should be able to predict risk more effectively. By understanding the metabolic shifts we may identify new targets for intervention. Specifically, we have identified changes in substrate level phosphorylation that favour reduced susceptibility to disinfectants/antibiotics but which compromises fitness.  Thus, we can assess if this is a common adaptation in a range of isolates from the field, using the OMNILOG. If proven, this would support reduced risk of survival through the food chain albeit at the expense of reduced susceptibilities. We need to dissect this adaptive metabolic response as many of the disinfectants have multiple components. For example, substrate level phosphorylation adaptive response to detergents may cause disaggregation of F0F1 ATP synthase located in the cytoplasmic membrane whereas the reduced susceptibility may be linked to oxidising agent stress response, possibly via SoxRS, RamA or MarRAB.  Again we can use the OMNILOG to assess and use this information to provide Defra with advice on use of disinfectants that avoid selection for mutants with reduced fitness where as oxidising agents promote the development of antibiotic resistance as they switch on efflux pumps. We also need to explore the transition from the single step mutants to multiple step mutants (as opposed to those generated from daily passage.  This may provide an insight into ways for effective disinfection without the transition to the mutants with reduced susceptibilities.Perhaps the most concerning feature of this work was the ease with which one disinfectant (QAC based) lead to the selection, with relative ease, of mutants with reduced susceptibility to ciprofloxacin (see table 2 appendix A1). The precise mechanisms behind this phenomenon remain unclear and certainly require much further consideration for two main reasons. First, ciprofloxacin is the antibiotic of choice in the event of antimicrobial intervention in treating advanced cases of Salmonellosis and secondly there is concern over the emergence of ciprofloxacin resistance in pathogens, apparently without the antibiotic being used. Understanding the

SID 5 (2/05) Page 23 of 25

mechanisms is likely to give insights to alternative measures that will reduce or eliminate this resistance emerging.

References to published material9. This section should be used to record links (hypertext links where possible) or references to other

published material generated by, or relating to this project.

SID 5 (2/05) Page 24 of 25

See relevant appendices for references cited in the text above

Publications arising from this project:-

Posters presented at ECCMID

K.A.G. Karatzas, M.A. Webber, L.J.V. Piddock, M.J. Woodward and T.J. Humphrey. 2006. Exposure of S. enterica serovar Typhimurium to disinfectants and development of multiple antibiotic resistance. In Proceedings of the 106th ASM General Meeting. Abstract and poster. Orlando, Florida, USA.

S. Cooles, M. A. Webber, L. P. Randall, N. G. Coldham, L. J. V. Piddock and M. J. Woodward. 2006. Does reduced susceptibility to disinfectants compromise colonisation and persistence of Salmonella enterica serovar Typhimurium in chickens? In Proceedings of the 16th European Congress of Clinical Microbiology and Infectious Disease. Abstract and poster. Nice, France. M. C. Bagnall, L. P. Randall, N. G. Coldham, A. Karatzas, T. Humphrey, L. J. V. Piddock and M. J. Woodward. 2007. Effect of multiple passages of Salmonella Typhimurium in the presence of disinfectants on susceptibility to antimicrobials, on persistence in the one-day-old chick model and efflux systems. 2007. In Proceedings of the 17th European Congress of Clinical Microbiology and Infectious Disease. Abstract and poster. Munich, Germany.

Papers

Webber MA, Woodward MJ and Piddock LJV. Disinfectant resistance in bacteria In 'Antimicrobial Resistance in Bacteria of Animal Origin' Edited by F. Aarestrup. 2005 Chapter 8. ASM press Washington. ISBN: 1-55581-306-2.

L. P. Randall, C. Clouting, F. A. Clifton-Hadley, R. H. Davies and M. J. Woodward. (2005). Farm disinfectants select for cyclohexane resistance, a marker of multiple antibiotic resistance, in Escherichia coli. J Appl Microbiol. 98:556-63.

L. P. Randall, S. W. Cooles, N. G. Coldham, E. G. Penuela, A. C. Mott, M. J. Woodward, L. J.V Piddock and M. A. Webber. Commonly used farm disinfectants can select for mutant Salmonella enterica serovar Typhimurium with decreased susceptibility to biocides and antibiotics. (submitted to AAC)

M. A. Webber, N. G. Coldham, S. Cooles, L. P. Randall, M. J. Woodward and Laura J.V.Piddock. Distinct mechanisms contribute to triclosan resistance in Salmonella enterica serovar Typhimurium. (Submitted to PNAS USA)

K. A. G. Karatzas, L. P. Randall, M. A. Webber, L. J.V. Piddock, T. J. Humphrey, M. J. Woodward and N. G. Coldham. Phenotypic and proteomic characterization of MAR derivatives of Salmonella enterica ser. Typhimurium selected following exposure to disinfectants. (submitted to Molecular Microbiology)

L. P. Randall, K. A. G. Karatzas, N. G. Coldham, L. J. V. Piddock and M. J. Woodward.. Spread of disinfectant passaged multiple antibiotic resistant (MAR) mutants of Salmonella enterica serovar Typhimurium in chicks and acquisition of fluoroquinolone resistance (in preparation for submission to JAC).

K. A. G. Karatzas, M. A. Webber, F. Jorgensen, M. J. Woodward, L. J.V. Piddock and T. J. Humphrey. Prolonged treatment of Salmonella enterica ser. Typhimurium with commercial disinfectants selects for multiple antibiotic resistance (MAR), increased efflux and reduced invasiveness. (in preparation for submission to JAC)

K. A. G. Karatzas, M. A. Webber, F. Jorgensen, M. J. Woodward, L. J.V. Piddock and T. J. Humphrey. Characterization of MAR isolates of Salmonella enterica ser. Typhimurium selected following prolonged treatment with triclosan. (in preparation for submission to JAC)

Book Chapters

M. A. Webber, M. J. Woodward and L. J. V. Piddock. (2005) Disinfectant resistance in bacteria In ‘Antimicrobial Resistance in Bacteria of Animal Origin’ Edited by F. Aarestrup. Chapter 8 pg 115-125. ASM press Washington. ISBN: 1-55581-306-2.

SID 5 (2/05) Page 25 of 25