Original Report

Cluster Analysis on Multiple Drugs Susceptibility Supplements Genotyping of Methicillin-Resistant Staphylococcus aureus .Junichi Yoshida, MD, MS, FACS; * Akiko Umeda, MD; + Toshiyuki Ishimaru, MD; * -and Motoichi Akao, MD*

ABSTRACT

Objective: To evaluate the typing power of cluster analysis of antimicrobial susceptibility.

Methods: Results of pulsed-field gel electrophoresis in 71 strains of methicillin-resistant St@~y/ococcus aureus were com- pared with cluster analysis of the diameter of growth inhibition in 11 drugs. Subjects were a consecutive series of patients (n = 71) from the wards and outpatient units of a community teaching hospital.

Results: The cluster analysis took 2 to 3 seconds once the data were entered into a computer. The sensitivity, specificity, and accuracy of the cluster analysis were 76.3%, 58.3%, and 73.2%, respectively, using genotyping as the reference.

Cc~nc/usions: The cluster analysis offered real-time epidemio- logic data at minimal cost and labor, warranting its cost-effec- tive role.

Key Words: antibiotyping, compute6 pulsed-field gel electrophoresis, Staphylococcus aureus

Int J Infect Dis 2001; 5:205-208.

Ichiyama et al reported on the usefulness of pulsed-field gel electrophoresis (PFGE) in the hospital epidemiology of methicillin-resistant StaphyZococcus aureus (MRSA) while referring to the limitation of antibiotyping as an epidemiologic probe. ’ However, antibiotyping, in the form of cluster analysis, may be a practical method, being inex- pensive and immediately ready upon the laboratory

*Infection Control Committee, Shimonoseki City Hospital, Shimono- seki, ‘Department of Bacteriology, Kyushu University Faculty of Medi- cine, Fukuoka, Japan.

Supported in part by a Special Research Fund for Infection Control, City of ShimonosekiJapan.

Address correspondence to Dr. JunichiYoshida, Shimonoseki City Hospital, l-13-1 Koyo-cho, Shimonoseki 750-8520, Japan. URL http://www.yoshi- daj.com.

report of susceptibility. The paucity of reports comparing PFGE and cluster analysis of antimicrobial susceptibility prompted the current study.

MATERIALS AND METHODS

Initially, subjects were a consecutive series of 72 isolates recovered from a total of 72 patients in a community hos- pital during the period May to July 1998. Later, one strain was accidentally contaminated in susceptibility; thus 71 strains remained for comparison. To ensure that selected sample were of varied origins, 55 strains were submitted from within the wards and 16 from outpatient units. Cul- ture samples were collected prior to initiation of chemotherapy, and only one sample per patient, recov- ered at first isolation of MRSA, was included. Methicillin- resistant S. duress was defined as S. aureus that was positive by Methicillin-Resistant ID Kit Screen Agar (Bec- ton Dickinson, Co., Ltd., Tokyo, Japan) but negative by oxacillin disk-diffusion.

Monodisk susceptibility results for 11 drugs were used for computation: ampicillin (10 kg), combination of ampicillin (10 kg) and sulbactam (10 pg), oxacillin (10 kg), pipemcillin (100 Fg), gentamicin (10 yg), arnikacin (30 pg), minocycline (30 kg), clindamycin (2 kg), lev- ofloxacin (5 kg), cefazolin (30 kg), and ceftizoxime (30 kg) (Becton-Dickinson Japan, Tokyo, Japan). The diame- ters of growth inhibition circles were measured at 24 hours and the data were entered into a personal com- puter (Presario, Compaq Japan, Tokyo, Japan). Hierarchi- cal cluster analysis was performed using SPSS 10.0 software (SPSS Inc., Chicago, IL).

The diameters of the zones of inhibition (in milli- meters) were used for cluster analysis. In the hierarchi- cal cluster analysis, these diameter data were used as the variables for the 11 drugs. When computing, the default setting of the software was used except in the output expression of the dendrogram. The default setting com- putes squared Euclidean distance, the sum of the squared distances over all the variables, for similarity correlation. For agglomeration schedule, the default between-groups linkage was used. In the dendrograms, the coefficients of

205

brought to you by COREView metadata, citation and similar papers at core.ac.uk

provided by Elsevier - Publisher Connector

206 International Journal of Infectious Diseases / Volume 5, Number $2001

******HIERARCHICAL CLUSTER ANALYSIS******

Dendrogram using Average Linkage (Between Groups)

Resealed Distance Cluster Combine

Figure 1. Computer printout of the dendrogram for 71 MRSA strains created by the diameters of growth inhibition circles in 11 drugs. Label = typing by pulsed-field gel electrophoresis; A-l to A-9 = superim- posed data describing cluster analysis grouping; Num = number of strains.

similarity were resealed from 0 to 25 along the hori- zontal axis called “Resealed Distance Cluster Combine” (Figure 1).

The same colony of MRSA undergoing the disk-dif- fusion tests was stored in preservation media (Keep Medium 0.8 mL; Nikken Biomedical Laboratory, Kyoto, Japan). Later chromosomal DNA was prepared as previ- ously described,2 digested with SmaI (Toyobo Co., Osaka, Japan). Then blocks were loaded onto an agarose 1% gel

in a PFGE apparatus (Bio-Rad Laboratories, Life Science Group, Hercules, CA). The pulse times of electrophoresis were 20 to 35 seconds for 20 hours and 40 to 70 seconds for the subsequent 7 hours at 5 V/cm. With a PFGE marker Lambda Ladder (New England, BioLabs, Beverly, MA) as a reference, the locations of bands were recorded in a computer and the hierarchical cluster analysis was performed. Finally the clonal identity was determined, permitting a deviation of up to three bands.

The cluster analysis of antibiotypes was evaluated by its sensitivity, specificity, and accuracy to determine clonal identity detected by PFGE. In detail, the true positive (TP) was defined as single clonal identity within a cluster where “Resealed Distance Cluster Combine” equalled 1. If a cluster represented a second group of any clonal group, all the strains were regarded as failure. Conversely the true negative (TN) was defined as unclassifiable geno- types for strains showing Resealed Distance Cluster Com- bine equalled 2 or more. Given the above,

Sensitivity = TP /(TP + FN) Specificity = TN /(TN + FP) Accuracy = (TP + TN)/(TP + FN + TN + FP)

where FN is false negative and FP false positive.

RESULTS

The 72 strains were divided into two groups for overnight growth inhibition tests; thus, in terms of time consumed, it took a total of 4 days to test the 11 antibiotics to gen- erate the antibiogram data. At computation, a single run took 2 to 3 seconds once the antibiotype data were entered into the computer. Fifty-nine strains were demon- strated within nine clusters, which were labelled A-l to A-9 (see Figure 1). Every cluster showed a characteristic profile of antibiogram (Table 1).

By PFGE, 53 strains were classified into seven chro- mosomal groups, namely A to G (Figure 2). Because one strain was lost in susceptibility, the classification of the remaining 52 strains was superimposed in the dendro- gram (see Figure 1).

The sensitivity, specificity, and accuracy of the clus-

ter analysis of susceptibility were 76.3%, 58.3%, and 73.2%, respectively. At evaluation, all the strains in groups

A-2 and A-4 were regarded as failure without matching chromosomal groups.

DISCUSSION

Epidemiologists in general consider that antibiotyping of MRSA is an inappropriate probe in endemic MRSA for the following reasons: (1) bacterial phenotypes are vul- nerable to change, (2) resistance against antimicrobials is transferable between organisms, and (3) antimicrobial susceptibility is subject to testing conditions in vitro and

* 1 2 3 4 5 6 7 8 9 10 11 1213 14 15 16 17 1819

A A A AAAA BBE 8 C CC CCC CC

Cluster Analysis of Antimicrobial Susceptibility / Yoshida et al 207

* 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 F F DDDDDDBDDDDDEE

* 20 21 22 23 24 25262728 2930 31 32 33

c cc C B B B EE B

* 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72

EE E8 ccc 0 6 G G

Figure 2. A-D, Actual PFGE of MRSA strains in sequence. Upper row increment of 50 kilobasepairs.

antimicrobial administration in vivo. To circumvent these shortcomings, in the present study, strains were collected over a limited period for the first detection of MRSA, to avoid antimicrobial pressure and subsequent phenotypic changes.

The authors have demonstrated the efficacy of clus- ter analysis using antibiogram and biotypes,” although the method suffered from a lack of bacteriologic evi- dence of typing or PFGE. At a community hospital in a

= strain number; lower row = chromosomal groups; * = marker with an

region where PFGE is scarcely available, cluster analysis of isolates of MRSA may be an invaluable supplement to PFGE, with methodology amenable to every hospital worldwide.

The prerequisites for this assumption may, however, be the use of multiple drugs for susceptibility testing. In the present study, 11 drugs of various categories of antimicrobials were used, because preliminary data showed that the more drugs, the better the accuracy of

208 International Journal of Infectious Diseases / Volume 5, Number 4,200I

Table 1. Antibiotic Susceptibility Profiles of MRSA Strains for Antibiotypes A-l to A-10

Drugs A-l A-2 A-3 A-4 A-5 A-6 A-7 A-8 A-9 A-10

Ampicillin R R R R R R R R R R Ampicillin/sulbactam R I I R R I I R I I Oxacillin R R R R R R R R R R Piperacillin R R R R R R R R R R Cefazolin R R R R R R R R R R Ceftizoxime R R R R R R R R R R Gentamicin S S S S R I S I S R Amikacin

6 I I R R R I I I I

Minocycline I S R S R R R R S Clindamycin R R R R R R R R R S Levofloxacin R R R R R I R S I I

R = resistant; I = intermediate: S = sensitive, all based on the standard of National Committee for Clinical Standards (NCCLS).

determining chromosomal typing. It has been reported that resistance against minocycline is transmitted by chromosomal gene tetM whereas plasmids as well as chromosomes are implicated in resistance against amino glycosides and clindamycin.* Pulsed-field gel elec- trophoresis is known to depict severed fragments of chromosomes, the difference of which would result in phenotypic differences. The difference in plasmids would also be expressed in the form of multiple antibiotypes as chromosomal group B appeared in antibiotypic group A-6 and separate group A-4.

In the study by Ichiyama and colleagues on geno- typing of MRSA, antibiotyping was performed using nine drugs, the susceptibility of which was classified into resis- tant, intermediate, and sensitive. Cluster analysis of their data would yield three clonal groups that would match three clusters (see Table 1)’ Thus, antibiotyping when analyzed by cluster analysis would augment discrimina- tory power even though using digitized data. Giacca and colleagues and Blanc and co-workers compared the clus- ter analysis of antibiograms with ribotyping and demon- strated the efficacy of quantitative antibiograms.5a6 The methodology differed from that of the present study in which an increased number of drugs was used and PFGE was used as a reference. The increased costs incurred by the use of multiple drugs was compensated by reducing labor and sharing the cost of running PFGE. Typing by PFGE has been acknowledged as more discriminatory than by ribotyping.’

The foremost advantages of the cluster analysis of susceptibility include its ease, decreased cost and labor, speediness, and no need of expensive equipment. Low specificity is a limitation of the cluster analysis; however, the 58% noted in this study may have been because the materials contained strains domestic to the particular hos- pital alone. Strains harvested from the outpatient units were included for possible foreign strains, but ironically,

PFGE showed that some clones were endemic across the outpatient and ward areas.

Nevertheless, for community hospitals in areas where PFGE is hardly available, the advantages outweigh the lim- itations in the analysis of nosocomial spread of MRSA. The cluster analysis of susceptibility, occasionally requires confirmation by PFGE, because these two methods are mutually complementary For this purpose, an attempt to compare cluster analysis on the band patterns of PFGE and that on an antibiogram may expedite the procedure.

REFERENCES

1. Ichiyama S, Ohta M, Shimokata K, Kato N, Takeuchi J. Genomic DNA fingerprinting by pulsed-field gel elec- trophoresis as an epidemiological marker for study of noso- comial infections caused by methicillin-resistant Staphylococcus aureus. J Clin Microbial 1991; 29:2690-2695.

2. Hu L-A, Umeda A, Kondo S, Amako K. Typing of Stapbylo- coccus aureus colonising human nasal carriers by pulsed- field gel electrophoresis. J Med Microbial 1995; 42:127-132.

3.Yoshida J, Nagata T, Nishioka Y Nose Y, Tanaka M. Outbreak of multidrug-resistant Staphylococcus aureus: a cluster analy- sis. J Clin Epidemiol 1996; 49:1447-1452.

4. Kioos WE. Staphylococcus. In: Collier L, Balows A, Sussman M, eds. Topley and Wilson’s microbiology and microbial infections. 9th ed. Vol. 2. Systemic bacteriology. London: Arnold, 1998:577-632.

5. Giacca M, Memo S, Trojan S, Monti-Bmgadin C. Cluster analy- sis of antibiotic susceptibility patterns of clinical isolates as a tool in nosocomial infection surveillance. Eur J Epidemiol 1987; 3:155-163.

6. Blanc DS, Petignat C, Moreillon P Wenger A, Bille J, Franci- oli l? Quantitative antibiogram as a typing method for the prospective epidemiological surveillance and control of MRSA: comparison with molecular typing. Infect Control Hosp Epidemiol 1996; 17:654-659.

7. Prevost G, Jaulhac B, Piemont Y. DNA fingerprinting by pulsed-field gel electrophoresis is more effective than ribo- typing in distinguishing among methicillin-resistant Staphy- lococcus aureus isolates. J Clin Microbial 1992; 30:967-973