Bacteroides fragilis is a Gram-negative anaerobic bacterium that, while being a key member of the human gut microbiota, is also a major pathogen in anaerobic infections, including bacteraemia with high mortality rates. Its pathogenicity is attributed to multiple virulence factors, such as capsular polysaccharides, and its ability to acquire a wide range of antimicrobial resistance (AMR) mechanisms, including cfiA-encoded metallo-β-lactamases that confer resistance to carbapenems [1]. Although B. fragilis has traditionally been considered predictably susceptible to anti-anaerobic agents, recent data suggest a global rise in AMR [2]. Despite this, AST for anaerobes remains uncommon due to the complexity and cost of conventional methods. EUCAST introduced a standardized disc diffusion method for selected anaerobes to address these challenges in 2022. However, research on methods for the direct identification (ID) and antimicrobial susceptibility profiling of anaerobic pathogens in positive blood cultures (BC) remains limited.

In this study, we aimed to establish a rapid, cost-effective, and practical method for the direct ID and AST of B. fragilis from positive BCs. The method was evaluated using spiked positive BC samples, and its performance was compared with that of the standard EUCAST culture-based AST.

A total of 92 BC bottles were processed in the experiments, as detailed in the following section, comprising 43 previously isolated and stored B. fragilis clinical strains tested in duplicate, and a single Bacteroides fragilis ATCC 25285 reference strain tested in six replicates. Five mL defibrinated sheep blood was spiked with one mL of bacterial suspension (final inoculum:1-2x104 CFU/mL) and transferred into BC bottles (BACTEC™ Lytic/10 Anaerobic F culture vials, BD, New Jersey). Following a positive signal, within 30 min of the blood culture bottles flagging positive, 1–2 drops of blood were inoculated onto Fastidious Anaerobe Agar (FAA) using a 0.8 × 38 mm needle. The plates were incubated in an anaerobic atmosphere for 24 h, and the AST was performed the next day according to the EUCAST standardized disk diffusion method for anaerobic bacteria (standard AST, sAST [reference method]) [3]. Another 125 μL of blood from BC bottles was spread over the FAA by swabbing. Antibiotic disks were applied, and the plates were incubated for 16–20 h in an anaerobic atmosphere (direct disk diffusion, dDD). Also, 1.5 ml of the BC broth was transferred to a microcentrifuge tube. After centrifugation for 5 min at 14,000 rpm, the supernatant was discarded and the pellet was suspended in 1 ml of phosphate-buffered saline (PBS, pH∼7.4). The centrifugation step was repeated, and two spots from the pellet were applied on the MALDI-TOF MS (Biotyper 4.1, Bruker Daltonics, GmbH, Bremen, Germany) target plate. Formic acid was applied to one of the spots. Then, the HCCA matrix was added, and ID analysis was initiated. Subsequently, from the pellet, a suspension of 1 McFarland standard was prepared in 0.85 % saline, and disk diffusion was performed according to routine procedure (pellet-AST, pAST). All plates with antibiotic disks were incubated for 16–20 h under anaerobic conditions. Anaerobic atmosphere was created with BD GaspakTM EZ anaerobe gas generating sachets placed in anaerobic jars (Oxoid AnaeroJar, Thermo Fisher Scientific, United States). Clostridium perfringens DSM 25589 strain with a metronidazole 5 μg disk was used to monitor the anaerobic atmosphere. A flow chart for the procedures used in the study is presented in Fig. S1.

Rapid ID from the pellet correctly identified 93.5 % (86/92) of isolates to the genus level and 85.9 % (79/92) to the species level. The ID rate increased to 95.6 % (88/92) at the genus level but remained at 85.9 % (79/92) at the species level by the addition of formic acid. It is essential to note that minor variations in the amount of biomass spotted on the MALDI-TOF target plate may have impacted identification performance, as the Bruker Biotyper is sensitive to target quantity [4]. Susceptibility of the isolates to meropenem, piperacillin-tazobactam, clindamycin, and metronidazole was 55.8 %, 74.4 %, 67.4 %, and 97.7 %, respectively, according to sAST. pAST showed an overall categorical agreement (CA) of 98.5 % with sAST. CA rates for meropenem, piperacillin-tazobactam, clindamycin, and metronidazole were 98.8 %, 96.5 %, 98.8 %, and 100 %, respectively. We recorded three major errors (ME) for piperacillin-tazobactam, one ME for meropenem, and one very major error (VME) for clindamycin. The overall CA between dDD and sAST was 69.5 %. CA rates were 63.9 % for meropenem, 48.8 % for piperacillin-tazobactam, 90.7 % for clindamycin, and 74.4 % for metronidazole. The high number of MEs (n = 99) and VMEs (n = 6) was observed. QC zone diameter ranges have not been established for the dDD method. For the ATCC strain tested by the dDD method, the zone diameters were below the lower limits of the sAST QC ranges in 79.2 % (19/24) of occasions, while the remaining five measurements fell within the QC ranges. The CAs and errors for all antibiotics are listed in Table 1. Distribution of zone diameters measured in pAST and dDD is shown in Fig. 1, Fig. 2. No distinct zone diameter cut-off was identified to discriminate between susceptible and resistant isolates in dDD (Fig. 2).

Timely microbiological diagnosis and the prompt availability of AST results are crucial in improving clinical outcomes, especially in patients with bloodstream infections (BSIs) [5]. In the traditional workflow, bacteria isolated from BCs are subcultured on solid media to enable ID and AST. This process typically takes 48–72 h to yield species-level ID and AST results, which can delay critical treatment decisions. Recently, several biochemical, MALDI-TOF MS-based, and molecular approaches have emerged that allow for the rapid detection of bloodstream pathogens directly from BCs [6,7]. Moreover, EUCAST and CLSI have developed rapid antimicrobial susceptibility testing protocols that can be performed directly from positive BC bottles for the most common aerobic pathogens [8,9].

Several studies evaluating direct AST methods for aerobic bacteria, including Enterobacterales and non-fermenters, have reported high categorical agreement (CA) rates between direct disk diffusion, automated systems (Vitek 2, BD Phoenix), and reference methods, generally ranging from 95 % to 99 % [[10], [11], [12], [13], [14], [15], [16], [17], [18]]. MEs and VMEs were consistently below 2 % across methods. Positive predictive values of ≥96 % were also reported for most organism–antibiotic pairs, ranging from 86 % to 100 % [17]. Notably, no comparable poor performance with β-lactam/β-lactamase inhibitor combinations or carbapenems was observed in these aerobic studies, suggesting that the lower CA values seen in the dDD part of our work may be specific to anaerobes. In previous studies using bacterial pellets for ID, the rates of correct ID ranged from 79 % to 100 % [11,[13], [14], [15]]. The differences in ID and AST results among the studies may be attributed to several factors, including the spectrum of microorganisms tested, the reference standard used (e.g., EUCAST vs. CLSI), variations in media or testing conditions, inter-laboratory differences in protocols and technical expertise, as well as differences in sample size and study design.

To the best of our knowledge, our study is the first attempt to perform AST directly from positive BCs of an anaerobic pathogen. The time to ID and AST report was shortened by 24 h with the rapid methods used in this study. Although direct testing methods such as dDD have demonstrated high CAs and predictive values for aerobic bacteria in previous studies, our results suggest that similar success has not yet been achieved for anaerobic pathogens. While direct AST methods have been optimized and validated extensively for aerobes, the application of such rapid techniques to anaerobes remains limited and lacks standardization. The low level of agreement observed between dDD and sAST in our study underscores the need for further research to develop and validate standardized protocols specifically tailored to anaerobic organisms. According to EUCAST, the standard anaerobic disk diffusion method uses a 1.0 McFarland inoculum, whereas the standard aerobic method employs a 0.5 McFarland suspension, and the EUCAST RAST method for aerobic bacteria utilizes a 100–150 μl inoculum. In this study, a higher inoculum than that recommended for the aerobic RAST method was not applied; however, future studies should investigate whether increasing the inoculum volume could improve the performance of the anaerobic dDD method. On the other hand, the performance of rapid ID and pAST was highly promising, supporting the feasibility of incorporating these methods into routine workflows for anaerobic bacteria. In our study, rapid ID from bacterial pellets correctly identified 93.5 % of isolates at the genus level and 85.9 % at the species level; genus-level accuracy increased to 95.6 % with the addition of formic acid. pAST showed excellent concordance with sAST, with an overall categorical agreement of 98.5 %, and agreement rates above 96 % for all antibiotics tested. These findings indicate that the pellet-based approach enables both accurate and substantially faster results compared to conventional methods. As shortening the time to reliable AST results remains a critical objective in clinical practice, our findings highlight the potential of pAST as a rapid and accurate alternative for anaerobic pathogens. Nonetheless, larger-scale studies are needed to validate these results and support their broader implementation.

This study has several limitations. First, the sample size is small, and the research was conducted at a single center, which may limit the generalizability of the findings. Second, this study was based on spiked samples rather than prospective clinical cultures, which may not fully reflect the method's performance under real-world conditions. Nevertheless, this design was appropriate for an initial feasibility evaluation. Third, only one metronidazole-resistant isolate was included, restricting the interpretation of results. However, due to the low prevalence of metronidazole resistance in our country, the inclusion of additional resistant isolates was not feasible [19,20]. In addition, the study did not evaluate the clinical impact of rapid ID and AST of B. fragilis bloodstream infections on patient outcomes. Finally, this approach may have limitations in polymicrobial samples, where the presence of multiple anaerobic species could interfere with accurate identification or susceptibility interpretation. Such situations are common in complex anaerobic infections, and this factor should be carefully considered when applying direct methods in routine practice.

In conclusion, the rapid ID and AST protocol (pAST) based on multistep centrifugation is rapid, easy to perform, and cost-effective. Reporting ID and AST results within 24 h after blood culture positivity may improve patient management by early and appropriate antimicrobial treatment. Further validation using a larger sample size and evaluating the impact on patient mortality and morbidity would provide valuable insights.