Large-scale nosocomial outbreaks of Serratia marcescens septicaemia in Japan have had a fatality rate of 20–60% within 48 h. As a countermeasure, a real-time PCR assay was constructed for the rapid diagnosis of S. marcescens septicaemia. This assay indeed detected S. marcescens in clinical blood specimens (at ca. 102 CFU ml−1), at a frequency of 0.5% in suspected cases of septicaemia. In mice, the assay provided estimates of blood S. marcescens levels at various infectious stages: namely, 107 to 108 CFU ml−1 at a fatal stage (resulting in 100% death), 104–105 CFU ml−1 at a moderately fatal stage (resulting in 50% or more death), and <103 CFU ml−1 at a mild stage (resulting in 100% survival), consistent with actual CFU measurements. Blood bacterial levels could be an important clinical marker that reflects the severity of septicaemia. The simultaneous detection of S. marcescens and the carbapenem resistance gene was also demonstrated.
Serratia marcescens is an opportunistic bacterium that causes nosocomial infections in intensive care units (ICUs) and neonatal intensive care units (NICUs) [1–3]. S. marcescens causes urinary tract infections , respiratory tract infections , conjunctivitis , endocarditis , meningitis , wound infections [5,9], or septicaemia the mortality of which reaches 39–50%[10,11].
In Japan, three large-scale outbreaks of S. marcescens septicaemia occurred between 1999 and 2002 [12–14]. A peripheral venous catheter inserted over a long period was a common factor in all three outbreaks. The health of numerous patients deteriorated rapidly and many subjects died before the blood culture results returned. The mortality within 48 h after a sudden increase in fever was 20% in the first outbreak, in Tokyo, in 1999; 60% in the second, in Osaka, in 2000; and 25% in the third, in Tokyo, in 2002.
This, together with the fact that, in such outbreaks, no correlation was observed between blood lipopolysaccharide (LPS) levels and mortality , strongly suggested the need for a new rapid and sensitive method and its clinical application in the diagnostic laboratory. Although the real-time PCR assay has been employed for detecting microbial pathogens in clinical blood samples [15–19], no such assay of S. marcescens (in blood) has been reported. In this study, we developed a specific real-time PCR assay for the detection of S. marcescens in blood and indeed diagnosed septicaemia from the results. We also evaluated the assay in a mouse infection model, and demonstrated the simultaneous detection of S. marcescens and the carbapenem resistance gene in one test tube.
2Materials and methods
The S. marcescens strains, 89 and 1–2, from the nosocomial outbreaks in Tokyo in 1999  and 2002  were employed in this study. Other S. marcescens strains used included S. marcescens ATCC29021 and clinical isolates (54 strains) from blood, sputum, urine, or faeces from patients in Tokyo, Yokohama, Chiba, and Niigata. All of the strains (132 strains) of other gram-negative bacterial species (18 species) used in this study were also clinical isolates from blood, sputum, urine, or faeces from patients in Tokyo, Chiba, Niigata, Bangkok (Tailand), and Dhaka (Bangladesh). They include: 13 strains of S. liquefaciens; 3 strains of S. plymuthica; 4 strains each of enteropathogenic Escherichia coli, enterotoxigenic E. coli, enteroinvasive E. coli, enterohemorrhagic E. coli, and enteroaggregative E. coli; 20 strains of Klebsiella pneumoniae; 15 strains of K. oxytoca; 3 strains of Enterobacter aerogenes; 5 strains each of Salmonella Enteritidis, S. Typhi, and S. Typhimurium; 3 strains of Citrobacter freundii; 3 strains of Proteus mirabilis; 3 strains each of Shigella dysenteriae and S. flexneri; 2 strains each of S. boydii and S. sonnei; 5 strains of Yersinia enterocolitica; 15 strains of Pseudomonas aeruginosa; 3 strains of Campylobacter jejuni/coli; 2 strains of Alcaligenes xylosoxidans; and 2 strains of Acinetobacter barumannii.
Carbapenem-resistant S. marcescens strains (4 strains) used in this experiment were isolated from clinical blood or urine specimens, and included strain NSM1 carrying the blaIMP-1 gene (GenBank Accession No. AB162950) isolated in Niigata, and strains TO1, TO2, and TO3 carrying the blaIMP-1 gene (GenBank Accession Nos. AB162947, AB162948, and AB162949) isolated in Tokyo. Carbapenem-resistant P. aeruginosa strain NPA1 carrying the blaIMP-10 gene (GenBank Accession No. AB195637) from a clinical blood specimen in Niigata, and A. xylosoxidans strain NAX1 carrying the blaIMP-10 gene (AB195638) from a urine specimen in Niigata were also employed.
2.2Media and bacterial growth
For bacterial growth, we used LB broth (Difco Laboratories, Detroit, MI) as the liquid medium, which was inoculated and incubated at 37°C to log phase with agitation. Nutrient agar (Eiken Chemical, Tokyo) was used as the solid medium.
Serogroups of the S. marcescens outbreak that occurred in Tokyo in 1999 and 2002 were analyzed using a Serratia O antisera kit (Denka Seiken, Tokyo) with polyclonal antibodies, following the manufacturer's directions.
2.4Pulsed-field gel electrophoresis
Total bacterial DNA was extracted and digested with Xba I or Spe I . The digested DNA was subjected to pulsed-field gel electrophoresis (PFGE) (1% agarose) with a lambda ladder as a standard molecular size marker (Bio-Rad, Hercules, CA, USA) and stained with ethidium bromide.
2.5Blood from volunteers and clinical blood samples
The human blood used for bacterial suspensions was from six healthy volunteers. Clinical blood samples were obtained from 370 patients with suspected septicaemia (who developed a fever). Each blood sample (10 ml) was put into blood culture bottles to examine aerobic and anaerobic microbial infections (including S. marcescens) in routine practice. Another 1 ml of each sample was kept in a refrigerator: 200 μl was subjected to a real-time PCR assay and the remainder was used for determining the number of CFU on nutrient agar.
2.6Preparation of bacterial DNA
In experiments where the specificity of the real-time PCR assay was examined, bacterial cells, grown on nutrient agar, were suspended in phosphate-buffered saline (PBS, pH 7.4) to 1 × 107 CFU ml−1. Bacterial DNA was extracted from 200 μl of the suspension, using the QIAamp DNA mini kit (Qiagen, Tokyo) following the manufacturer's instructions. The DNA was finally dissolved in 200 μl of Tris–hydrochloride, pH 8.0. In experiments where the sensitivity was examined, S. marcescens cells, freshly cultured on nutrient agar, were serially diluted in PBS or in human blood (1 to 1 × 107 CFU ml−1). S. marcescens DNA was then extracted from 200 μl of each dilution and finally dissolved in 200 μl of Tris–hydrochloride, pH 8.0. The extraction of bacterial DNA from clinical blood samples (200 μl each) and blood samples from infected mice (200 μl each) was performed in the same way. Twenty-microliter aliquots of the DNA solution were used for the real-time PCR assay.
2.7PCR primers and probe
For the detection of S. marcescens by a real-time PCR assay, the primers (SMSF and SMSR) and probe (SMSP) were designed based on the 16S rRNA gene sequence of S. marcescens strain DSM 30121 (GenBank Accession No. AJ233431): SMSF, 5′-GGTGAGCTTAATACGTTCATCAATTG (nucleotide position: 435–460); SMSR, 5′-GCAGTTCCCAGGTTGAGCC (595–613); SMSP, 5′-TGCGCTTTACGCCCAGTAATTCCGA (534–558). The primers SMSF and SMSR amplified a fragment 179 bp in size. The probe SMSP contained the reporter dye FAM (6-carboxyfluorescein; Applied Biosystems, Foster City, CA) at the 5′ end and the quencher dye TAMRA (6-carboxytetramethylrhodamine; Applied Biosystems) at the 3′ end.
To detect the carbapenem resistance gene (blaIMP-1), the primers (blaF and blaR) and probe (IMP-R) were designed based on the blaIMP-1 gene sequence of the carbapenem-resistant S. marcescens strain NSM1: blaF, 5′-GTTTGTGGAGCGTGGCTATAAAA (nucleotide position: 238–260); blaR, 5′-GATCGAGAATTAAGCCACTCTATTCC (305–330); IMP-R, 5′-AGGCAGCATTTCCTCTCATTTTCATAGCG (265–293). The primers blaF and blaR amplified a fragment 93 bp in size. The probe IMP-R contained the reporter dye VIC™ (Applied Biosystems) at the 5′ end and the quencher dye TAMRA at the 3′ end.
2.8Real-time PCR assay
The real-time PCR was performed with a ABI 7900HT sequence detector (Applied Biosystems), following the manufacturer's instructions. The reaction mixture (50 μl) contained 400 nM of the primers, 80 nM of the probe, 25 μl of Taqman universal prepared mixture and 20 μl of the template (bacterial DNA prepared as above). The cycling conditions were an initial single cycle for 10 min at 95°C (to activate AmpliTaq Gold) and 50 cycles of two-temperature cycling consisting of 15 s at 95°C (for denaturation) and 1 min at 60°C (for annealing and polymerization). The intensity of fluorescence (ΔRn) was calculated by subtracting the baseline fluorescence from the actual fluorescence signal data. The threshold cycle (CT) was defined as the cycle number at which the reporter fluorescence exceeded the threshold value, a parameter defined as 10 SD above the baseline fluorescence. In experiments where both S. marcescens and the carbapenem resistance gene were simultaneously detected in one test tube, the reaction mixture (50 μl) contained the primer and probe sets both for S. marcescens (SMSF, SMSR, and FAM-labeled SMSP) and for the blaIMP-1 gene (blaF, blaR, and VIC™-labeled IMP-R). The real-time PCR assay using SYBR Green I dye (Applied Biosystems) was performed, following the manufacturer's directions.
2.9Serratia marcescens infection in mice
The mice used were specific pathogen-free male C57BL/6 mice (CLEA Japan Inc., Tokyo). S. marcescens (outbreak-derived strains 89 and 1–2) was grown on a nutrient agar plate at 37°C, and colonies were suspended in PBS. The bacterial suspension (100 μl) at a concentration of 1 × 107, 3 × 107, 6 × 107, 1 × 108, 1.7 × 108, 3 × 108, 1 × 109, or 3 × 109 CFU ml−1 was used for intraperitoneal (i.p.) inoculation. Each infection group contained 12 mice each. To determine the LD50, the survival of the infected mice was monitored every 3 h after the S. marcescens inoculation for up to 60 h, and then every 6 h for up to 2 weeks. In experiments where the S. marcescens cell concentrations in blood were determined, whole-blood samples (400 μl each) were obtained by cardiac tapping at 0 min, 45 min, 1.5 h, 3 h, 6 h, 9 h, 12 h, and 24 h after the S. marcescens injection. The blood samples were subjected to an assay of the actual CFU count by agar plating as well as the real-time PCR assay.
3Results and discussion
3.1Characteristics of outbreak-derived S. marcescens
O serogroups of the two outbreak strains 89 and 1-2 were untypable (OUT). The highest titre in the serogroup assay was observed with serogroup O6 (1:16 for both strains 89 and 1-2). The titre with serogroup O14 was 1:8 for strain 89 and 1:4 for strain 1–2. The titre with the other serogroups (including O12) examined was <1:2. Pulsed-field gel electrophoresis with Xba I or Spe I revealed that the two strains differed markedly, indicating that they were of different origins (not clonally related).
3.2Specificity and sensitivity of the real-time PCR assay
To check the specificity of the primer-probe set (SMSF, SMSR, and FAM-labeled SMSP), a large battery of S. marcescens isolates (57 strains) of different sources and origins, including the outbreak-derived strains 89 and 1–2 and strain ATCC29021, and 132 strains of other gram-negative bacterial species (18 species) were examined. They were suspended at a concentration of 1 × 107 CFU ml−1, and examined by the real-time PCR assay. All of the S. marcescens strains examined produced positive results. In contrast, all of the strains of other gram-negative bacterial species, S. liquefaciens, S. plymuthica, diarrheagenic E. coli (enteropathogenic, enterotoxigenic, enteroinvasive, enterohemorrhagic, enteroaggregative), K. pneumoniae, K. oxytoca, E. aerogenes, S. enterica (Enteritidis, Typhi, Typhimurium), C. freundii, P. mirabilis, S. dysenteriae, S. flexneri, S. boydii, S. sonnei, Y. enterocolitica, P. aeruginosa, C. jejuni/coli, A. xylosoxidans, and A. barumannii, gave negative results. Human blood itself from 6 healthy volunteers (without bacteria) also gave negative results. The data indicated the real-time PCR assay developed in this study is highly specific.
To examine the sensitivity of the real-time PCR assay, S. marcescens (strain 89) cells were serially diluted in PBS or in human blood (1 to 1 × 107 CFU ml−1). Bacterial DNA was then extracted from each dilution and used for the real-time PCR assay (Fig. 1A). When the CT values were plotted against the log10 of the S. marcescens concentrations, a good linearity was observed over the range from 5 to 1 × 107 CFU ml−1 in PBS and 2 × 101 to 1 × 107 CFU ml−1 in blood samples (Fig. 1B). The detection limit was 5 CFU ml−1 for PBS suspensions and 2 × 101 CFU ml−1 for blood suspensions. Very similar results were also obtained with the outbreak strains 1–2. Thus, the real-time PCR assay is highly sensitive.
3.3Assay of clinical blood samples
We examined 370 clinical blood samples from patients suspected of having septicaemia (who developed a fever). Of these, samples from two patients (0.54%) produced positive results for S. marcescens in culture as well as in the real-time PCR assay (for the real-time PCR assay, DNA was extracted directly from the patient's blood, not from blood culture bottles). When the CT value of the two clinical blood samples was plotted on the experimental calibration line (constructed for S. marcescens suspended in human blood) shown in Fig. 1B, the bacterial concentration in blood was estimated to be 6.5 × 101 CFU ml−1 for patient (a) and 1.4 × 102 CFU ml−1 for patient (b). The actual number of S. marcescens in blood for patients (a) and (b) was determined, on nutrient agar, to be 1.1 × 101 and 2.1 × 101 CFU ml−1, respectively.
The reason why the exact S. marcescens CFU count was under-estimated (compared with the estimate obtained with the real-time PCR assay) is not known, however, since the blood samples were kept in a refrigerator before the analysis, bacterial viability might have decreased during storage.
3.4Estimation of blood S. marcescens levels in mice
When mice were injected intraperitoneally with S. marcescens (strain 89), as shown in Fig. 2A, the LD50 was determined to be 6 × 106 CFU. When the mice were injected with 3 × 107 CFU, S. marcescens levels in blood reached a plateau at 105–106 CFU ml−1, then increased rapidly to 107–108 CFU ml−1 (at 12 h post-injection), which was absolutely fatal (all mice died 12–15 h after the injection) (Fig. 2B, a). When an inoculum of 6 × 106 CFU per mouse was used, S. marcescens levels in blood reached a plateau at 105–106 CFU ml−1, then decreased slightly to 104 CFU ml−1 (at 12 h post-injection). After that, bacterial levels started to increase in some mice and decreased in others, resulting in ca. 50% death from 54 h post-injection, and in a large standard deviation (SD) in the data. (Fig. 2B, b). When a smaller inoculum (3 × 106 CFU per mouse) was used, S. marcescens levels in blood once again reached a plateau at 105 CFU ml−1 (as a spike), before rapidly dropping to 103 CFU ml−1 (at 12 h post-injection) (Fig. 2B, c). All the mice were alive even 2 weeks after the injection.
All the changes of bacterial CFU levels in blood were rapidly and accurately demonstrated by the real-time PCR assay, as shown in Fig. 2B (a–c). Very similar results were also obtained when another outbreak-derived strain (1–2) was employed (data not shown).
The bacterial cell numbers in blood could therefore be a marker reflecting the severity of septicaemia. That is, in mice, at 12 h post-injection (where all the infected mice were still alive), blood bacterial levels of >107, 104–105, and <103 CFU ml−1 were 100% fatal, ca. 50% fatal, and 0% fatal (100% survival), respectively.
3.5Simultaneous detection of S. marcescens and the carbapenem resistance gene
Next, we simultaneously detected S. marcescens and the carbapenem-resistance gene (blaIMP-1) in one test tube using the real-time PCR assay (Fig. 3). With the probe method (Fig. 3A), the primers SMSF and SMSR and probe SMSP (labeled with FAM) detected S. marcescens DNA (Fig. 3A, a and b) but not P. aeruginosa DNA (Fig. 3A, c and d). Meanwhile, the primers blaF and blaR and probe IMP-R (labeled with VIC) detected the carbapenem resistance gene of both S. marcescens (carrying the blaIMP-1 gene) (Fig. 3A, a) and P. aeruginosa (carrying the blaIMP-10 gene) (Fig. 3A, c). Carbapenem-susceptible strains of S. marcescens or P. aeruginosa gave negative results (Fig. 3A, b and d). The detection limit of the real-time PCR assay using blaF, blaR, and IMP-R (labeled with VIC) for the carbapenem-resistant S. marcescens (strain NSM1) was 3 × 101 CFU ml−1 in PBS.
We also employed the SYBR Green method (Fig. 3B). The primers SMSF and SMSR detected S. marcescens (Fig. 3B, a and b) but not P. aeruginosa (Fig. 3B, c and d). The melting point (Tm) obtained with 6 strains of S. marcescens was 83.4 ± 1.1°C. The primers blaF and blaR detected the carbapenem resistance gene of both S. marcescens (blaIMP-1Fig. 3B, a) and P. aeruginosa (blaIMP-10Fig. 3B, c). The Tm value obtained with the 4 blaIMP-1 genes (of S. marcescens) was 76.9 ± 0.4°C. Carbapenem-susceptible S. marcescens (Fig. 3B, b) and P. aeruginosa (Fig. 3B, c) gave negative results. Carbapenem-resistant A. xylosoxidans (carrying the blaIMP-10 gene) was also analyzed using the real-time PCR assay; again, blaF and blaR gave a peak, but SMSF and SMSR did not (data not shown). The Tm value for the two blaIMP-10 genes (of P. aeruginosa and A. xylosoxidans) was 76.8°C.
The blaIMP-1 and blaIMP-10 sequences differed by only one base (G for blaIMP-1 and T for blaIMP-10 at the 145th nucleotide) and one amino acid (Val for blaIMP-1 and Phe for blaIMP-10 at the 49th amino acid). The sequences were the same for the primers blaF and blaR and probe IMP-R.
Thus, this assay rapidly detected carbapenem-resistant S. marcescens (carrying the blaIMP-1 gene). It also detected the carbapenem-resistance gene (the blaIMP-10 gene) carried by P. aeruginosa and A. xylosoxidans. Carbapenem has been recommended for the treatment of S. marcescens infections in Japan. Although the strains derived from the three large nosocomial outbreaks in Japan were susceptible to carbapenems [12–14,21], carbapenem-resistant S. marcescens were isolated from clinical specimens, at a rate of 5% or less . The resistance to carbapenem in S. marcescens is mostly encoded by the IMP-1-type metallo-β-lactamase gene (blaIMP-1) [22,23]. The blaIMP-1 gene is also distributed in highly resistant P. aeruginosa[22,23].
In conclusion, a real-time PCR assay was established for the rapid and specific detection of S. marcescens septicaemia. The time required to complete the assay was only 2.5 h, and only 200 μl of blood was needed. The assay accurately estimated S. marcescens cell numbers in blood, which could be a clinical marker reflecting the severity of septicaemia (as demonstrated in mice), although further studies on potential clinical applications are necessary. Regarding the incidence of S. marcescens infection, in a hospital in Niigata in 2003, blood culture examinations were positive in 170 (9.2%) of 1846 clinical blood specimens from patients who developed a fever. Gram-negative rods were detected in 22.9% of the positive specimens, and S. marcescens in 7.7%. Thus, the overall incidence of S. marcescens in blood specimens was 0.16%. The incidence obtained in the present study was slightly higher, probably due to the relatively small number of patients examined or due to the limited number of wards (mainly surgical wards) examined. The assay also allowed for the rapid and simultaneous detection of S. marcescens and the common carbapenem resistance gene (blaIMP-1). Thus, this study provides a countermeasure, in terms of a rapid diagnosis, to nosocomial outbreaks of S. marcescens septicaemia (including those due to carbapenem-resistant clones).
This work was supported by a Grant-in-Aid for Scientific Research from the Ministry of Education, Culture, Sports, Science and Technology, Japan.