Received 10 January 2005; returned 8 February 2005; revised 23 March 2005; accepted 5 April 2005

Journal of Antimicrobial Chemotherapy (2005) 55, 921 927 doi:10.1093/jac/dki146 Advance Access publication 9 May 2005 Phenotypic variability of Pseudomonas aeruginosa in sputa from patients with acute infective exacerbation of cystic fibrosis and its impact on the validity of antimicrobial susceptibility testing J. E. Foweraker 1 *, C. R. Laughton 1, D. F. J. Brown 2 and D. Bilton 3 JAC 1 Department of Microbiology, Papworth Hospital, Papworth Everard, Cambridge CB3 8RE, UK; 2 Health Protection Agency, Clinical Microbiology and Public Health Laboratory, Addenbrooke s Hospital, Cambridge CB2 2QW, UK; 3 Department of Chest Medicine, Papworth Hospital, Papworth Everard, Cambridge CB3 8RE, UK Received 10 January 2005; returned 8 February 2005; revised 23 March 2005; accepted 5 April 2005 Objectives: To investigate the variability in antimicrobial susceptibility of Pseudomonas aeruginosa from sputa of patients with cystic fibrosis, to compare testing individual colonies of the same morphotype either separately or combined and to study the reproducibility of testing antimicrobial susceptibility within and between laboratories. Methods: One hundred and one sputa were cultured. Four colonies of each P. aeruginosa morphotype were suspended. Susceptibility to 12 agents by disc diffusion was tested individually or by pooling the four suspensions. A sputum sample containing four morphotypes of one genotype of P. aeruginosa was used to study reproducibility. Susceptibility was tested in duplicate by eight biomedical scientists in one laboratory and by routine procedures in seven different laboratories. Results: There was a mean of four morphotypes of P. aeruginosa per sputum and three antibiograms per morphotype. In some cases, all four colonies of a single morphotype had different antibiograms. The susceptibility profiles of single isolates of P. aeruginosa correlated poorly with pooled cultures, with the pooled tests missing resistance. Results from one sample tested in duplicate by eight biomedical scientists in one laboratory and in seven other laboratories did not correlate well. The wide range of zone sizes in disc diffusion tests illustrated the variation in susceptibility of 48 colonies from one sputum sample. Conclusions: The role of conventional antimicrobial susceptibility testing is questionable once P. aeruginosa chronically infects the cystic fibrosis lung. A range of susceptibility patterns is seen, even within a morphotype. Routine test results are not reproducible and underestimate resistance. Keywords: CF, P. aeruginosa, susceptibility testing Introduction Pseudomonas aeruginosa is an important pathogen in patients with cystic fibrosis (CF). It causes acute episodes of infection and a more rapid decline in lung function than is seen in patients without the organism. With chronic infection, P. aeruginosa exists as multiple variants with a range of colony morphologies including the classical mucoid phenotype. Although colony types or morphotypes differ, they may have similar genetic composition as shown by pulse field gel electrophoresis. 1 A variety of antimicrobial susceptibility patterns can be found among isolates from the same sputum and consequently testing a single colony of P. aeruginosa may not be representative of the susceptibility of the whole population. Mixed morphotype testing has therefore been recommended. 2 For this, single colonies of different morphology (morphotypes) are suspended together to prepare an inoculum for one test of antimicrobial susceptibility. Published reports on mixed morphotype testing have not, however, investigated whether there is variation in antimicrobial susceptibility between individual colonies with the same... *Corresponding author. Tel: + 44-1480-364305; Fax: + 44-1480-364780; E-mail: juliet.foweraker@papworth.nhs.uk... 921 q The Author 2005. Published by Oxford University Press on behalf of the British Society for Antimicrobial Chemotherapy. All rights reserved. For Permissions, please e-mail: journals.permissions@oupjournals.org

Foweraker et al. morphotype, and do not indicate how many colonies of each morphotype should be picked for susceptibility testing. Controlled studies have shown that patients with acute exacerbations of CF with P. aeruginosa improve if treated with anti-pseudomonal antibiotics. 3 However, patients can respond to antibiotics to which the P. aeruginosa is reported as resistant. 4 This may be because the resistant isolate is not the cause of symptoms, either because it is a minor part of the infecting population or it is less pathogenic than other isolates in the mixture. In contrast, patients with isolates that on routine laboratory testing are fully susceptible to the antibiotics used may not improve. 5 This may be because resistant isolates in a mixed population are missed on selection of colonies for susceptibility testing. There were three objectives of our study: (i) to investigate the degree of variation in antimicrobial susceptibility within each sputum sample by testing multiple individual colonies of the same morphotype of P. aeruginosa; (ii) where there were differences in antimicrobial susceptibility between isolates of the same morphotype, to compare the results of tests on individual isolates with those obtained when the different isolates were mixed and susceptibility was measured in a single disc diffusion test; (iii) to investigate the reproducibility of testing sputum containing isolates with multiple susceptibility patterns, both within one laboratory when tests are set up by different staff and between different hospital laboratories Methods Specimens Sputum was collected from 40 episodes of acute infective exacerbation in 24 adults who were chronically colonized with P. aeruginosa (with a maximum of three exacerbations per patient). Sputum was collected just before antibiotics were started (day 0) and on days 6 7 and 13 14 of treatment. An exacerbation was defined as a significant fall in lung function, measured as >200 ml reduction in the forced expiratory volume in 1 min (FEV 1 ), together with three or more of: increased cough or shortness of breath; increased sputum volume; increased sputum purulence; haemoptysis; fever (>37.58C); new or increased crackles heard in the chest; new shadowing on chest X-ray; weight loss. Most patients received an initial treatment with intravenous ceftazidime plus tobramycin. Other combinations, usually of a different b-lactam plus an aminoglycoside, were used for patients who were allergic to ceftazidime. The antibiotics used were usually those to which the patient had responded in the treatment of a previous exacerbation. Sputum culture Sputum was processed within 4 h of expectoration. If culture was delayed for >30 min, the sputum was refrigerated at 48C. Each sputum was homogenized with an equal volume of 0.1% dithiothreitol (Pro-Lab Diagnostics, Neston, UK). The homogenized specimen was diluted 1 in 4000 and 1 in 40 000 in sterile distilled water (SDW). It was inoculated using a spiral plater (Don Whitley Scientific, Shipley, UK) on to heated blood agar (HBA; PHLS Media Services, Leeds, UK), pseudomonas agar base with cetrimide, fusidic acid and cephaloridine selective agar supplement (PCFC agar) and pseudomonas agar base with cetrimide and sodium nalidixate supplement (PCN agar). Plates were incubated for 48 72 h at 378C in either 5% CO 2 (HBA) or air (PCFC & PCN). PCFC and PCN agars were supplied by Oxoid Ltd, Basingstoke, UK. Aliquots of the homogenized sputum were added to an equal volume of glycerol broth (20% glycerol in nutrient broth) and stored at 808C. Total viable counts of bacteria were calculated from the growth on PCFC and PCN agars. Plates were read using the Sorcerer automated colony counting system (Perceptive Instruments, Haverhill, UK). Bacteria that were Gram-negative, oxidase positive and were mucoid or pigmented on pseudomonas selective agar were identified as P. aeruginosa. Bacteria without pigment or a mucoid phenotype were identified using API 20NE (Biomerieux, Basingstoke, UK). Bacteria that could not be speciated by these methods were identified by the Laboratory of Healthcare Associated Infection, Health Protection Agency, Colindale, UK. A single observer described different morphotypes of P. aeruginosa, by their appearance on primary growth on PCFC on the basis of size, texture, colour and mucoidity. The amount of growth as a proportion of the total was estimated for each morphotype. Four colonies of each morphotype were selected and each individual colony was subcultured on HBA, suspended in glycerol broth and stored on sterile glass beads at 808C. Antimicrobial susceptibility testing Individual colony (method 1). Four colonies of each morphotype were picked separately, subcultured on HBA to check for purity and individually tested for antimicrobial susceptibility by the BSAC standardized disc diffusion method. 6 The antimicrobial agents tested were amikacin, aztreonam, ceftazidime, colistin, ciprofloxacin, gentamicin, imipenem, meropenem, netilmicin, piperacillin, piperacillin/tazobactam and tobramycin. Zone diameters were measured with electronic callipers (Bowers Metrology, West Drayton, UK) with the result automatically inserted into an Excel spreadsheet for analysis. Susceptibility was interpreted according to BSAC criteria. Mixed multiple colonies of the same morphotype (method 2). Equal volumes of the four suspensions prepared for susceptibility testing of individual colonies of the same morphotype were mixed to test the combined susceptibility using the BSAC disc diffusion method as above. Reproducibility of CF sputum susceptibility testing procedures In order to examine the reproducibility of susceptibility testing of isolates from one sputum, we used a frozen sputum sample that contained a variety of morphotypes of one genotype of P. aeruginosa. Four examples of each of the morphotypes from this sample had previously been compared using Xba I genome macrorestriction analysis resolved by PFGE. The restriction fragment profiles all had >85% Dice coefficients of similarity using the unweighted pair group method with arithmetic averages. The specimen was thawed and re-cultured to confirm that the morphotypes found on processing the fresh specimen were still present. The sputum was diluted 1 in 1000 in SDW and 10 ml of this was plated onto multiple sets of HBA, Blood and PCFC agars. The plates were incubated for 24 h at 378C in5%co 2 (Blood, HBA) or air (PCFC). They were checked to ensure that all sets were comparable for density of growth and that three morphotypes were present (a fourth morphotype was known to take more than 48 h to culture and was not visible after incubation for 24 h). Plates were distributed in order to examine the reproducibility of testing within one clinical diagnostic laboratory and the reproducibility between seven different laboratories. For the former, two sets 922

Testing susceptibility of P. aeruginosa in CF sputum of plates were given to eight biomedical scientists in the diagnostic clinical laboratory at Papworth Hospital with instructions to process the plates further according to the local protocol. The biomedical scientists were not told that the two sets of plates were from the same sputum. In order to test reproducibility between laboratories, sets of plates were sent via routine in-house road transport to seven laboratories in the Eastern Region of England. Six laboratories received the plates the same day they were dispatched, the seventh the next day. A control set of plates was returned to the organizing laboratory the same day via the same transport system. It was requested that only the methods routinely used in each laboratory for testing and recording the results should be used. A questionnaire was sent with the plates to record the methods used for picking isolates and susceptibility testing. After the eight biomedical scientists at Papworth Hospital had read the susceptibility plates with a template, the zone diameters from all 16 sets of susceptibility tests were independently measured with callipers to provide information on the degree of susceptibility and variability in the zone sizes of the isolates examined. Results One hundred and one sputa from 40 exacerbations in 24 patients were examined, 40 taken just before starting antibiotics (day 0), but only 29 from days 6 7 and 24 from days 13 14 as not all patients continued to expectorate once on treatment. In some patients, symptoms did not resolve after antimicrobial treatment for 14 days and they either continued on the same drugs or changed to different antibiotic treatment for a further 14 21 days. Six sputa were received from days 20 21 and two after 28 days of treatment. All but five sputa contained P. aeruginosa. Alcaligenes spp. were cultured from six sputa, two mixed with P. aeruginosa and one with Stenotrophomonas maltophilia. One organism could not be identified by the reference laboratory but was closest to Pandoraea sp. by sequence analysis. As the poor penetration of anti-pseudomonal antibiotics leads to low concentrations in sputum, isolates of intermediate susceptibility were included with resistant isolates to simplify some of the analyses. Susceptibility related to morphotype Of the sputa with P. aeruginosa, all but one had more than one morphotype. In only one sample did all isolates of all morphotypes have the same antimicrobial susceptibility profile. All other sputa had P. aeruginosa with a variety of susceptibility patterns in the same sample, including differences between colonies of the same morphotype. There was the same degree of variability seen after 7 and 14 days of antibiotic treatment as in the sputa taken just before starting treatment. Overall, there was a mean of four morphotypes of P. aeruginosa in each sputum (with a range of 1 6). There was a mean of six antibiograms (range 1 12). The mode number of antibiograms per morphotype in each sample was three with a range of 1 4. In some sputa, each of four colonies of a single morphotype had a different susceptibility pattern. Susceptibility tested individually (method 1) and in combination (method 2) There was a poor correlation when the results of testing the colonies singly were compared with the susceptibility result of the combined suspension of the colonies of each morphotype of P. aeruginosa. The overall susceptibility results were the same by both methods for only 11 out of 60 sputa. Resistance to one or more antibiotics was missed by combination testing in 45 sputa, whereas when colonies were tested individually, resistance was missed in four sputa. The results were analysed for the sputa from all exacerbations and also by limiting the analysis to the examination of one exacerbation per patient in case any individual had several exacerbations with an unusual organism that could lead to bias. This was not found to be the case. A more detailed analysis is shown in Table 1. This gives the results for the different antibiotics, showing the number of sputa in which a resistant P. aeruginosa was missed using the different methods of susceptibility testing. In most cases, antibiotic resistance was missed in the mixtures, but occasionally a resistant phenotype was detected only in combined testing. Differences were seen with all antibiotics. Reproducibility of susceptibility testing by mixed multiple colony method on the same morphotype in one laboratory None of the eight individuals processing the plates identified and tested the dwarf morphotype that took 72 h to grow. The combined results of the different antimicrobial susceptibility profiles of the isolates from each set of plates gave an overall antibiogram. No individual had the same susceptibility results for the same sputum tested twice (Table 2). The distribution of zone diameters for P. aeruginosa in the 16 replicated sputum samples indicates that the variation in susceptibility was not just a result of the zone diameters being close to susceptibility breakpoints, but rather that there was a range of resistant and susceptible organisms within the same population (Figure 1). Table 1. Failure to detect an antibiotic-resistant P. aeruginosa in a sputum sample when mixtures of colonies of each morphotype were tested (method 2) compared with testing the same colonies individually (method 1) (60 sputum samples contained 274 morphotypes of P. aeruginosa) Antibiotic Number of sputa where P. aeruginosa resistant to an antibiotic was missed (n = 60) mixtures of colonies of each morphotype tested all individual colonies of each morphotype tested Amikacin 15 6 Aztreonam 12 2 Ceftazidime 17 5 Ciprofloxacin 9 1 Colistin 1 0 Gentamicin 14 5 Imipenem 9 4 Meropenem 13 0 Netilmicin 12 3 Piperacillin 7 6 Piperacillin/tazobactam 15 6 Tobramycin 11 2 923

Foweraker et al. Table 2. Susceptibility of P. aeruginosa from one sputum sample tested in duplicate by eight different biomedical scientists (the results are for the most resistant morphotype reported for each of the duplicate tests) Number of reports of overall susceptibility a in duplicate tests (n =8) Antimicrobial agent S, S S, I S, R R, I R, R Amikacin 0 0 0 2 6 Aztreonam 0 0 0 0 8 Ceftazidime 1 0 4 1 2 Ciprofloxacin 8 0 0 0 0 Colistin 8 0 0 0 0 Gentamicin 0 0 0 0 8 Imipenem 0 0 2 0 6 Meropenem 0 0 5 0 3 Netilmicin 0 0 0 3 5 Piperacillin 5 0 3 0 0 Piperacillin/tazobactam 8 0 0 0 0 Tobramycin 0 2 1 2 3 a S, susceptible; I, intermediate; R, resistant. Methods used in seven diagnostic clinical laboratories All laboratories re-incubated the primary plates, either overnight or for a further 24 h. For susceptibility testing, five laboratories used the BSAC standardized disc diffusion method, one used Stokes method and one used a breakpoint method for their firstline antibiotics and BSAC disc diffusion for their second-line antibiotics. Five used P. aeruginosa NCTC 10662 as a control organism and two used P. aeruginosa ATCC 27853. Five incubated the plates overnight and three for 24 h. In two laboratories, incubation was extended for a further 24 h if there was no growth overnight. The five laboratories using only the BSAC method read zone sizes with a template, the laboratory using BSAC for second line antibiotics measured the zone sizes. For susceptibility testing, a single colony of each morphotype was tested in four laboratories and in three laboratories several identical colonies of each type were tested. The biomedical scientists processing the samples were asked to note the details of the method used. On two occasions the method described was different from that laboratory s standard operating procedure (SOP) for handling sputum from patients with CF. For example, in one case the SOP required the combination of several identical colonies for each test, but the biomedical scientist tested a single colony of each morphotype. Contrary to the BSAC recommendation, two laboratories used 10 mg instead of 30 mg amikacin discs and four used 300 mg polymyxin B instead of 25 mg colistin discs. Reproducibility of susceptibility testing in seven diagnostic clinical laboratories There was considerable variation in results from the seven laboratories examining cultures plated from a single sputum containing four morphotypes of P. aeruginosa. All laboratories missed the slow-growing dwarf morphotype. In two laboratories, only a mucoid morphotype was recognized as P. aeruginosa. In one laboratory, four colonies of the mucoid morphotype were tested separately and in another, just one colony of the mucoid Figure 1. Zone diameters in disc diffusion susceptibility tests on multiple isolates of Pseudomonas aeruginosa cultured from a single sputum. The length of each line represents the number of isolates with that zone diameter. Antibiotics tested were: amikacin (AMK); aztreonam (ATM); ceftazidime (CAZ); ciprofloxacin (CIP); colistin (CT); gentamicin (GEN); imipenem (IMP); meropenem (MEM); netilmicin (NET); piperacillin (PIP); piperacillin/tazobactam (TZP); and tobramycin (TOB). 924

Testing susceptibility of P. aeruginosa in CF sputum Table 3. Susceptibility of isolates of P. aeruginosa (pooled results) from the same sputum sample tested in seven clinical diagnostic laboratories morphotype was picked for susceptibility testing. Three laboratories recognized two morphotypes one mixed five colonies of each morphotype for susceptibility testing and two laboratories selected only one colony per morphotype. Two laboratories recognized three morphotypes one picked one colony per morphotype and another mixed 2 3 colonies to test each morphotype. Not all laboratories tested the same antibiotics. The agents tested (number of laboratories testing antibiotic) were as follows: amikacin (6), aztreonam (6), ceftazidime (7), ciprofloxacin (7), colistin or polymyxin (6), gentamicin (7), imipenem (2), meropenem (7), netilmicin (3), piperacillin/tazobactam (5) and tobramycin (6). Consistent results were found for colistin (polymyxin), piperacillin/tazobactam, and imipenem only (Table 3). Discussion Number of laboratories reporting a Antibiotic S I R NT Amikacin 0 1 5 1 Azlocillin 1 0 0 6 Aztreonam 1 0 5 1 Ceftazidime 4 0 3 0 Ciprofloxacin 3 0 4 0 Gentamicin 1 0 6 0 Imipenem 0 0 2 5 Meropenem 3 0 4 0 Netilmicin 0 1 2 4 Piperacillin 1 0 0 6 Piperacillin/tazobactam 5 0 0 2 Polymyxin/colistin 6 0 0 1 Tobramycin 2 1 3 1 a S, susceptible; I, intermediate; R, resistant; NT, not tested. We have shown far more variation in the susceptibility of individual colonies of P. aeruginosa in the sputum of a patient with an acute exacerbation of CF than has been previously reported. The variation was not just a consequence of borderline susceptibility results that can fall either side of the breakpoint on repeated testing. In many cases, organisms of the same morphotype had different antibiograms. Occasional mucoid morphotypes with a fairly uniform susceptibility profile were seen, but even where a mucoid morphotype was susceptible to an antibiotic, there was a wide variation in zone size between different mucoid colonies picked from the same sputum. Some variation in susceptibility of P. aeruginosa from patients with CF has been reported previously. In 1979, Thomassen et al. 7 reported that different morphotypes of P. aeruginosa cultured from CF sputum had different susceptibility profiles in 51% of sputa, but individuals of the same morphotype with a different susceptibility were only found in 15% of samples. In most cases, the variation was attributed to zones of inhibition close to breakpoints. However, the study was limited by the small range of antibiotics tested (tobramycin, carbenicillin, gentamicin, kanamycin and tetracycline) and others have shown that changes in susceptibility in vitro can occur without the morphology changing. 8 Recommended methods for susceptibility testing have required mixing different morphotypes of P. aeruginosa for susceptibility testing. The standard method published by the UK Health Protection Agency recommends the use of the BSAC disc diffusion method for testing mixed multiple colonies from the growth on primary isolation plates and to report the susceptibility of the most resistant. 6 The UK Cystic Fibrosis Trust advises sampling a representative of each morphotype, and testing by the BSAC disc method. 9 However, it is unclear if a single colony of each morphotype should be mixed with single colonies of other morphotypes for one susceptibility test, or representative colonies should be tested individually. Our results suggest that both recommendations will lead to a major underestimate of antibiotic resistance. Some authors have considered that mixing P. aeruginosa of different morphotypes to perform a single disc susceptibility test is reasonably predictive of the susceptibility pattern of the most resistant isolate tested singly. Similar results were achieved when single colonies were subcultured and tested individually, or when the individual colony suspensions were combined in equal volumes for mixed morphotype testing. 2 If, however, the suspension was formed by mixing colonies of each morphotype picked directly from the primary isolation plates there was less correlation with results where different examples of each morphotype were tested singly. This is not unexpected, as we have shown that colonies of the same morphotype can have different susceptibility patterns. In another study, the broth microdilution MIC for each morphotype was tested separately with the inoculum prepared by suspending 5 10 colonies of the same morphotype. This was compared with the MIC of a mixture of equal proportions of each suspension (of each morphotype) to give a mixed morphotype test. 10 In each case, the MIC of the mixed morphotype test was within two dilutions of the most resistant single morphotype. Again the authors did not examine the variation in susceptibility between individuals of the same morphotype. Others have shown that mixed morphotype testing in a broth microdilution MIC test poorly predicted the susceptibility of the most resistant isolate tested separately, this time using the criterion of MICs greater than one dilution being different. 11 Resistance found in individual examples of each morphotype was missed in mixed morphotype testing. We have shown that there is so much variation in the susceptibility of individual colonies of the same morphotype that picking and mixing only single colonies of each morphotype would lead to an underestimate of antibiotic resistance. In addition, our results suggest that even mixing several colonies of each morphotype for mixed single morphotype testing is unreliable. It is not clear why purified single colonies of the same morphotype give a different susceptibility result when in a mixed suspension. There may be interactions between bacteria, although this would be expected to be less likely when susceptibility is tested by agar diffusion rather than in broth dilution. Alternatively, the individual phenotypes may be unstable. The consequence of trying to test susceptibility of such a mixed population of P. aeruginosa is seen in the results of repeated testing of the same sample by different individuals in the same diagnostic clinical laboratory. There were differences in susceptibility between tests set up by different staff and even when the same individual tested the same sputum twice. Results apparently 925

Foweraker et al. depended on which colonies of each morphotype are picked to test. The major differences in the results from seven diagnostic laboratories testing the same sample were not surprising given the variation between individual colonies of P. aeruginosa of one genotype in a single sputum and the differences in the methods of colony selection. In addition, the practice in some laboratories of re-incubating disc diffusion susceptibility plates if there was no growth after overnight incubation could lead to resistant bacteria appearing susceptible, as very slow growth may result in increased zone diameter. On the basis of the antibiograms detected by each laboratory, there would be unacceptable differences in the antibiotics that could be recommended to the clinician by the different laboratories (Table 4). Variability within a single genotype of P. aeruginosa in CF sputum is not just seen in antimicrobial susceptibility, but has also been shown for alginate production, where bacteria with different mutations leading to mucoidity can be demonstrated within the same sputum sample (Professor J. Govan, University of Edinburgh, personal communication), and for type III secretion phenotypes. 12 P. aeruginosa is known for its high level of adaptability, especially in chronic infection in the CF lung, where organisms live in biofilms as well as in a planktonic state. There has been particular interest in the dwarf phenotype, also known as small colony variant (SCV), which is more resistant to antibiotics and has the characteristics of hyperpilation and increased adherence to host cells that would make them likely to form biofilms. 13 Antibiotic resistance in SCVs has been related to changes in the growth environment, and a gene switch mechanism similar to that seen in phase variation in other bacteria has been proposed. 13 Others suggest that variation in susceptibility in SCVs is due to mutation and selection. 14 It was notable that in our reproducibility study, no laboratory extended the incubation of the primary plates for long enough to pick up the SCVs that took 72 h to grow. Most laboratories incubate culture plates from CF sputum for a maximum of 48 h. In a recent study, it was found that 11 out of 30 CF patients were colonized by a hypermutatable strain of P. aeruginosa, whereas no hypermutators were found in 75 non-cf patients acutely infected with P. aeruginosa. 15 This ability to mutate frequently promotes diversification and adaptive mutations such as Table 4. Appropriate antibiotics for use based on susceptibility results that would be reported to a clinician from seven different hospital clinical laboratories testing the same sputum sample (based on data from the reproducibility study see Table 3) Laboratory Agents to which isolates were reported susceptible 1 aztreonam, ceftazidime, ciprofloxacin, meropenem, tobramycin, gentamicin 2 colistin, piperacillin/tazobactam 3 colistin, piperacillin/tazobactam 4 colistin, piperacillin/tazobactam, meropenem 5 ceftazidime, ciprofloxacin, colistin, piperacillin/ tazobactam 6 ceftazidime, ciprofloxacin, colistin, meropenem, piperacillin, piperacillin/tazobactam 7 azlocillin, ceftazidime, colistin, tobramycin resistance to antibiotics. Hypermutation occurring during culture in vitro would not explain the degree of variation seen in our work, as hypermutation produces variants at a rate of one in 1000 at most. In the CF sputum, bacteria grow in a wide range of conditions in different parts of a biofilm and will be exposed to different effective antibiotic concentrations. It may be that in each micro-niche there is a different selective pressure for a particular level of resistance, including the situation where organisms are fully protected from the effects of antibiotics because of the local environment or their growth phase. Hence, the bacterial population in one patient could contain bacteria with a range of antimicrobial susceptibility. Expectorated sputum contains fragments of different parts of the biofilm as well as planktonic forms, so bacteria with a wide range of antimicrobial susceptibility patterns may be found. As the patient receives different courses of antibiotics, bacteria with different combinations of susceptibility to a range of antibiotics could evolve. There has been little published on the implications for testing in routine clinical laboratories of the variation in antimicrobial susceptibility in a sputum sample. It was suggested that to get a representative picture, up to 20 colonies should be tested from each specimen. 16 This was deemed to be impractical for routine laboratories, but no alternative approach for accurate analysis was proposed. 17 We have found that in some cases, all four colonies of one morphotype had different susceptibility patterns, and therefore even testing four colonies of each morphotype may not be enough. Mixing individuals of the same morphotype for a combined diffusion susceptibility test does not solve the problem. Present testing methods in diagnostic laboratories miss resistant populations of P. aeruginosa in CF sputum, but in order to improve reliability and reproducibility even more individual colonies may need to be tested than in our study. It has been questioned whether antibiotic resistance in vitro relates to treatment failure and whether conventional antimicrobial susceptibility testing has any place in patient management once P. aeruginosa is established in the CF lung as a chronic infection. 5 The correlation between measuring antibiotic susceptibility using our proposed extensive testing method and the response of the patient to the recommended antibiotics should be established, and is the subject of a current study. Other options that need further investigation are direct sputum testing, testing susceptibility in biofilms and examining antibiotic combinations (synergy testing). 18 21 Results of synergy testing of P. aeruginosa must be critically assessed. Different methodology (kill curves, chequerboard), can give different answers 22 and we do not know if the results of synergy tests on different bacteria from a single sputum will be the same. Any susceptibility tests of antibiotic combinations or in biofilms must take into account the variability and therefore the need for adequate sampling of this complex population of P. aeruginosa in the CF lung. Again, the correlation between measuring antibiotic susceptibility using these methods and the response of the patient to the recommended antibiotics should be tested. Acknowledgements This study was supported by a grant from the UK Cystic Fibrosis Trust. We thank Dr L. Sharples and Miss F. Cafferty of the MRC Biostatistics Unit, Cambridge, for their advice and assistance. 926

Testing susceptibility of P. aeruginosa in CF sputum References 1. Breitenstein S, Walter S, Bosshammer J et al. Direct sputum analysis of Pseudomonas aeruginosa macrorestriction fragment genotypes in patients with cystic fibrosis. Med Microbiol Immunol 1997; 186: 95 9. 2. Wolter J, Kotsiou G, McCormack J. Mixed morphotype testing of Pseudomonas aeruginosa cultures from cystic fibrosis patients. J Med Microbiol 1995; 42: 220 4. 3. Regelmann W, Elliott G, Warwick W et al. Reduction of sputum Pseudomonas aeruginosa density by antibiotics improves lung function in cystic fibrosis more than do bronchodilators and chest physiotherapy alone. Am Rev Resp Dis 1990; 141: 914 21. 4. Smith A, Doershuk C, Goldmann D et al. Comparison of a b-lactam alone versus b-lactam and an aminoglycoside for pulmonary exacerbation in cystic fibrosis. J Pediatr 1999; 134: 413 21. 5. Smith A, Fiel S, Mayer-Hamblett N et al. Susceptibility testing of Pseudomonas aeruginosa isolates and clinical response to parenteral antibiotic administration. Lack of association in cystic fibrosis. Chest 2003; 123: 1495 502. 6. Andrews J. BSAC standardized disc susceptibility testing method (version 3). J Antimicrob Chemother 2004; 53: 713 28. 7. Thomassen M, Demko C, Boxerbaum B et al. Multiple isolates of Pseudomonas aeruginosa with differing antimicrobial susceptibility patterns from patients with cystic fibrosis. J Infect Dis 1979; 140: 873 80. 8. Fyfe J, Govan J. Chromosomal loci associated with antibiotic hypersensitivity in pulmonary isolates of Pseudomonas aeruginosa. J Gen Microbiol 1984; 130: 825 34. 9. Cystic Fibrosis Trust. Antibiotic treatment for cystic fibrosis: Report of the UK Cystic Fibrosis Trust s Antibiotic Group. Bromley, Kent: CF Trust, 2000. 10. Van-Horn K. Mixed-morphotype broth microdilution susceptibility testing of Pseudomonas aeruginosa from cystic fibrosis patients. J Clin Microbiol 1993; 31: 458 9. 11. Morlin G, Hedges D, Smith A et al. Accuracy and cost of antibiotic susceptibility testing of mixed morphotypes of Pseudomonas aeruginosa. J Clin Microbiol 1994; 32: 1027 30. 12. Jain M, Ramirez D, Seshardri R et al. Type III secretion phenotypes of Pseudomonas aeruginosa strains change during infection of individuals with Cystic Fibrosis. JClinMicrobiol2004; 42: 5229 37. 13. Drenkard E, Ausubel F. Pseudomonas biofilm formation and antibiotic resistance are linked to phenotypic variation. Nature 2002; 416: 740 3. 14. Haussler S, Ziegler I, Lottel A et al. Highly adherent small-colony variants of Pseudomonas aeruginosa in cystic fibrosis lung infection. J Med Microbiol 2003; 52: 295 301. 15. Oliver A, Canton R, Campo P et al. High frequency of hypermutable Pseudomonas aeruginosa in cystic fibrosis lung infection. Science 2000; 288: 1251 4. 16. Govan J, Nelson J. Microbiology of lung infection in cystic fibrosis. Br Med Bull 1992; 48: 912 30. 17. Gilligan P. Microbiology of airway disease in patients with cystic fibrosis. Clin Microbiol Rev 1991; 4: 35 51. 18. Serisier DJ, Jones G, Tuck A et al. Clinical application of direct sputum sensitivity testing in a severe infective exacerbation of cystic fibrosis. Pediatr Pulmonol 2003; 35: 463 6. 19. Ceri H, Olson ME, Stremick C et al. The Calgary biofilm device: new technology for rapid determination of antibiotic susceptibilities of bacterial biofilms. J Clin Microbiol 1999; 37: 1771 6. 20. Lang BJ, Aaron SD, Ferris W et al. Multiple combination bacteriocidal antibiotic testing for patients with cystic fibrosis infected with multiresistant strains of Pseudomonas aeruginosa. American J Resp Crit Care Med 2000; 162: 2241 5. 21. Muscovitz SM, Foster JM, Emerson J et al. Clinically feasible biofilm susceptibility assay for isolates of Pseudomonas aeruginosa from patients with cystic fibrosis. J Clin Microbiol 2004; 42: 1915 22. 22. Cappelletty DM, Rybak MJ. Comparison of methodologies for synergism testing of drug combinations against resistant strains of Pseudomonas aeruginosa. Antimicrob Agents Chemother 1996; 40: 677 83. 927