The Majority of Bacillus subtilis Strains Isolated From Blood Cultures Were Derived From Traditional Japanese Fermented Soybeans Natto: A Single-center Retrospective Study

Abstract

Background

Natto consumption has been associated with improved cardiovascular and survival outcomes. Bacillus subtilis variant natto (B. subtilis var. natto) is essential for soybean fermentation in natto production. Recently, several cases of B. subtilis var. natto bacteremia have been reported, although the contribution of B. subtilis var. natto remains unclear. This study assessed the prevalence and clinical impact of B. subtilis var. natto in B. subtilis bacteremia.

Methods

This retrospective cohort study analyzed medical records of patients with positive B. subtilis blood cultures at Tenri Hospital from 1 April 2016 to 31 March 2023. True bacteremia required portal-of-entry or, for intra-abdominal/abscess, 2-physician adjudication; if no portal, 2 or more same-day blood culture-positive sets. Genetic testing confirmed B. subtilis var. natto through specific mutations in the bioF and bioW genes.

Results

Of 4634 positive blood cultures, 70 (1.5%) were identified as B. subtilis, with 69 (99%) classified as B. subtilis var. natto. Of these, 25 cases (36%) were confirmed as true bacteremia. The primary diagnoses included intra-abdominal infections, pneumonia, and urinary tract infection. The median patient age was 79 years, with 9 being women. The median Pitt bacteremia score was 0 (interquartile range: 0–1). Seven patients (28%) required urgent surgery or endoscopic procedures, while 4 (16%) died within 30 days.

Conclusions

Most B. subtilis blood strains isolates were B. subtilis var. natto. This bacteremia carried significant severity, with a 16% 30-day mortality rate and 28% requiring urgent interventions. Clinicians should not dismiss B. subtilis var. natto, a probiotic strain, as harmless.

In a retrospective study from a Japanese hospital, 69 of 70 Bacillus subtilis blood isolates were B. subtilis var. natto, of which 25 were confirmed as true bacteremia, with a 16% 30-day mortality rate and 28% requiring urgent interventions.

Bacillus subtilis, a well-characterized Gram-positive bacillus, is a key model organism for physiological and metabolic research because of its accessible genetics and versatile genetic manipulation tools, including targeted gene deletions, promoter replacements, and heterologous expression of biosynthetic gene clusters [1]. B. subtilis comprises multiple subspecies, including B. subtilis variant natto (B. subtilis var. natto), which is essential for fermenting soybeans to produce natto. This fermentation process is facilitated by the unique enzymatic activity of B. subtilis var. natto, particularly its production of nattokinase, a fibrinolytic enzyme with potential health benefits. Natto is a traditional Japanese food with a sticky texture, prepared from fermented soybeans (Figure 1). It pairs well with warm rice and is a staple food in Japanese cuisine, with increasing international exports in recent years [2]. Unlike B. subtilis, natto bacteria require biotin for growth, which facilitates their classification through biotin-supplemented media [3]. Numerous studies have documented the health benefits of natto consumption. In addition to its probiotic properties, natto consumption is associated with a reduced risk of cardiovascular diseases [4] and lower mortality [5].

Figure 1.

Natto, sold in single-serving packs in Japan (above), is known to increase in viscosity and improve in flavor when thoroughly mixed. A significant portion of the Japanese population consumes natto with hot rice (below).

By contrast, some cases of B. subtilis bacteremia associated with probiotic use have been reported [6, 7]. In clinical microbiology laboratories, B. cereus is the most frequently isolated Bacillus species, while B. subtilis accounts for ∼10%–30% of Bacillus isolates, depending on the setting [8, 9]. Among B. subtilis isolates, recent genomic studies suggest that a subset corresponds to B. subtilis var. natto, although the exact proportion remains unknown due to limitations in routine identification methods [10]. Furthermore, natto consumption has been reported as a risk factor for B. subtilis bacteremia [11]. Notably, recent genetic analyses of B. subtilis detected in blood cultures have identified several cases of B. subtilis var. natto bacteremia, including elderly or immunocompromised patients who had recently consumed natto and developed bacteremia likely of gastrointestinal origin [12, 13]. However, the proportion of B. subtilis bacteremia cases attributable to B. subtilis var. natto remains unclear. The current limitations of matrix-assisted laser desorption/ionization time-of-flight mass spectrometry (MALDI-TOF MS) technology makes it difficult to differentiate B. subtilis from B. subtilis var. natto. Therefore, genetic analysis is required to accurately identify B. subtilis var. natto [14, 15]. Mutations in the bioF and bioW genes are reliable molecular markers [16]. The bioF genes of B. subtilis var. natto exhibit a characteristic partial deletion of ∼50 bp, while the bioW gene contains a nonsense mutation [16]. These mutations have not been observed in standard laboratory strains such as B. subtilis 168 or in environmental isolates, indicating high strain specificity [15]. They are consistently detected in commercial natto strains and are associated with biotin auxotrophy, a phenotypic hallmark of B. subtilis var. natto. Given the increasing global export of natto [2], it is crucial to determine the proportion of B. subtilis var. natto bacteremia among all B. subtilis bacteremia cases to gain a deeper understanding of its clinical significance, including its prevalence, clinical progression, severity (including mortality rates), and antimicrobial susceptibility.

Therefore, in this single-center retrospective study, we performed a genetic evaluation of B. subtilis detected in blood cultures and investigated the general prevalence of B. subtilis bacteremia and patient background along with the clinical course and outcomes.

METHODS

Study Design

This retrospective cohort study was conducted at Tenri Hospital, a 715-bed hospital in Nara, Japan. To ensure methodological rigor, we followed the Strengthening the Reporting of Observational Studies in Epidemiology statement for reporting observational studies [17]. The Institutional Review Board of Tenri Hospital approved the study protocol (no. 1429). Written informed consent was not required owing to the retrospective nature of the study. An opt-out notice was available on the hospital website, allowing participants to withdraw their data if they chose.

Patients and Baseline Characteristics

We reviewed cases of patients with B. subtilis isolated from blood cultures from the Microbiology for Clinical Investigation database at Tenri Hospital between 1 April 2016 and 31 March 2023. The portal of entry was defined as the site of infection adjudicated by 2 board-certified infectious disease physicians on the basis of medical-record review. B. subtilis bacteremia was defined as cases in which B. subtilis was detected in one or more sets of blood cultures and was subsequently treated with effective antibiotics, and the following 2 sets of criteria:

1. When a presumed site of infection was present:

   • Pneumonia, urinary tract infection, and skin and soft tissue infection were classified as true bacteremia if Gram-positive rods typical of B. subtilis var. natto were isolated from corresponding clinical specimens.

   • Intra-abdominal infections and other abscess formations were classified as true bacteremia when 2 board-certified infectious disease physicians, incorporating the radiologists’ imaging interpretations, independently determined that the B. subtilis isolate represented true bacteremia.

2. When no portal of entry was identified:

   • Cases without an identifiable portal of entry were considered true bacteremia only if B. subtilis was detected in 2 or more blood culture sets on the same day; all others were classified as contaminated blood cultures.

We initially identified B. subtilis using MALDI-TOF MS (MALDI Biotyper^® version 3.1, Bruker Daltonics, Billerica, MA, USA) for accurate identification of bacterial species [18]. We extracted the following parameters from patients’ medical records: age, sex, site of diagnosis, underlying diseases, history of natto consumption before diagnosis, Pitt bacteremia score [19], and the updated Charlson Comorbidity Index (CCI) at the time of bacteremia diagnosis, which was systematically assessed for every individual [20]. We also extracted microbiological data, including the time to positivity in blood cultures, the number of positive blood culture sets, and the presence of polymicrobial bacteremia, including the simultaneous detection of microorganisms other than B. subtilis. Finally, we extracted data on the bacteremia etiology, invasive treatments, such as endoscopic procedures and surgeries, and mortality within 30 days.

Statistical Analysis

To explore factors associated with mortality, we performed univariate analyses comparing clinical variables between patients who died and those who survived. All statistical analyses were performed using R for macOS (version 4.5.1; R Foundation for Statistical Computing, Vienna, Austria; https://cran.r-project.org/bin/macosx/). Categorical variables (eg, female sex and nosocomial infection) were compared using Fisher's exact test, and continuous variables (eg, age, antibiotic duration, and Pitt bacteremia score) were analyzed using the Mann–Whitney U test. Missing values were excluded on a per-analysis basis, and two-sided P-values <.05 were considered statistically significant. Categorical variables are presented as numbers and percentages, and continuous variables as medians and interquartile ranges (IQRs). In addition, we compared clinical outcomes between the contamination and true bacteremia groups, restricted to death and urgent intervention. Group differences were evaluated with Fisher's exact test, and effect sizes were expressed as risk differences (RD; risk in the true bacteremia group minus that in the contamination group) with 95% confidence intervals estimated primarily by the Newcombe (Wilson score) method and, when sparse counts occurred, by the Agresti–Caffo correction.

Gene Sequencing Analysis

Genomic DNA was extracted from all clinical isolates using the QIAamp DNA Mini Kit (QIAGEN, Venlo, Netherlands) according to the manufacturer's instructions. To confirm strain identity, we analyzed the partial deletion of bioF and the nonsense mutation in bioW using Sanger sequencing following a previously described method [16].

Microbiological Information

All clinical isolates identified as B. subtilis var. natto were evaluated to determine the minimum inhibitory concentrations of penicillin, ampicillin, clavulanic acid/amoxicillin, ceftriaxone, imipenem, meropenem, amikacin, gentamicin, erythromycin, clindamycin, doxycycline, levofloxacin, ciprofloxacin, vancomycin, linezolid, rifampicin, and sulfamethoxazole/trimethoprim for the organism. We performed susceptibility tests for B. subtilis var. natto using the broth microdilution method with cation-adjusted Mueller–Hinton broth according to the guidelines in Clinical and Laboratory Standards Institute (CLSI) M45-Ed3 (2016) [21]. Testing was performed using the OPTPANEL MP (Kyokuto Pharmaceutical Industrial Co., Ltd., Japan) and calcium-adjusted Mueller–Hinton broth from the same manufacturer. Isolates were first precultured at 35°C under ambient atmospheric conditions for 24 h to prepare the inoculum, and MICs were visually determined after 16–20 h of incubation for susceptibility testing in accordance with the CLSI M45 criteria for Bacillus spp. (excluding B. anthracis).

RESULTS

Of the 4634 positive blood culture samples, we detected 70 blood culture-positive cases (1.5%) of B. subtilis. Based on gene sequencing analysis using the partial deletion of bioF and nonsense mutation of bioW, we identified 69 strains as B. subtilis var. natto and 1 strain as B. subtilis (Figure 2). The results of gene sequencing analysis of bioF and bioW for all strains are provided in Supplementary Figures 1. Among these, 44 were classified as contamination cases (64%), whereas the remaining 25 were considered definitive cases of bacteremia. Table 1 summarizes the clinical and microbiological characteristics of the 25 true cases of B. subtilis var. natto bacteremia. The median age of patients was 79 years (IQR, 73–85 years), and 9 were women (36%). Notably, 23 cases (92%) were community-onset, and 5 patients (20%) were immunocompromised. None of the cases had a documented history of natto consumption before disease onset, preventing further evaluation. The primary clinical diagnosis was intra-abdominal infections (13 cases), pneumonia (6 cases), and urinary tract infections (5 cases). Polymicrobial bacteremia was observed in 9 cases (36%). The coisolated organisms included Escherichia coli, Pseudomonas aeruginosa, Clostridium perfringens, Enterococcus avium, Ruminococcus gnavus, Eggerthella lenta, Gemella morbillorum, and coagulase-negative staphylococci (Staphylococcus epidermidis, S. caprae, and S. simulans). The median Pitt bacteremia score was 0 (IQR, 0–1), and the median updated CCI was 2 (IQR, 0–4). The median duration of antibiotic treatment was 9 days (IQR, 5–13), and 21 cases were successfully treated with noncarbapenem β-lactam antibiotics.

Figure 2.

Classification of 70 blood culture-positive cases of Bacillus subtilis strains. B. subtilis, Bacillus subtilis.

Table 1.

Clinical and Microbiological Characteristics of 25 Cases of Bacillus Subtilis Var. natto Bacteremia

	Deceaseda (n = 4)	Survivors (n = 21)	P-value
Age (years, IQR)	79.5 (IQR: 70.0–82.8)	78.0 (IQR: 73.0–85.0)	1.0
Female (n, %)	2 (50.0%)	7 (33.3%)	.60
Time to positivity in blood cultures (hours, IQR)	24.1 (IQR: 23.1–24.7)	22.8 (IQR: 19.7–31.3)	.91
Patients with 2 positive blood culture sets (n, %)	1 (25.0%)	5 (23.8%)	1.0
Nosocomial infection (n, %)	0 (0%)	2 (9.5%)	1.0
Immunosuppression (n, %)	0 (0%)	5 (23.8%); chemotherapy with cytotoxic agents: 4, steroids: 1,	.55
Pitt bacteremia score (points, IQR)	5.0 (IQR: 1.8–9.0)	0.0 (IQR: 0.0–1.0)	.005
Updated Charlson comorbidity index (points, IQR)	4.0 (IQR: 3.0–4.0)	2.0 (IQR: 0.0–3.0)	.30
Polymicrobial bacteremia (n, %) (Microorganisms other than B. subtilis strains)	2 (50.0%)	7 (33.3%)	.60
Abdominal infection (n, %)	2 (50.0%)	11 (52.4%)	1.0
Urinary tract infection (n, %)	0 (0%)	5 (23.8%)	.55
Pneumonia (n, %)	2 (50.0%)	4 (19.0%)	.23
Retroperitoneal abscess (n, %)	0 (0%)	1 (4.8%)	1.0
Duration of antibiotic treatmentb (days, IQR)	6.0 (IQR: 4.0–7.2)	10.0 (IQR: 7.0–13.0)	.07
The number of cases treated with carbapenem antibiotics (n, %)	1 (25.0%)	3 (14.3%)	.53
Patients who required urgent surgery or endoscopic procedure (n, %)	0 (0%)	7 (33.3%) surgery: 6, ERCP 1	.30

All patients who required urgent surgery or endoscopic procedures were alive.

IQR, interquartile range; n, number of patients; ERCP, endoscopic retrograde cholangiopancreatography.

^a“Deceased” refers to patients who died within 30 days.

^bCalculation of the median duration of antimicrobial therapy, excluding deceased patients.

Table 2 provides the details of patients who required emergency surgery or intervention and those who died within 30 days. In this study, 7 patients (28%) underwent urgent surgery or endoscopic procedures, whereas 4 (16%) died within 30 days. Of these 11 patients, 6 were men and 5 were women, with an age range of 46–91 years. Furthermore, all cases had community-acquired bacteremia, and 9 (82%) had underlying conditions. Notably, only 2 patients (18%) were undergoing immunosuppressive therapy. Of the 7 patients who required emergency surgery or intervention, the most common condition was peritonitis (4 patients), followed by acute cholecystitis (3 patients); all 7 patients exhibited clinical improvement following the intervention. In contrast, of the 4 patients who died within 30 days, 2 died due to intra-abdominal infections, and 2 died of pneumonia. Supplementary Table 1 provides detailed information on all patients infected with B. subtilis var. natto bacteremia, excluding those listed in Table 2. Based on statistical analysis comparing deceased and surviving patients, the Pitt bacteremia score was the only variable significantly associated with mortality. No other factors—including polymicrobial bacteremia, immunosuppression, and the use of carbapenem antibiotics—showed a statistically significant difference (Table 1). When outcomes were compared between the contamination and true bacteremia groups, the proportion requiring urgent surgery or endoscopic intervention was higher in the true bacteremia group (28.0% [7/25] vs 0.0% [0/44]; RD, 0.280; 95% CI, .097–.452; P < .001), whereas 30-day mortality did not differ significantly (16.0% [4/25] vs 6.8% [3/44]; RD, 0.092; 95% CI, −.056–.285; P = .245) (Supplementary Table 2).

Table 2.

Details of 7 Patients With Bacillus subtilis Var. natto Bacteremia Requiring Urgent Intervention (No. 1–7) and 4 Patients With Bacillus subtilis Var. natto Bacteremia Who Died Within 30 Days (No. 8–11)

No.	Sex	Age	Community-onset or Nosocomial Onset	Underlying Diseases	Immunosuppression	CCI	PBS	The Number of Positive Blood Culture Sets	Polymicrobial Bacteremia	Diagnosis	Antibiotic Choice	Duration Of Antibiotic Treatment	Urgent Intervention or Deatha
1	M	49	Community	None	No	0	2	1	No	Cholangitis	CMZ →CPFX →ABPC	14	ERCP/alive
2	M	73	Community	Choledocholithiasis	No	0	0	1	Yes (Staphylococcus simulans)	Cholecystitis	CMZ	4	Cholecystectomy/alive
3	F	71	Community	Colon peritoneal carcinomatosis	Yes (chemotherapy)	7	0	1	No	Peritonitis due to colon obstruction	ABPC/SBT →CMZ	8	Colostomy/alive
4	M	46	Community	None	No	0	0	2	Yes (Eggerthella lenta and Escherichia coli)	Gangrenous appendicitis	CMZ	13	Appendectomy/alive
5	F	77	Community	Dementia, epilepsy	No	2	1	1	No	Peritonitis due to sigmoid colon diverticular perforation	MEPM	10	Sigmoidectomy/alive
6	F	85	Community	Dementia, IPMN	Yes (chemotherapy)	10	1	2	Yes (Clostridium perfringens)	Peritonitis due to transverse colon perforation	MEPM	7	Transverse colectomy/alive
7	M	82	Community	COPD, Alcoholism, Lumbar fracture, Cerebral Aneurysm	No	3	0	1	No	Cholecystitis	CMZ →MEPM	11	Cholecystectomy/alive
8	M	79	Community	Type 2 diabetes, dementia, hepatitis C	No	4	8	2	No	Pneumonia	ABPC/SBT + AZM →VCM + MEPM	8	None/died
9	M	80	Community	Type 2 diabetes, dementia	No	4	2	1	No	Cholecystitis	MEPM	5	None/died
10	F	43	Community	Hydrocephalus, epilepsy, intellectual disability	No	0	12	1	GPC (unidentified)	Panperitonitis due to colonic perforation	None	0	None/died
11	F	91	Community	Type 2 diabetes, dementia, poststroke	No	4	1	1	Yes (Staphylococcus caprae)	Pneumonia	ABPC/SBT	7	None/died

Abbreviations: M, male; F, female; IPMN, intraductal papillary mucinous neoplasm; COPD, chronic obstructive pulmonary disease; CCI, the updated Charlson Comorbidity Index; PBS, the Pitt Bacteremia Score; GPC, Gram-positive cocci; CMZ, cefmetazole; CPFX, ciprofloxacin; ABPC, ampicillin; ABPC/SBT, ampicillin/sulbactam; MEPM, meropenem; AZM, azithromycin; VCM, vancomycin; ERCP, endoscopic retrograde cholangiopancreatography.

^aAll patients who required urgent surgery or endoscopic procedures were alive.

Table 3 presents the antibiotic susceptibility of B. subtilis var. natto. Remarkably, susceptibility to almost all evaluated antimicrobial agents was 100%, except for one isolate, which was evaluated as having intermediate susceptibility to clindamycin.

Table 3.

Antimicrobial Susceptibility of 69 Bacillus subtilis Var. natto Strains as Classified by Clinical and Laboratory Standards Institute

Antibiotics	Range (μg/mL)	MIC50 (μg/mL)	MIC90 (μg/mL)	CLSI			CLSI Susceptibility Breakpoint (μg/mL)
				Susceptible No. (%)	Intermediate No. (%)	Resistant No. (%)	Susceptible	Intermediate	Resistant
PCG	≤0.06	≤0.06	≤0.06	69 (100)	0 (0)	0 (0)	≤0.12	…	≥0.25
ABPC	≤0.06	≤0.06	≤0.06	69 (100)	0 (0)	0 (0)	≤0.25	…	≥0.5
AMPC/CVA	≤0.06/0.03	≤0.06/0.03	≤0.06/0.03	N.D.			N.D.
CTRX	≤0.06–4	0.25	1	N.D.			N.D.
IPM	≤0.12	≤0.12	≤0.12	69 (100)	0 (0)	0 (0)	≤4	8	≥16
MEPM	≤0.12	≤0.12	≤0.12	69 (100)	0 (0)	0 (0)	≤4	8	≥16
AMK	≤0.25–2	0.5	1	69 (100)	0 (0)	0 (0)	≤16	32	≥64
GM	≤0.25–1	≤0.25	≤0.25	69 (100)	0 (0)	0 (0)	≤4	8	≥16
EM	0.06–0.25	0.12	0.25	69 (100)	0 (0)	0 (0)	≤0.5	1∼4	≥8
CLDM	0.12–1	0.25	0.5	68 (98.6)	1 (1.4)	0	≤0.5	1∼2	≥4
DOXY	≤0.06–0.12	≤0.06	≤0.06	69 (100)	0 (0)	0 (0)	≤4	8	≥16
LVFX	≤0.03–0.25	0.06	0.12	69 (100)	0 (0)	0 (0)	≤2	4	≥8
CPFX	≤0.03–0.25	≤0.03	0.06	69 (100)	0 (0)	0 (0)	≤1	2	≥4
VCM	0.12–1	0.25	0.25	N.D.			N.D.
LZD	0.5–2	1	2	N.D.			N.D.
RFP	≤0.03–0.06	≤0.03	≤0.03	69 (100)	0 (0)	0 (0)	≤1	2	≥4
ST	≤0.3/0.015 −0.6/0.03	0.6/0.03	0.6/0.03	69 (100)	0 (0)	0 (0)	2/38	…	4/76

PCG, penicillin; ABPC, ampicillin; AMPC/CVA, clavulanic acid/amoxicillin; CTRX, ceftriaxone; IPM, imipenem; MEPM, meropenem; AMK, amikacin; GM, gentamicin; EM, erythromycin; CLDM, clindamycin; DOXY, doxycycline; LVFX, levofloxacin; CPFX, ciprofloxacin; VCM, vancomycin; LZD, linezolid; RFP, rifampicin; ST, sulfamethoxazole/trimethoprim; MIC, minimum inhibitory concentration; CLSI, Clinical and Laboratory Standards Institute; N.D., no data.

DISCUSSION

In this study, B. subtilis accounted for 1.5% of all blood cultures, 99% of which were genetically identified as B. subtilis var. natto. This overall positivity rate for B. subtilis is consistent with previous reports, which found that B. subtilis—regardless of strain—accounted for 0.8%–1.9% of bloodstream isolates in both Japanese and international settings [22, 23]. To the best of our knowledge, this is the first case series in which a detailed genetic analysis of B. subtilis bacteremia was conducted to determine whether it is derived from natto bacteria. To date, only 2 reports have summarized B. subtilis bacteremia in Japan. Hashimoto et al reported 10 cases of B. subtilis bacteremia [24]. Aoyagi et al conducted a case-control study and identified natto consumption as a risk factor for B. subtilis bacteremia, with an odds ratio of 3.3 [11]. However, neither of these studies incorporated genetic confirmation of the isolates. Our study is novel and suggests that when B. subtilis is detected in blood cultures in Japan, it is highly likely to be derived from natto bacteria. In other countries, there have been reports of bacteremia following the ingestion of B. subtilis-containing supplements, implying the possibility of direct gastrointestinal invasion by B. subtilis spores from the supplements [6, 7]. B. subtilis bacteremia is rare. However, it occurs worldwide, and some cases are attributed to the oral ingestion of foods or supplements containing B. subtilis.

According to our definition in this study, 25 out of 69 cases (36%) of B. subtilis var. natto detected in blood cultures were identified as true cases of bacteremia, and most of the B. subtilis var. natto bacteremia cases were of community-onset rather than hospital-acquired infections. This contrasts with the findings on Bacillus cereus bacteremia, which also originates from Bacillus species [25, 26], and bacteremia caused by the probiotic Clostridium butyricum MIYAIRI 588 strain, both of which are predominantly of nosocomial origin [27]. The predominance of community-onset B. subtilis var. natto bacteremia cases was likely influenced by differences in bacterial exposure levels. While natto is not included in the hospital diet at our institution—supporting the predominance of community-onset cases—nosocomial transmission may still occur, potentially through healthcare workers who have consumed natto and subsequently contaminated hands or devices. Fresh natto contains approximately 7.5 × 10⁹ CFU/g of B. subtilis var. natto [28], and ingesting such a large bacterial load could potentially lead to bacteremia. High-dose exposure may facilitate transient intestinal colonization and translocation into the bloodstream, particularly in individuals with compromised mucosal barriers. In addition, healthy patients capable of consuming natto are more likely to be found in outpatient settings, whereas patients requiring probiotics like MIYA-BM^Ⓡ are more often hospitalized. The most frequent foci of bacteremia included cases with no identifiable source, intra-abdominal infections, urinary tract infections, and pneumonia. Previous studies have documented cases of B. subtilis var. natto causing pneumonia [29], meningitis [30], and lower gastrointestinal perforation [31], highlighting the diverse range of potential portals of entry for bacteremia, as observed in our study. Moreover, a considerable proportion of severe cases were observed among patients with B. subtilis var. natto bacteremia. Notably, 7 patients required emergency interventions such as surgery or endoscopic procedures, all of whom recovered without mortality within 30 days. However, another 4 patients (16%) died within 30 days, demonstrating that B. subtilis var. natto bacteremia can result in severe and life-threatening complications. Overall, severe outcomes were observed in 11 out of the 25 cases (44%). In previous reports of B. subtilis var. natto bacteremia [12, 13, 29–32], only 1 surgical case has been documented [9], and importantly, there have been no documented cases of death. These findings suggest that B. subtilis var. natto bacteremia is associated with more severe clinical outcomes than previously recognized, warranting further investigation into its pathogenesis and optimal management strategies.

Based on the microbiological analysis, antimicrobial susceptibility testing of all isolated B. subtilis var. natto strains showed susceptibility to nearly all the antibiotics tested. Reportedly, B. subtilis var. subtilis is susceptible to penicillin, ampicillin, cefmetazole, imipenem, meropenem, clavulanic acid/amoxicillin, tazobactam/piperacillin, clindamycin, moxifloxacin, and metronidazole. B. subtilis var. subtilis was susceptible to nearly all antimicrobial agents reported in previous studies [12, 30–32]. As shown in Table 2 and Supplementary Table 1, a substantial number of patients in our cohort achieved clinical improvement with narrow-spectrum antibiotics, including penicillin and cephalosporin. These findings further reinforce the importance of antimicrobial stewardship in guiding treatment decisions. Once B. subtilis var. natto bacteremia is identified, antimicrobial stewardship should be implemented based on susceptibility results.

This study had some limitations. First, as a single-center study conducted in Japan, the prevalence of B. subtilis var. natto in B. subtilis bacteremia may not be generalizable to other institutions and countries. Given that natto is rarely consumed outside Japan, this high proportion may reflect a country-specific trend requiring further international investigation. Second, due to the study's retrospective nature, the classification of true bacteremia versus contamination was inherently limited by the available data, introducing potential uncertainty regarding the accuracy of diagnosis. Furthermore, several studies [7, 11, 33] defined bacteremia as cases in which B. subtilis was detected in blood cultures, employing a broader definition than that used in our study. By contrast, we applied stricter criteria—considering clinical context, culture sites, and expert review—to improve diagnostic precision, rather than relying solely on blood culture results. Generally, the incidence of true bacteremia when Bacillus species are detected in blood cultures is considered very low, ranging from 0% to 6.4% [34, 35]. However, our findings suggest that when B. subtilis var. natto is detected, a higher rate of true bacteremia should be expected. Third, regarding the history of natto consumption, we attempted to gather information through a comprehensive chart review; however, no documentation was available in any case, making evaluation impossible. Given this limitation, natto consumption should be investigated when cases of B. subtilis var. natto bacteremia are encountered. Fourth, our study could not definitively determine whether bacteremia was a direct cause of death or simply a marker of underlying patient vulnerability. The Pitt bacteremia score was significantly associated with the 30-day mortality rate. This indicates that initial severity is a critical factor for patient outcomes. We hypothesize that this higher initial severity is largely attributable to the patients’ underlying health conditions. Supporting this view, the median updated CCI, a measure of comorbidity burden, was numerically higher in deceased patients (4; IQR 3–4) compared to the entire cohort (2; IQR 0–4). This suggests that pre-existing comorbidities contribute to mortality risk by predisposing patients to a more severe initial presentation of bacteremia. Fifth, it is currently impossible to calculate the exact proportion of natto consumers who develop bacteremia. Survey data indicate that approximately 80% of Japanese consume natto at least once weekly [5]. Further studies on the potential incidence of bacteremia related to natto consumption would require epidemiological investigations incorporating large-scale databases and prospective cohort studies. Sixth, Bacillus spp. in nonblood specimens such as urine or sputum were identified only by Gram stain, without species-level confirmation. As Gram-positive rods are often not considered true pathogens, full identification is frequently omitted, potentially leading to misclassification of the infection source. Finally, the proportion of B. subtilis var. natto among the B. subtilis strains present in the environment remains unclear. If a significant portion of B. subtilis in the living environment were to be replaced by B. subtilis var. natto, infection can occur through environmental exposure, even without direct consumption of natto. Genomic surveillance will be needed to clarify whether it is food-associated or environmentally widespread.

CONCLUSION

This study revealed that the frequency of B. subtilis detected in blood cultures was 1.5%, with the majority originating from B. subtilis var. natto. True bacteremia accounted for 36% of the cases, 28% required urgent intervention, and the 30-day mortality rate was 16%. B. subtilis var. natto should not be routinely dismissed as a contaminant, particularly in intra-abdominal infections or in regions with high natto consumption.

Supplementary Data

Supplementary materials are available at Open Forum Infectious Diseases online. Consisting of data provided by the authors to benefit the reader, the posted materials are not copyedited and are the sole responsibility of the authors, so questions or comments should be addressed to the corresponding author.