Single-donor apheresis platelet samples were collected from 6 healthy voluntary blood donors at the Chengdu Blood Center between May 2022 and November 2023, comprising a total of 12 therapeutic units. The study was approved by the Ethics Review Committee of the Institute of Blood Transfusion, Chinese Academy of Medical Sciences (Ethics Approval Number: 2021011). For analysis, we defined one analytical sample per donor (n = 6). From each analytical sample, 10 mL was allocated to metagenomic sequencing and 100 mL to long-term culturomics. Processing was conducted in three laboratory batches (two analytical samples per batch), with one PBS-only culture control included in each batch and handled in parallel using the same culture and day-30 metagenomic workflow.

Collections and sampling followed the blood center’s routine, industry-standard antisepsis and sterility procedures. Phlebotomy sites were prepared using 2% chlorhexidine gluconate in 70% isopropyl alcohol, with a friction scrub for 30–60 s covering an area >5–7 cm in diameter, followed by 30 s of air-drying; no-touch venipuncture was performed, and the site was re-prepped if repalpated. For donors with chlorhexidine sensitivity, 10% povidone–iodine (alcohol-based two-step) was used as an alternative per institutional policy. We diverted the initial 20–30 mL of blood into an integrated first-diversion pouch before the main collection. All sampling was performed via sterile ports using new sterile needles/syringes, after wiping the port membrane with 70% isopropyl alcohol and allowing it to dry. Sample aliquots were processed in a biological safety cabinet with dedicated, pre-cleaned instruments.

To monitor background and process-related contamination in this low-biomass setting, we included sterile PBS culture controls in parallel with platelet samples. For each culture run, 5 mL of sterile PBS was inoculated into the same aerobic and anaerobic fluid nutrient medium used for samples and incubated under identical conditions and schedule (temperature, atmosphere, agitation, transfer scheme, and total 30-day enrichment). Whenever sample cultures were subcultured to solid media, an aliquot of the corresponding control broth was plated onto Columbia blood agar to serve as a plating control. At day 30, the remaining volumes of control broths were subjected to DNA extraction, library preparation, and shotgun metagenomic sequencing alongside study samples, using the identical host-depletion, bioinformatic profiling, and reporting thresholds.

Using liquid culture media plated on blood agar plates, single bacterial colonies were isolated. DNA was extracted from these single colonies, and PCR amplification was performed. Electrophoresis revealed an approximately 1.4 kb DNA band. Subsequent Sanger sequencing and BLAST analysis confirmed the isolation of a total of 90 bacterial strains from blood agar plates. The bacterial species identified from the six samples are shown in Figure 2 . No colony growth was observed on control plates inoculated with sterile PBS, and nucleic acid extracted from the negative control samples at the end of the culture period showed “Too Low” concentrations as measured by Qubit, confirming the absence of contamination during the culture process.

The isolated bacteria were classified into 3 phyla, 5 classes, 5 orders, 7 families, 9 genera, and 23 species ( Figure 3 ). At the phylum level, Firmicutes dominated, with the highest number of strains isolated. The dominant genera included Bacillus and Brevibacillus, with specific species such as B. brevis and Bacillus sp. Y1 showing high relative abundances (13.3% and 15.6%, respectively). Among Actinobacteria, Cutibacterium acnes represented 14.4% of the total abundance. Proteobacteria was primarily represented by Brevundimonas aurantiaca, accounting for 16.7% of the relative abundance.

At the genus level, Brevundimonas, Brevibacillus, and Cutibacterium were the dominant genera, collectively accounting for 44.4% of the total abundance. Notably, Brevibacillus brevis, B. aurantiaca, and C. acnes were repeatedly isolated in at least five samples, suggesting that these species possess strong competitive advantages under the culture conditions.

High-throughput sequencing using the Illumina NovaSeq 6000 platform generated 646.14 GB of raw data and 464.42 GB of clean data. The sequencing quality metrics were as follows: Q20 > 90%, Q30 > 85%, error rate = 0.03%, and an average GC content of 45.70%. Analysis of the metagenomic data from six platelet samples revealed the relative abundances of microbial taxa at the phylum, class, order, genus, and species levels.

Across samples, within-sample (α) diversity showed high richness but variable evenness (Observed richness 849–1083 genera; Shannon H′ 1.23–3.63; Simpson 1−D 0.307–0.925; Pielou’s J 0.176–0.521). Sample 5 exhibited a “high-richness/low-evenness” pattern (Shannon H′=1.23; Pielou’s J = 0.176), indicating dominance by one or a few taxa despite many detected genera, whereas samples 2–3 showed both high richness and high evenness (Shannon H′ ≥ 3.37; Simpson 1−D ≥ 0.892). Between-sample (β) diversity was substantial: mean pairwise dissimilarity was 0.73 (Bray–Curtis; range 0.26–0.97), 69.25 (Aitchison/CLR Euclidean; 42.39–101.07), and 0.24 (Jaccard; 0.14–0.32). PCoA based on Bray–Curtis explained 53.5% and 34.3% of variance on PC1 and PC2, respectively (Aitchison: 59.3%/21.6%; Jaccard: 46.7%/26.3%), yielding dispersed ordinations without tight clustering—consistent with heterogeneous community structure across donors. Distance matrices were moderately correlated for Bray versus Aitchison (Spearman ρ=0.67, p≈0.033) and weaker for comparisons involving Jaccard (ρ≤0.58), reflecting the difference between abundance-based and presence/absence metrics.

No colony growth was observed on blood agar for the PBS culture control during the 30-day enrichment. The PBS control processed in parallel failed library qualification: library qPCR was negative (Ct undetermined), consistent with insufficient yield/library-preparation failure. Accordingly, this control was considered non-informative for metagenomic contamination assessment and was excluded from sequencing and downstream taxonomic analyses.

Unlike refrigerated red blood cells or frozen plasma, platelet products require room-temperature storage under continuous agitation to maintain functionality, creating favorable environments for bacterial proliferation. The elevated risk of bacterial contamination in platelet products has been well-documented in hemovigilance systems from Europe and North America, where routine bacterial screening is now standardized to mitigate transfusion-associated risks (Levy et al., 2018).In contrast, China has yet to implement systematic bacterial culturing screening protocols for platelet products (National Health Commission of the People’s Republic of China, 2019). To mitigate adverse reactions during platelet transfusion, many physicians in China administer hydrocortisone or dexamethasone to patients either before or during the transfusion to prevent fever and other symptoms (Dazhang et al., 2002). While this practice may reduce reported rate of transfusion-related bacterial infections, and could potentially lead to underdiagnosis and delayed therapeutic interventions in clinical practice (Delaney et al., 2016).

Our data demonstrate a diverse bacterial consortium within platelet samples, raising serious concerns about their potential impact on Chinese blood recipients, particularly those immunocompromised populations. Culturomic analyses confirmed the presence of S. epidermidis (Becker et al., 2014; Vuong and Otto, 2002), E. coli (Russo and Johnson, 2003; Pitout and Laupland, 2008), and S. aureus (Chambers and Deleo, 2009; Tong et al., 2015), as well as C. acnes (Achermann et al., 2014; Vollmer et al., 2014) detected via metagenomics, were all well-documented pathogens associated with bloodstream infections (BSIs). Notably, emerging opportunistic pathogens such as D. tsuruhatensis (linked to biofilm-mediated catheter-associated BSIs (Ranc et al., 2018).) and Brevundimonas spp. (e.g., B. vesicularis and B. diminuta, implicated in immunocompromised-host infections (Ryan and Pembroke, 2018)) were also detected. Other bacteria identified, including R. insidiosa (persistence in medical equipment (Ruggieri et al., 2024; Alnimr, 2023) and carbapenem-resistant (Fang et al., 2019)), B. cereus (associated with hematologic malignancy-related BSIs (Bottone, 2010; Gaur et al., 2001)) are generally considered environmental microorganisms but may pose infection risks under specific conditions. Some bacteria detected in this study, such as B. sp. AS-1 and B. sp. Tri-49 from the genus Bosea, B. brevis, B. velezensis, B. coreaensis, B. licheniformis, and B. thuringiensis from the genus Bacillus, and C. vibrioides from the genus Caulobacter, currently lack direct evidence of pathogenicity, however, phylogenetically related strains [e.g., B. thiooxidans in catheter infections (Skipper et al., 2020)] and clinical reports suggest they may occasionally act as opportunistic pathogens (Justesen et al., 2007; Moore and Gitai, 2020).

These findings emphasize the vulnerability of high-risk cohorts—immunosuppressed patients (e.g., post-transplant or chemotherapy recipients), neonates, and the elderly—to severe outcomes from platelet-transmitted BSIs. Comprehensive risk stratification and vigilant monitoring are imperative to mitigate morbidity in these populations.

The identification of diverse microbial communities in platelet products not only underscores contamination risks but also challenges the historical paradigm of blood sterility. Over the past two decades, advancements in microscopy and bioinformatics technologies have significantly advanced our understanding of microbial presence in the bloodstream (Whittle et al., 2018; Li et al., 2018; Bhattacharyya et al., 2017; Paisse et al., 2016; Potgieter et al., 2015; Dinakaran et al., 2014; Rajendhran et al., 2013; Nikkari et al., 2001). For instance, Potgieter et al. visualized bacteria associated with blood cells in various blood products using electron microscopy (Potgieter et al., 2015), while Damgaard et al. isolated viable bacteria from 62% of healthy donor samples (Damgaard et al., 2015). Currently, evidence regarding microorganisms detected in human blood is not consistent. A large-scale sequencing study of 9,770 healthy individuals found no shared, stable core blood microbiome, but did confirm the detection of bacteria in 16% of individuals, with nearly 5% harboring the same bacterial species (Tan et al., 2023). In healthy individuals, microorganisms detected in blood can be interpreted along two clinically distinct patterns: (i) bona fide pathogens may transiently enter the bloodstream but are rapidly cleared by innate and adaptive immunity, leaving no sustained clinical syndrome; this phenomenon is well documented for viridans group streptococci and other mucosa-derived taxa, with bacteremia typically peaking within minutes and declining thereafter (Martins et al., 2024; Wilson et al., 2021); (ii) some studies have reported viable bacteria at very low levels in freshly drawn blood from self-reported healthy donors (Damgaard et al., 2015)—potentially consistent with low-metabolic or VBNC (viable but non-culturable) -like states (Oliver, 2010). The existence and biological significance of these so-called ‘co-existing’ bacteria in blood remain controversial and have therefore attracted widespread attention (Païssé et al., 2016; Damgaard et al., 2015, D’aquila et al., 2021).

In this context, our inference about a putative endogenous co-existing bacteria is grounded in a conservative, clinically anchored interpretation of low-abundance signals. First, transient translocation of bona fide commensals and pathobionts can occur and is typically self-limited in immunocompetent hosts (Lehtiniemi et al., 2005; Whittle et al., 2018). This aligns with our findings of species linked to these body sites, including common skin commensals(C. acnes, S. epidermidis, B. velezensis, B. cereus, B. licheniformis, and S. aureus (Fournière et al., 2020; Dréno et al., 2020)), gut microbes (B. thuringiensis and E. coli (Brenchley and Douek, 2012) (Di Marco Barros et al., 2022)), and oral commensals (Brevundimonas, Paenibacillus, Delftia, and Staphylococcus (Sarkar et al., 2017; Nasidze et al., 2009; Yamashita and Takeshita, 2017; Horliana et al., 2014)). Once in the bloodstream, these bacteria may enter a dormant state, consistent with the concept of “viable but non-culturable” (VBNC) bacteria, which maintain metabolically active but cease division, regaining activity and proliferation under favorable condition (Wagley et al., 2021; Mu et al., 2021; Kell and Pretorius, 2015; Potgieter et al., 2015). However, DNA-only metagenomics cannot by itself adjudicate viability or a viable-but-non-culturable (VBNC) state in vivo, and such claims require orthogonal evidence (e.g., culture resuscitation, rRNA/metatranscriptomic signals, or chemical exclusion of dead-cell DNA). Accordingly, we regard ‘healthy bacteremia’ as a speculative concept and interpret our low-abundance findings with caution.

The possible existence of blood co-existing bacteria has raised questions about their functional roles in the human bloodstream. Certain blood commensal bacteria, such as C. acnes and S. epidermidis, can evade immune detection and establish long-term colonization through biofilm formation, or finely tuned metabolic and virulence adaptations (Becker et al., 2014) (Achermann et al., 2014) (Lê-Bury et al., 2024), highlighting their potential adaptive mechanisms for survival in the bloodstream and warranting further investigation. In interpreting low-abundance blood detections, host context is pivotal. Recent case evidence shows that immunocompromised patients may harbor invasive, hyper-aerotolerant Campylobacter jejuni capable of causing bacteremia, underscoring that barrier-site organisms can exhibit atypical, blood-intruding phenotypes in specific hosts (Zeng et al., 2025). Consistent with prior work on hyper-aerotolerant Campylobacter, enhanced oxygen tolerance may facilitate survival outside microaerophilic niches and increase the likelihood of extraintestinal invasion (Oh et al., 2019). Interestingly, healthy individuals with a lower abundance of Proteobacteria and a higher abundance of Actinobacteria in their blood are reportedly at a reduced risk of type 2 diabetes (Amar et al., 2011), while bacterial 16S rRNA gene concentrations tend to increase with age (Yun et al., 2019),though the precise roles of specific bacteria in health and disease progression remain unclear and warrant further investigation.

Tan et al. (2023) proposed that the blood microbiome is not a stable, resident community, but rather represents a sporadic and transient presence of microorganisms originating from other body sites (Tan et al., 2023). If a core microbial community existed in the bloodstream, targeted purification strategies could potentially be employed in clinical practice to ensure transfusion safety. However, the existence of a heterogeneous and fluctuating microbiome presents a greater challenge to transfusion medicine. Our 30-day extended culturomic approach revealed the microbial diversity within platelet concentrates. In natural environments, certain bacteria exist in oligotrophic states (Ho et al., 2017). While high nutrient concentrations in culture media may have little effect on fast-growing and resilient bacteria, they can harm slow-growing, less adaptive bacteria, even leading to their death (Bartelme et al., 2020). We speculate that the 30-day extended culture period gradually reduced nutrient levels in the media, creating conditions more favorable for the growth and isolation of oligotrophic bacteria (Zhang et al., 2021). In our unpublished study, we explored the temporal dynamics of microbial communities by analyzing bacterial abundance and composition at specific time points (Days 3, 5, and 10). Fast-growing aerobes (e.g., R. insidiosa) peaked by Day 5, whereas anaerobes (C. acnes) thrived throughout storage, reflecting their metabolic adaptations. Additionally, P. spp. and S. maltophilia exhibited significant changes in abundance at different time points, suggesting dynamic shifts in their growth patterns during storage. Such temporal profiling identifies high-risk contaminants, advocating for early screening to preempt pathogen proliferation.

This study integrated and compared culturomics and metagenomics to emphasize their complementary strengths in detecting bacterial contamination in platelet samples. Culturomics isolated viable strains (e.g., B. aurantiaca) for phenotypic analyses, such as antibiotic susceptibility and metabolic profiling, which are critical for understanding pathogenic potential and designing interventions. However, it favors fast-growing or high-abundance microorganisms (Lagier et al., 2015; Browne et al., 2016), often overlooking low-abundance or slow-growing species requiring specific growth conditions. In contrast, metagenomics provided a culture-independent approach, detecting unculturable and rare taxa (e.g., B. sp. DS20) by capturing the genetic diversity and functional potential of microbial communities. This method is particularly suitable for detecting bacteria that are difficult to culture under laboratory conditions, such as obligate intracellular pathogens or extremophiles. Nonetheless, metagenomics cannot confirm viability and faces challenges like reduced sensitivity for low-abundance microbes, taxonomic resolution limitations, and depth biases (Boolchandani et al., 2019). Future Integration of long-read sequencing technologies (e.g., nanopore sequencing (Ying et al., 2022)) could mitigate these gaps, enhancing detection of rare or slow-growing pathogens.

Sequence-based surveys of blood or platelet products are intrinsically vulnerable to reagent-borne DNA (“kitome”), environmental/handling carry-over, well-to-well exchange, and index misassignment, all of which can distort community profiles at low biomass. Prior work has shown that contaminant DNA is ubiquitous in extraction kits and varies by kit/batch, with disproportionate impact on low-biomass shotgun and amplicon data; current best-practice recommendations emphasize rigorous negative controls and conservative interpretation (Salter et al., 2014; Eisenhofer et al., 2019).

What we did in this study. (i) At collection, we followed the blood center’s industry-standard antisepsis and sterility procedures (alcoholic chlorhexidine prep with specified scrub/dry times, first-diversion of the initial 20–30 mL, sterile-port access, no-touch technique). In the laboratory, all aliquots were handled in a Class II biosafety cabinet using dedicated, pre-cleaned instruments and aerosol-resistant tips to limit operator-derived contamination. (ii) Each processing batch included a PBS-only culture control that underwent the same 30-day enrichment and day-30 metagenomic workflow as study samples; these blanks were used to monitor process background.

Known risks we monitored but did not fully quantify. We recognize risks from well-to-well exchange during plate-based extractions and index misassignment (“index hopping”) during multiplexed sequencing. These phenomena can spuriously spread reads across samples and inflate apparent diversity. For low-biomass datasets, we recommend adopting conservative interpretive rules—namely: treat low-abundance taxa plausibly linked to skin or reagents with particular caution; prioritize signals only when (a) they exceed blank maxima, (b) are reproducible across assays/media, or (c) are concordant with clinical/biological context (Eisenhofer et al., 2019; Costello et al., 2018). Although our run design avoided mixing obvious high-biomass libraries with platelet samples, and blanks were interleaved to flag cross-talk, the present dataset was not powered to estimate these effects precisely; future runs will adopt unique dual indexing and tighter plate layouts to further suppress misassignment and well-to-well transfer (Davis et al., 2018).