Introduction

Bacillus cereus is ubiquitous in the environment. Bacillus spores are widespread in soil, dust, water and the hospital environment [1]. B. cereus contamination of laundered linen was associated with nosocomial outbreaks of B. cereus infections [2–4]. In a clinical setting, B. cereus infection frequently occurs as emetic and diarrhoeal food poisoning [1]. B. cereus is an opportunistic pathogen because it sometimes causes fatal infection to immunosuppressed hospitalised patients [5]. We usually consider B. cereus as a contaminant when it is isolated from clinical specimens. However, B. cereus causes a number of systemic and local infections, including fulminant bacteraemia, central nervous system involvement (meningitis and brain abscesses), endophthalmitis, catheter-related bloodstream infection (CRBSI) and pneumonia [1, 6]. Of these various infections, B. cereus bloodstream infection (BSI) seems to be the most severe condition. However, clinical patient reviews for all inpatients have not been performed, except for immunocompromised patients [5, 7].

Most B. cereus isolates are resistant to β-lactams, such as penicillins and third-generation cephalosporins [8]. Thus, vancomycin is used as a first-line drug [9]. However, we need data for alternative agents, such as teicoplanin, imipenem, clindamycin and trimethoprim-sulfamethoxazole, because the antimicrobial susceptibilities of clinical B. cereus isolates are rarely reported. Daptomycin and linezolid have emerged as promising antimicrobials for infections due to Gram-positive pathogens [10].

B. cereus produces several virulence factors, including toxins and enzymes [11]. Certain B. cereus isolates, closely related to B. anthracis with specific virulence genotypes, can cause severe and even fatal infections in patients who appear to be otherwise healthy [12]. Whether virulence genes of B. cereus BSI isolates determine their pathogenicity is an issue open to question because it has not yet been investigated.

To reveal the microbiological characteristics of B. cereus BSI isolates that cause nosocomial infections, we investigated the epidemiology, environmental contamination, antimicrobial susceptibility towards antibiotics, including daptomycin and linezolid, virulence genotypes and genetic similarity. We also characterised the clinical presentation of B. cereus BSI through analyses of risk factors for BSI and mortality in association with virulence genotypes.

Methods

Settings

From January 2008 to September 2013, all patients with B. cereus blood culture were identified through the records of the clinical microbiology laboratory at Kyoto University Hospital, a tertiary-care, 1,182-bed university hospital in Japan. For each month, the number of B. cereus isolates from all of the specimens and the number of patients with B. cereus BSI were counted. BSI was defined as the isolation of B. cereus from at least two blood cultures from separate sites at the same time with signs and symptoms of infection (such as fever and chills) [13]. We considered the blood to be contaminated if patients had specific infection sites due to non-B. cereus pathogens that improved without using antimicrobial agents with in vitro activities against B. cereus and when only one blood culture out of two or more was positive for B. cereus. The Ethics Committee of Kyoto University Graduate School and Faculty of Medicine (E2001) approved this study and waived the need for obtaining informed consent from each patient.

Environmental culture survey

We conducted an environmental survey because a significant number of patients contracted BSIs during September 2013. The freshly received towels, pyjamas and bed sheets were collected in September 2013 from three randomly chosen wards representing each hospital building. We crumpled 20 × 20-cm (400-cm2) pieces of them with 40 ml of sterilised water. We spread 100 μl of this fluid on sheep blood agar plates and aerobically incubated the plates at 37 °C for 24 h. The bacterial load was estimated by counting the colonies on agar plates which were identified as B. cereus.

Identification and antibiotic susceptibility

Species identification was performed with BBL Crystal (Becton Dickinson, Sparks, MD, USA) Gram-positive identification panels. Antibiotic susceptibility was evaluated by microdilution using Dry Plate Eiken (Eiken, Tokyo, Japan) and interpreted according to the Clinical and Laboratory Standards Institute (CLSI) criteria (CLSI document M45-A2) [14]. The antibiotics tested were the following: ampicillin, cefazolin, imipenem, levofloxacin, gentamicin, erythromycin, clindamycin, trimethoprim-sulfamethoxazole, teicoplanin and vancomycin. The isolates with intermediate susceptibility were classified as resistant. Etest susceptibility testing of daptomycin and linezolid was performed using cation-adjusted Mueller–Hinton agar plate following the CLSI recommendations (CLSI document M07-A9) [15].

Virulence factors

To detect the virulence genes of BSI isolates, polymerase chain reaction (PCR) assays were performed with the enterotoxin FM (entFM) gene [16], the enterotoxin T (bceT) gene [17], the haemolytic enterotoxin complex (hblACD) genes [18], the non-haemolytic enterotoxin (NHE) complex (nheABC) genes, the cytK gene [11] and the piplc gene [19].

Genomic fingerprinting

We analysed BSI isolates, isolates obtained by environment survey and isolates that were regarded as contaminated in August and September 2013. A randomly amplified polymorphic DNA PCR (RAPD-PCR) technique, based on the use of the Escherichia coli M13 phage-derived primer PM13 (5′-GAGGGTGGCGGCTCT-3′), was performed to analyse the genetic relatedness [10]. The profiles obtained by RAPD were analysed with GelCompar II version 4.6 (Applied Maths, Sint-Martens-Latem, Belgium). Cluster analysis was applied using the unweighted pair-group method based on the Dice coefficient. Isolates displaying more than 90 % similarity were considered as belonging to the same cluster [10].

Clinical characteristics and risk factors for mortality

We, as infection control team doctors, investigated all BSI patients. Antimicrobial agents were recommended for 14 days after the last positive blood culture and resolution of signs and symptoms. Based on the Infectious Diseases Society of America (IDSA) guidelines [9], we also recommended removal of intravenous catheters when B. cereus bacteraemia occurred.

The medical records of patients who were diagnosed as having BSI or who had contaminated blood culture were reviewed. We excluded patients with polymicrobial BSI. For each year, the first BSI episode was included. The following data were recorded: age, gender, hospitalisation before/after BSI, severity, underlying conditions (solid malignancy, cardiovascular disease, haematologic malignancy and central nervous system disease), Charlson index [20], assisted ventilation, indwelling urinary catheter, drain, central or peripheral venous catheterisation, parenteral nutrition, enteral nutrition, corticosteroid therapy, other immunosuppressive therapy, prior antibiotic therapy within 3 months, prior surgical procedures within 3 months, antibiotic treatment agents and mortality (30-day and in-hospital). Nosocomial BSI was defined as BSI that occurred ≥48 h after hospital admission. CRBSI was defined as the isolation of B. cereus from the catheter tip culture or with no apparent source of BSI according to the IDSA guidelines [9]. Empirical therapy was defined as the first antibiotic treatment initiated at the time of blood culture and continued for more than 24 h. Adequate treatment was defined as treatment with antimicrobial drugs which had in vitro susceptibility.

The risk factors for BSI were evaluated by comparison between nosocomial patients with BSI and contamination. We also investigated the risk factors for in-hospital mortality among patients with nosocomial infections having BSI. In addition to clinical information, the presence of virulence genes was compared between survived BSI and fatal BSI patients.

Statistical analyses

Categorical variables were compared using Fisher’s exact test. Continuous variables were compared using the Mann–Whitney U-test. A p-value < 0.05 was considered to be statistically significant. Multivariate logistic regression analysis was used to determine the independent risk factors for B. cereus BSI. All variables with a p-value < 0.1 on univariate analyses were used in the multivariate logistic regression model. We conducted our statistical analysis using Stata version 11.2 (StataCorp, College Station, TX, USA).

Results

From January 2008 to September 2013, an average of 3,091 specimens per month was submitted to the clinical laboratory. The number of specimens was not affected by seasonal variations. B. cereus was positive in an average of 2.1 specimens per month and BSI patients were identified at an average of 0.7 per month. Figure 1 shows that as many as a total of 82 of 71,975 specimens was positive from June to September (an average of 3.4 per month), but only a total of 24 of 71,895 specimens was positive for B. cereus from January to April (an average of 1.0 per month, p < 0.001). The increase in infected cases was observed during summer irrespective of year. A total of 51 patients were diagnosed as B. cereus BSI and 41 as contaminated. The number of patients per month that developed B. cereus BSI increased from 0.2 between January and April to 1.4 between June and September (5 patients vs. 34 patients, p < 0.001), showing the similar trend for all specimens positive for B. cereus.

Fig. 1
figure 1

Number of specimens positive for Bacillus cereus and patients with B. cereus bloodstream infection (BSI) along with the average monthly ambient temperatures in Kyoto city. The black bars indicate summer season between June and September. A total of 82 of 71,975 specimens were positive from June to September, but only a total of 24 of 71,895 specimens were positive for B. cereus from January to April (p < 0.001). The increase in summer was observed irrespective of the year. The number of patients that developed B. cereus BSI increased from 0.2 patients between January and April to 1.4 between June and September (p < 0.001), showing the similar trend for all specimens positive for B. cereus

Full size image

Environmental culture survey

During September 2013, five patients from different wards were diagnosed as having BSIs. Environmental contamination was suspected and the culture survey was conducted. From a cleaned pyjama from one ward, 88 colony-forming units/cm2 of B. cereus were isolated. B. cereus was not detected from pyjamas from the other two wards. Towels and sheets of all three wards were negative for B. cereus.

RAPD

Figure 2 shows that the environmental isolate detected during September 2013 was different from all of the 51 BSI and ten contaminated isolates. Among BSI isolates, three clusters of two isolates were found. The two isolates in a cluster were obtained in the same ward during October 2011 and November 2011. Isolates in the other clusters were obtained during the 2-month period but did not share the ward.

Fig. 2
figure 2

Dendrogram obtained by cluster analysis of randomly amplified polymorphic DNA polymerase chain reaction (RAPD-PCR) fingerprint patterns of the 51 BSIs, ten contaminated blood cultures and one environmental B. cereus isolate. The environmental isolate was different from all BSI and contaminated isolates. Among the BSI isolates, three clusters of two isolates were found. The two isolates (strains 9048 and 9235) in a cluster were obtained from the same ward during September 2011 and October 2011. Isolates in the other clusters were obtained during the 2-month period but from different wards

Full size image

Antibiotic susceptibility

The minimum inhibitory concentration (MIC) distributions and susceptibility of antimicrobial agents against the 51 bloodstream B. cereus isolates are detailed in Table 1. Vancomycin (100 %), imipenem (98 %) and levofloxacin (94 %) showed high susceptibility rates, whereas low susceptibility rates were observed for cefazolin (37 %) and ampicillin (3.9 %). The MICs required to inhibit the growth of 90 % of organisms were 4 μg/mL for daptomycin and 2 μg/mL for linezolid.

Table 1 Minimum inhibitory concentration (MIC) distributions of antimicrobial agents against the 51 bloodstream Bacillus cereus isolates
Full size table

Clinical characteristics and risk factors for BSI and mortality

All 51 BSI patients were diagnosed as having nosocomial BSI. Among 41 patients who were diagnosed as contaminated blood culture, four outpatients were excluded and 37 nosocomial patients were included for further analysis. The clinical characteristics of patients with BSI and contamination, and potential risk factors for developing BSI, are shown in Table 2. Among the 51 BSI patients, no neonates (≤28 days old) were diagnosed as B. cereus BSI. There was no significant difference in age, sex, Charlson index or days of hospitalisation prior to positive blood culture between the BSI and contaminated patients. Solid organ tumour (35 %) was the major underlying condition in both the BSI and contaminated patients. Only 12 % of BSI patients had haematological malignancy. Vascular catheters were a common factor in all of the BSI patients. In the univariate analysis, factors significantly associated with B. cereus BSI were recent surgery [odds ratio (OR) 2.98, 95 % confidence interval (CI) 1.14–7.76, p = 0.026] and urinary catheter (OR 9.58, 95 % CI 3.54–26.0, p < 0.001). In the multivariate analysis, urinary catheter (OR 6.93, 95 % CI 2.40–20.0, p < 0.001) was the independent risk factor associated with B. cereus BSI, whereas haematological malignancy (OR 1.15, 95 % CI 0.23–5.71, p = 0.89), recent surgery (OR 2.94, 95 % CI 0.90–9.60, p = 0.074) and central venous catheter (OR 0.67, 95 % CI 0.19–2.34, p = 0.53) were not associated with B. cereus BSI.

Table 2 Clinical features of the 51 patients with Bacillus cereus bloodstream infection (BSI) and the 37 patients with contaminated blood culture
Full size table

B. cereus BSI was associated with 20 % of in-hospital mortality and 10 % of 30-day mortality. The median duration of onset of BSI to death was 29 days. All of the BSI patients were diagnosed as having CRBSI. Risk factors for in-hospital mortality among the 51 BSI patients are shown in Table 3. Urinary catheter (OR 12.3, 95 % CI 0.67–225, p = 0.045; this P value calculated by Fisher’s exact test shows statistical significance. Due to the nature of the data, the OR and CI could only be calculated using an approximate method) and higher Charlson index (OR 1.99, 95 % CI 1.26–3.12, p = 0.0073) were significant factors for in-hospital mortality in the univariate analysis. The kind of catheters (central or peripheral), other medical device, medications and method for administering nutrients were not significant risk factors.

Table 3 Risk factors for in-hospital mortality among the 51 patients with Bacillus cereus bloodstream infection
Full size table

Virulence factors

Virulence genes for the 51 BSI isolates are summarised in Table 3. Most isolates from the BSI group carried nheABC (88 %). Profiles of the nhe, entFM, hblCDA, cytK, bceT and piplc genes were not significant risk factors for in-hospital mortality.

Discussion

This study elucidated the epidemiology of B. cereus nosocomial infections, including seasonal trend, association of environmental contamination and genetic relationship of BSI isolates. We also determined the risk factors for developing BSI and in-hospital death among inpatients.

Some studies indicated that B. cereus nosocomial infections occurred notably in summer, suggesting an association with high room temperature [2, 3]. In our hospital, the number of B. cereus-positive cases showed a similar tendency. Moreover, we also found that the higher isolation rate of B. cereus from the bloodstream occurred from June to September. The increased bacterial load in a patient, which was supported by the number of specimens positive for B. cereus, may be associated with an increase of BSIs. Thus, monitoring the number of B. cereus from clinical samples has the potential for a novel way to predict an increase of BSIs. This strategy seems to be time-saving and easy to perform.

Complete removal of B. cereus from linen is impractical, and there are no universally acceptable levels in washed linen, sheets and pyjamas [3]. Thus, these can be a source of pollution and can cause outbreaks [2]. In this study, B. cereus was detected from a fresh unused pyjama from one floor. However, the RAPD-PCR profile pattern of this isolate was not related to BSI isolates and contaminated isolates from blood culture in the same period. Furthermore, most of the BSI isolates were unrelated, even in the outbreak period that occurred in September 2013. The route of the current B. cereus contamination could not be determined. However, these results suggest that the increase of BSI in summer was not associated with a specific strain. The contribution of common environmental source or patient-to-patient transmission was not likely. Thus, we support the hypothesis that an increase in temperature promotes the growth of each B. cereus strain, subsequently causing an increase of nosocomial infections.

B. cereus shows a wide range of variation in virulence genotypes. Recent reports of B. cereus encouraged us to seek the genetic determinants responsible for its virulence [3, 5, 12]. Our study was not able to explain the virulence determinants that affected the clinical presentation. The nheABC gene is the major cytotoxic membrane-damaging factor [21]. In B. cereus food-bourne diarrhoeal disease, nheABC was considered as the main virulence factor and was found in more than 90 % of isolates [12, 16, 22]. nheABC was also dominant in our BSI isolates. The detection rates of entFM, hblCDA, cytK, bceT and piplc among BSI isolates were lower than those in the previous studies investigating food poisoning isolates　[11, 19, 22]. We speculated that BSI isolates have different microbiological characteristics from food poisoning isolates.

Antimicrobial susceptibility testing is important in guiding therapeutic decisions. In this study, 100 % of isolates were susceptible to vancomycin. This keeps vancomycin as a first empirical therapy of defence against B. cereus. Breakpoints for daptomycin, linezolid and teicoplanin have not been shown in the CLSI or European Committee on Antimicrobial Susceptibility Testing (EUCAST) guidelines. Using CLSI MIC breakpoints established for Corynebacterium spp., only 16 % of BSI isolates were susceptible to daptomycin (≤1 μg/mL) and 95 % were susceptible to linezolid (≤2 μg/mL). Linezolid may become one of the alternative drugs when first-line therapy fails or in the event of adverse events requiring treatment discontinuation. More investigations are needed in order to elucidate their clinical utilities.

Previous studies showed that B. cereus infection often develops in hospitalised neonates [3] or hospitalised patients with haematological malignancies [5]. In our study, no neonates were diagnosed as B. cereus BSI during the study period. Patients with haematological malignancies were fewer in BSI patients compared to contaminated patients (p = 0.054). Recent surgery was dominant in BSI patients (45 %) and was a significant factor for developing BSI (p = 0.026). Wound contamination with B. cereus can easily occur due to the ubiquitous presence and the capacity to persist in gauze [1]. Wound contamination, perioperative use of antibiotics and medical devices may be associated with BSI development. Urinary catheter is another significant predictor and the only independent predictor of B. cereus BSI. However, B. cereus is not considered to cause urinary tract infection [23]. The reason why urinary catheter is related to the development of BSI seems uncertain. The presence of urinary catheter or some other medical devices may promote the acquisition and colonisation of B. cereus because B. cereus can adhere to artificial devices via biofilm formation [1].

B. cereus BSI was associated with an in-hospital mortality of 20 % in our hospital. Urinary catheter and higher Charlson index were significant factors for mortality. It is reasonable that patients with medical devices like urinary catheters and comorbid illnesses may be at a greater risk of complications or death. All BSI patients were diagnosed as CRBSI. The ability of B. cereus to form a biofilm matrix and its adherence properties accounted for its relation with central venous catheter infections [6]. Central venous catheters were the major focus of BSI and a risk factor for mortality [6]. Peripheral catheters might have significantly contributed to the development of BSI in our study because 69 % of the BSI patients did not have central venous catheters but had peripheral catheters. We must be careful when treating central venous as well as peripheral catheters in order to avoid CRBSIs. Unnecessary vascular catheters must be removed as soon as possible, and we always have to reconsider whether patients really need vascular catheters. Considering that urinary or vascular catheters underlay B. cereus BSI and the number of BSIs increased in summer, adjusting room temperatures according to seasons may be effective in protecting such high-risk patients from BSI.

The major limitation of this study is the relatively small number of BSI patients. A greater number of patients was needed to make a multivariate model and to elucidate the association of virulence genotypes. Investigations for skin or urine colonisation with B. cereus were not conducted, but they would help in understanding the mechanisms of BSI development.

In conclusion, this study demonstrated that the number of B. cereus isolates and BSIs increased in summer without clonal spread of a specific strain. B. cereus BSI can develop in all hospitalised patients and was associated with 20 % of in-hospital mortality. We should be careful in regards to the potential risk for B. cereus BSI among inpatients who have indwelling catheters, including vascular or urinary catheters, or those who have undergone recent surgery.