Molecular diversity of Proteus mirabilis isolates producing extended-spectrum β-lactamases in a French university hospital


Corresponding author and reprint requests: M. Biendo, Laboratoire de Bactériologie et Hygiène, CHU Nord, Place V, Pauchet, 80054 Amiens cedex 1, France


Between February 1997 and December 2002, 3340 hospitalised patients yielded samples positive for Proteus mirabilis, of whom 45 (1.3%) were colonised/infected by P. mirabilis producing extended-spectrum β-lactamases (ESBLs). The gross incidence of patients colonised/infected by ESBL-producing P. mirabilis was 1.61/105 days of hospitalisation, with 20% of isolates being collected from patients in urology wards, most frequently (53.3%) from urine samples. Seventeen (37.7%) of the 43 isolates were obtained from samples collected within 48 h of hospitalisation, indicating that they were community-acquired. Isoelectric focusing assays and sequencing identified the TEM-24, TEM-92 and TEM-52 ESBLs. Pulsed-field gel electrophoresis revealed eight pulsotypes (I–VIII), with the two most common pulsotypes, IV and VI, comprising ten (23.3%) and 12 (26.6%) isolates, respectively. These pulsotypes were considered to represent epidemic strains and spread in various wards of the hospital.


Most infections caused by organisms producing extended-spectrum β-lactamases (ESBLs) have been described as nosocomial [1–4], although such infections have also been described in non-hospitalised patients [5–7]. Some data suggest that infections caused by ESBL-producing organisms might be an emerging problem in outpatients in different countries [5–7]. Among Enterobacteriaceae, Proteus mirabilis is the second most common cause of urinary tract infections (UTIs), and is also an important cause of nosocomial infections [8]. Wild-type strains of P. mirabilis are usually susceptible to β-lactams. However, a progressive increase in β-lactam resistance, mediated by the production of acquired β-lactamases, has occurred in this species [9]. Plasmid-mediated ESBLs, including TEM-type derivatives active against expanded-spectrum cephalosporins, have also started spreading in P. mirabilis[10]. In studies performed in French hospitals, the proportion of ESBL-producing isolates rose from 0.8% of P. mirabilis isolates in 1991 [11] to 2.4% in 1996 [12], 3.7% and 6.9% in 1998 [13,14], and 14.2% in 2000 [8]. Within the genus, P. mirabilis is by far the commonest species in which ESBLs have been recognised, with occasional outbreaks of infections caused by an epidemic resistant strain and the spread of self-transferable plasmids from one organism to another [14]. The present study used the technique of pulsed-field gel electrophoresis (PFGE) to assess the relatedness of P. mirabilis isolates thought to be responsible for outbreaks of nosocomial infections occurring in various wards in Amiens University Hospital, France. Analysis of the PFGE results was combined with data on antimicrobial resistance patterns and β-lactamase types.

Materials and methods

Study population and bacteria

This was a retrospective 6-year study (February 1997 to December 2002) of 122 978 patients admitted to Amiens University Hospital, France. Bacteria isolated from these patients were identified by the ID32 system (bioMérieux, Marcy l'Etoile, France) and tested for ESBL production by the double-disk synergy test (see below). Control strains were Escherichia coli K12R111 producing TEM-1, P. mirabilis CF529 producing TEM-92, and Enterobacter aerogenes producing TEM-24 and SHV-4. Plasmids pBR322 and pUC8 (TEM-1-encoding plasmids of 4.362 kb and 2.678 kb, respectively), EA 600 (a TEM-24-encoding plasmid of 85 kb) and CF529 (a TEM-92-encoding plasmid of 50 kb) were used for comparisons.


Catheter colonisation or infection was recorded when P. mirabilis was cultured from an intravascular catheter (≥ 103 CFU/mL) according to the method of Brun-Buisson [15] in the absence or presence, respectively, of clinical symptoms. Symptomatic patients with positive urine cultures (with fewer than three different pathogens isolated at ≥ 105 CFU/mL, or ≥ 103 CFU/mL associated with a leukocyturia of > 104/mL) were considered to have UTI. Asymptomatic patients with a permanent urinary catheter and positive cultures (with fewer than three different organisms isolated at ≥ 105 CFU/mL, or without a bladder catheter and two consecutive positive urine cultures of the same microorganism at ≥ 105 CFU/mL) were considered to have asymptomatic bacteriuria [16]. When the interval from admittance to isolation of the organism was ≤ 48 h, with no previous hospitalisation, the infection was considered to be community-acquired [17].

Antimicrobial susceptibility testing

Susceptibility to antimicrobial agents was tested by the disk diffusion method on Mueller–Hinton agar (Bio-Rad, Ivry sur Seine, France) according to the recommendations of the Comité de l'Antibiogramme de la Société Française de Microbiologie (CA-SFM) [18]. The antimicrobial agents used in this study were ampicillin, ticarcillin, piperacillin, amoxycillin–clavulanic acid, cephalothin, cefoxitin, latamoxef, cefotaxime, ceftazidime, cefepime, cefpirome, imipenem, aztreonam, gentamicin, tobramycin, netilmicin, amikacin, isepamicin, trimethoprim–sulphamethoxazole, pipemidic acid and ofloxacin. Bacteria were classified as susceptible, intermediate or resistant as described by the CA-SFM [18]. MICs of cefotaxime, ceftazidime, aztreonam, cefepime and cefpirome were determined alone or in combination with a fixed concentration of clavulanate (2 mg/L) by a serial two-fold macrodilution procedure in Mueller–Hinton broth [19]. Double-disk synergy tests were performed on Mueller–Hinton agar as described previously [20], with a central amoxycillin–clavulanic acid disk and disks of third-generation cephalosporins (ceftazidime, cefepime and cefpirome) and a monobactam (aztreonam) placed 30 mm (centre to centre) from each other. The test was considered to be positive for ESBL production when the bacterial growth had a ‘champagne cork’ appearance [20].

β-Lactamase assay

Isoelectric focusing of crude β-lactamase extracts was performed with polyacrylamide gels containing ampholines of pH 3.5–9.5 (ampholine PAG plate; Amersham Biosciences, Little Chalfont, UK), using the procedure recommended by the manufacturer, and an LKB 2117 Multiphor II apparatus (Amersham Biosciences). The isoelectric points (pIs) of the β-lactamases studied were determined by comparison with reference enzymes (TEM-1, pI 5.4; TEM-92, pI 6; TEM-24, pI 6.5; and SHV-4, pI 7.8).

PFGE typing

Isolates were typed by determining PFGE SfiI DNA macrorestriction patterns with the Genepath Group 5 Reagent kit (Bio-Rad) according to the manufacturer's recommendations. PFGE was performed with a contour-clamped homogeneous electric field system (CHEF-DR II, Bio-Rad) at 14°C and 200 V for 19.7 h. Molecular Analyst software (Bio-Rad) was used to analyse the DNA restriction patterns and determine their similarity, based on calculation of the Dice similarity coefficient and use of the UPGMA algorithm (unweighted pair-group method using arithmetic averages).

Agarose gel electrophoresis of plasmids and PCR amplifications

Plasmid DNA was extracted with the RPM kit (QBiogene, Calsbad, CA, USA), used according to the manufacturer's instructions. The crude extract was used directly for electrophoresis, which was performed in agarose 0.7% w/v gels for 2 h at 100 V. Plasmid sizes were determined by comparison with the plasmid size standards. PCR amplifications and electrophoresis were performed as described previously [21,22], but with primers TEM-A (5′-ATGAGTATTCAACATTTCCGTG) and TEM-B (5′-TTACCAATGCTTAATCAGTGAG). Initial denaturation was at 95°C for 3 min, followed by 30 cycles of 95°C for 1 min, 55°C for 1 min and 72°C for 1 min, and a final extension at 72°C for 1 min.

DNA sequencing

The PCR-amplified blaTEM genes were purified using the QIA-quick PCR purification kit (QIAgen, Courtaboeuf, France) according to the manufacturer's recommendations, and were then sequenced by PCR cycle sequencing with Dye Ex chemistry (QIAgen), followed by automated analysis on an ABI PRISM 310 Genetic analyser (Applied Biosystems, Foster City, CA, USA).


Incidence of ESBL-producing P. mirabilis infection

During the study period, 122 978 patients were hospitalised, of whom 3340 (2.7%) yielded samples positive for P. mirabilis. Of these, 45 (1.3%) patients were colonised/infected by ESBL-producing isolates. These 45 patients were admitted to various wards of the hospital and comprised 27 (60%) males and 18 (40%) females, with a mean age of 63 years (range 23–94 years). The gross incidence of patients infected by an ESBL-producing P. mirabilis isolate was 1.61/100 000 days of hospitalisation. The urinary tract was the most frequent site of isolation (53.4%), followed by the bronchopulmonary tract (17.7%) and miscellaneous suppuration sites (17.7%). Urological, miscellaneous medical, surgical, long-stay and middle-stay wards accounted for 62.2% of the cases of colonisation/infection (Table 1). Patient records revealed that P. mirabilis was associated with other ESBL-producing microorganisms in 20% (9/45) of infections: in UTIs, with E. coli (two cases), Klebsiella pneumoniae (one case) and Providencia stuartii plus E. coli (one case); in bronchopulmonary tract infections, with K. pneumoniae (one case) and Ent. aerogenes (two cases); in femoral catheter infections, with Ent. aerogenes (one case); and in suppuration, with K. pneumoniae (one case). The following risk factors were present in 24.2% (12/45) of patients: indwelling urinary catheter (50%; 6/12); assisted ventilation (33.3%; 4/12); and intravascular catheter (16.6%; 2/12). Seven patients died in hospital (three patients hospitalised in the urology ward, one in the nephrological intensive care unit (ICU), one in a polyvalent ICU, one in the neurology ward and one in the dermatology ward). Death was not related directly to P. mirabilis infection, but was probably caused by underlying diseases.

Table 1.  Data for the first ESBL-producing isolate of Proteus mirabilis from each patient
Isolate no.Date of isolation (month/year)SpecimenPatient genderHospital wardNo. of days of hospitalisation before isolatea
  • ICU, intensive care unit; M, male; F, female.

  • a

    The median time to acquisition of ESBL in long-stay and rehabilitation wards was 32 days vs. 16.2 days in short-stay wards.

  • b

    These patients were hospitalised 2 months previously and the isolate was considered to be hospital-acquired.

 203/1997UrineMMaxillo-facial surgery23
 507/1997UrineMFunctional rehabilitation72
 610/1997Jugular catheterMLong-stay ward40
 710/1997UrineMVascular surgery25
 812/1997SuppurationMLong-stay ward72
 901/1998UrineMPolyvalent ICU93
1006/1998UrineFMiddle-stay ward25
1302/1999Bronchial fluidFNeurology13
1404/1999Bronchial fluidFRespiratory medicine1
1504/1999UrineMFunctional rehabilitation1
1605/1999UrineFMiddle-stay ward26
1810/1999UrineMPlastic surgery11
1901/2000UrineMFunctional rehabilitation1
2609/2000Bronchial fluidMPolyvalent ICU1
2912/2000Femoral catheterFNephrological ICU315
3103/2001UrineMInternal medicine1
3204/2001UrineFNephrological ICU1
3305/2001UrineMLong-stay ward34
3606/2001Bronchial fluidMRespiratory ICU1
3806/2002WoundMPlastic surgery18
4107/2002Renal abscessFNephrology4
4210/2002SputumMLong-stay ward1
4310/2002Bronchial fluidMRespiratory medicine  1b
4411/2002UrineFNeurology  1b

Seventeen (37.7%) ESBL-producing P. mirabilis isolates were identified in samples collected within 48 h of hospitalisation. These patients were admitted from their respective homes, had not been hospitalised during the previous 6 months or transferred from another hospital, and were considered to be community-acquired cases of infection. Twenty-eight (62.3%) cases of P. mirabilis infection were identified ≥ 72 h after admission and were considered to be hospital-acquired.

Antimicrobial susceptibilities

The 45 P. mirabilis isolates gave a positive disk potentiation when cefotaxime, ceftazidime, cefepime, cefpirome and aztreonam disks were used with amoxycillin–clavulanic acid. These isolates showed decreased susceptibilities to amino-, carboxy- and ureido-penicillins, most cephalosporins (including cefotaxime, ceftazidime, cefepime, cefpirome) and aztreonam, but remained susceptible to cephamycins (cefoxitin, latamoxef) and imipenem. β-Lactam MICs were increased for the 45 clinical isolates (Table 2) compared to the reference strain E. coli ATCC 25922, which had MICs ≤ 1 mg/L. Variable susceptibility patterns were observed for cefotaxime (MIC range 2–128 mg/L), ceftazidime (4–64 mg/L), cefepime (2–64 mg/L), cefpirome (2–32 mg/L) and aztreonam (2–32 mg/L). For all isolates, the MICs of cefotaxime, ceftazidime, cefepime, cefpirome and aztreonam decreased to ≤ 0.25 mg/L in the presence of clavulanic acid 2 mg/L. MICs of cefoxitin, latamoxef and imipenem for the same isolates were 0.25–1, 0.06–0.25 and 0.25–0.5 mg/L, respectively. With regard to aminoglycosides, 62.2% of the 45 isolates were resistant to gentamicin, tobramycin and netilmicin, 20% were resistant to tobramycin, netilmicin and amikacin, and 17.8% were resistant to gentamicin, tobramycin, netilmicin and amikacin, suggesting the presence of AAC(3), AAC(6′), and AAC(3) + AAC(6′), respectively. All 45 isolates were also resistant to trimethoprim–sulphamethoxazole, pipemidic acid and ofloxacin.

Table 2.  Results of analytical isoelectric focusing and molecular characterisation of 45 isolates of Proteus mirabilis producing extended-spectrum β-lactamase
Isolate no.β-Lactamase characterisation MIC (mg/L)PFGE patterns
  1. CTX, cefotaxime; CAZ, ceftazidime; FEP, cefepime; CPO, cefpirome; ATM, aztreonam; PFGE, pulsed-field gel electrophoresis.

126.0 + 5.9TEM-92 + TEM-5232648816I

Identification of β-lactamases

The presence of ESBLs was suggested by the results of isoelectric focusing and synergy tests. All 45 isolates yielded amplicons in PCRs for blaTEM, while none were positive in PCRs for blaSHV.

Sequencing of blaTEM genes was performed on three isolates with β-lactamase pIs of 5.9, 6.0 and 6.5. Analysis of the deduced protein sequence, compared to that of blaTEM-1, showed three amino-acid substitutions for the pI 5.9 enzyme (Glu-104 → Lys; Met-182 → Thr; Gly-238 → Ser) and four amino-acid substitutions for the pI 6.0 enzyme (Gln-6 → Lys; Glu-104 → Lys; Met-182 → Thr; Gly-238 → Ser). These protein sequences are identical to those of TEM-52 and TEM-92, respectively [21,23]. The enzyme of pI 6.5 had five substitutions reported previously [10] (Gln-39 → Lys; Glu-104 → Lys; Arg-164 → Ser; Ala-237 → Thr; Glu-240 → Lys), corresponding to TEM-24. Overall, the results suggested that 24 (55.6%) isolates produced TEM-24, 15 (33.3%) produced TEM-92, four (8.8%) produced TEM-52, and one (2.2%) (no. 12) produced both TEM-92 and TEM-52. Plasmid visualisation in agarose gels showed that the TEM-24 gene was located on an 85-kb plasmid, the TEM-92 gene on a 50-kb plasmid, and the TEM-52 gene on a 70-kb plasmid. Isolate no. 12 contained both the 50-kb and the 70-kb plasmids.

PFGE analysis and epidemiology of ESBL-producing P. mirabilis isolates

The 45 SfiI PFGE profiles grouped [24] into eight pulsotypes (I–VIII; Table 2). The predominant types were pulsotypes VI (12 isolates; 26.6%) and IV (ten isolates; 22.3%). The 12 isolates belonging to pulsotype VI were obtained between 1997 and 2001 from patients in different care units. Similarly, the ten isolates belonging to pulsotype IV were collected between 2001 and 2002 from patients hospitalised in various wards: three in urology, two in geriatrics, and the remaining five in different care units. Seven isolates belonged to pulsotype I, of which three were collected from patients hospitalised in the same care unit (functional rehabilitation), two from patients in the urology ward, and two others from patients in different care units. All of these isolates were recovered between 1977 and 2000. Five isolates were collected in 2000 and belonged to pulsotype II; these were obtained from patients hospitalised in different wards, with the exception of two patients who were in the same geriatric ward. Three isolates (two from urology and one from a geriatric ward) belonging to pulsotype III were isolated in 2000, and one additional isolate was obtained in 1999 from the urology ward. The two isolates of pulsotype V and the four isolates of pulsotype VII were collected between 1997 and 2001. The single isolate of pulsotype VIII was obtained in 2002. These results are shown in more detail in Tables 1 and 2.

The first isolate of ESBL-producing P. mirabilis was obtained in this hospital in February 1997 from a patient admitted to the urology ward (patient 1). During the following months (March to December 1997), multiresistant isolates of P. mirabilis were obtained from another patient hospitalised in the urology ward (patient 3) and from patients in maxillofacial surgery (patient 2), neurosurgery (patient 4), functional rehabilitation (patient 5), vascular surgery (patient 7), and a long-stay ward (patients 6 and 8) (Tables 1 and 2).


The overall prevalence rate (1.3%) of ESBL-producing P. mirabilis observed in this study was lower than rates reported elsewhere, and the present study did not confirm the suggestion that ICUs play a major role in the intra-hospital dissemination of epidemic multiresistant bacteria [25]. In the present study, the urology, long-stay and rehabilitation wards, which together accounted for 35.5% of all patients colonised/infected with ESBL-producing P. mirabilis, played a predominant role in intra- and inter-ward dissemination. Many of these isolates were obtained from elderly patients (mean age 63 years) hospitalised in the urology (20%) or long-stay and rehabilitation wards (15.5%), and most (53.3%) were collected from urine, with a higher rate for men (62.5%) than for women (37.5%). This higher rate could be explained by the frequency of UTIs observed in these units and by the presence of P. mirabilis infections in elderly patients with severe underlying disease or chronic illness. These findings confirm those reported previously [13].

In the present study, the high percentage of community-acquired cases (37.7% of ESBL colonised/infected patients) differed from previously published figures of 8.7%[26], 1.5%[27] and 0.3%[28]. These variations might be explained by the recruitment of different patient types and variable epidemiological situations in different hospitals or regions. The proportion of community-acquired infections observed in the present study may be an overestimation, caused by factors such as hospitalisation at home, in private healthcare centres, clinics, nursing homes, a stay in follow-up care wards, empirical antibiotic therapy or ‘contact’ with another carrier (e.g., in the same family). Colodner et al.[6] conducted a study in Afula, Israel, and found that the independent risk factors for ESBL production in isolates from non-hospitalised patients were previous hospitalisation in the past 3 months, antibiotic treatment in the past 3 months, age > 60 years, diabetes and male gender. Most patients hospitalised in ICUs are transferred to a general acute-care unit, followed by a rest, nursing or retirement home. Some of these patients may carry ESBL-producing Enterobacteriaceae for prolonged periods, and continued carriage of such strains may contribute to their extra-hospital spread [29]. In addition, community-acquired strains producing ESBLs might be selected from the existing gastrointestinal flora when it is exposed to broad-spectrum antimicrobial agents [30].

The levels of resistance to cefotaxime, ceftazidime, cefepime, cefpirome and aztreonam varied according to the isolate and enzyme type involved, but all ESBL-producing isolates were detected by the double-disk synergy test. The relative MICs of cefotaxime and ceftazidime were in accord with either the presence of an ESBL with high hydrolytic activity against cefotaxime (TEM-92 or TEM-52) or an ESBL with high hydrolytic activity against ceftazidime (TEM-24) [10,14,23,31]. TEM-92 and TEM-52 were encoded by large plasmids that differed in size, but which were associated with similar β-lactam resistance markers. In contrast, the TEM-24 plasmid was associated with different β-lactam resistance markers. The results agreed with those reported previously concerning TEM-92, TEM-52 and TEM-24, which are encoded by 50-, 70- and 85-kb plasmids, respectively [10,23], although Pai et al.[31] observed a great variability in plasmid size (between 71 and 100 kb) among TEM-52-producing E. coli and K. pneumoniae isolates from Korea.

TEM-24 is produced predominantly by strains of Ent. aerogenes[32], and the first Ent. aerogenes strains producing TEM-24 were reported in our hospital in 1996, where they were responsible for an epidemic that occurred between 1998 and 1999 [33]. The first isolates of P. mirabilis producing TEM-24 were observed in our hospital in February and March 1997. A TEM-92-producing P. mirabilis strain was first isolated in a French hospital in 1998 [14], and it was during the same year that a P. mirabilis isolate producing TEM-92 was identified in our laboratory. TEM-52 has been reported previously in K. pneumoniae[23,31], and subsequently in P. mirabilis isolates from Italy [14]. The first P. mirabilis isolate producing TEM-52 in our hospital was isolated in January 1999.

The presence of these plasmid-mediated β-lactamases in a hospital is a cause for concern. According to the literature, there is an increasing incidence of this kind of resistance in many French hospitals. Polyclonal spread of P. mirabilis isolates was observed in most wards of the hospital. The easy transfer of the genes conferring resistance indicates a risk of transmission to other enterobacteria that could lead to treatment failure. The diversity of β-lactamases, and their association with resistance to aminoglycosides, quinolones and trimethoprim–sulphamethoxazole, suggests that resistant strains of P. mirabilis could become a problem in the context of nosocomial infection. The frequency of infection in the community means that resistant strains can be constantly re-introduced into a hospital. The present investigation also demonstrated the existence of several clusters of P. mirabilis isolates. Certain strains were disseminated widely in the hospital, and were maintained during the study period, while others showed limited spread over time.


We thank W. Sougakoff (Laboratoire de Bacteriologie–Hygiène, Faculté de Médecine, Pitié-Salpétrière, Paris) for the reference strain of P. mirabilis CF 529, and D. Sirot (Department of Bacteriology and Virology, Medical Faculty of Auvergne University) for the E. coli strains encoding the reference β-lactamase TEM-1 (K12R111). We also thank GlaxoSmithKline Pharmaceuticals for lithium clavulanate powder.